JP2007249978A - Image measurement method, image measurement apparatus, and image measurement program storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technique for measuring the three-dimensional geometrical information of a plane and the positional information of a point, etc. from an optical flow pattern or a stereographic image. <P>SOLUTION: When defining p<SB>0</SB>, p<SB>1</SB>as the respective positions of an arbitrary point at two different times in a space appeared in an image, when the inside of the space is viewed from a predetermined observation point, and when p<SB>inf</SB>as the position of the point after the lapse of an infinite time, and p<SB>c</SB>as the position of the point at the time, when the plane including the point is overlaid on the observation point, the time at which the direction of the plane and the plane is overlaid on the observation point is obtained, by using a cross-ratio äp<SB>inf</SB>p<SB>0</SB>p<SB>1</SB>p<SB>c</SB>} determined by the four points, p<SB>inf</SB>, p<SB>0</SB>, p<SB>1</SB>and p<SB>c</SB>. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image measurement method for measuring the position and orientation in a space from points and surfaces appearing in an image, an image measurement apparatus for implementing the image measurement method, and an image measurement program for realizing the image measurement. Is stored in the image measurement program storage medium.

  In order to move mobile robots, automobiles, airplanes, and the like in accordance with the environment, it is necessary to measure the environment three-dimensionally from a moving image such as a camera. Let us consider what kind of three-dimensional measurement is being performed by humans when visually landing and walking.

  FIG. 1 shows an optical flow pattern reflected in the retina of the pilot. From this pattern, the pilot perceives the slope of the runway (three-dimensional azimuth) and “how many seconds it will take to reach the runway if it continues as it is” to land the airplane accurately. In other words, the pilot is landing from this pattern by measuring the three-dimensional orientation of the plane (runway) and “time to cross the plane”.

  Consider the case where we walk in the hallway. When walking in the direction of hitting the hallway wall, the optical flow pattern described above is generated on the retina, and the time taken to cross the wall from that pattern (time to collide with the wall) is measured, and the wall is measured simultaneously The avoidance movement is made from the three-dimensional direction in the direction to avoid it. On the other hand, when walking parallel to the wall, it is measured that it will not collide indefinitely (that is, it will collide after an infinite time has passed), and it will continue walking in that direction. In this way, even a winding corridor can be walked with accurate avoidance of the walls. In addition, when going inside the office, it is possible to walk while avoiding “objects composed of planes” such as whiteboards, desks, and lokers. Furthermore, when driving a car, the same “three-dimensional measurement of a plane” is performed to run on a highway or put in a garage.

  In this way, we accurately measure the three-dimensional geometrical information (the three-dimensional orientation of the plane and the time until it crosses the plane) of an object composed of a plane-there are many such objects. Movement is possible. Even a curved object can be spatially recognized by measuring three-dimensional geometric information of a “plane group in contact with it”.

  If such “three-dimensional geometric information of a plane” can be measured from an image, it becomes possible to move a mobile robot, a car, an airplane, etc. while adapting to the environment or avoiding an obstacle.

  For each velocity element of the optical flow pattern in Fig. 1, that is, the motion in the local region (local motion) in Fig. 1, a technique for measuring it from the moving image has been reported so far. Kaihei 05-16595, JP-A 05-165957, JP-A 06-044364, JP-A 09-081369; “two-dimensional correlation and convolution one-dimensionally along the ρ coordinate of the Hough plane ”Kawakami and Okamoto, IEICE Technical Report, vol. IE 96-19, pp. 31-38, 1996; ami, S. and Okamoto, H., Vision Research, vol. 36, pp. 117-147, 1996).

  However, a method for integrating this optical flow pattern and measuring “three-dimensional geometric information of a plane (three-dimensional orientation of the plane, time to cross the plane, shortest distance to the plane)” has been reported so far. Not.

  In addition, a technique for measuring three-dimensional geometric information (such as their three-dimensional orientation and the shortest distance to them) of a straight line or a cylinder in a space from a moving image has been reported so far (Japanese Patent Publication No. 03-52106). , Japanese Patent Publication No. 06-14356, Japanese Patent Publication No. 06-14355, Japanese Patent Publication No. 06-10603, Japanese Patent Laid-Open No. 02-816037; “Measurement of three-dimensional azimuth and distance of line segment by spherical mapping,” Inamoto , Computer Vision Study Group, vol. 45-2, pp. 1-8, 1986; “Realization of monocular stereoscopic vision with bird mimics,” Science Asahi, June issue, pp. 28-331987; “Measurement in three dimensions” by motion stereo and spatial mapping, "Morita, T. et al. et al., CVPR, pp. 422-428, 1989; “Motion stereo vision system,” Inamoto, Y. et al., Proceeding of '91 ISART, pp. 239-246, 1991; Nuclear power generation facility work robot development commissioned research results report (extreme work robot technology research association) ”section 4.2.2.1).

  However, no method for measuring the three-dimensional geometric information of a plane has been reported so far.

  It is an object of the present invention to provide a technique for measuring plane three-dimensional geometric information, point position information, and the like from an image such as an optical flow pattern. Note that the measurement of the three-dimensional geometric information includes measurement of the shortest distance to the plane, as will be described later. It is another object of the present invention to provide a technique for measuring planar three-dimensional geometric information from a stereo image.

The first image measurement method among the image measurement methods of the present invention is an arbitrary measurement in the measurement space that appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. The measurement positions of the points at two different measurement times are _denoted by p ₀ and p ₁ , respectively, and the direction of the measurement point relative to the observation point is the same as the moving direction v between the two measurement times. P _{inf is} the position of the measurement point after the infinite time has elapsed in the planned movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued. when measuring plane including the measurement point in the continuous state is a position of the measurement points in the overlay time overlapping the observation point and p _c,
The four positions _p inf measurement _point, p _0, p 1, using the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations determined by _{p c,} the orientation of the measurement plane, And / or it is characterized in that a physical quantity indicating the superposition time at which the measurement plane overlaps the observation point is obtained.

Here, in the first image measurement method, the compound ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent calculation, the measurement point, two measurement in the two measurement time Instead of the positions p ₀ and p ₁ , the measurement point p ₀ of the measurement point at one of the two measurement times and the two measurement positions p of the measurement point at the two measurement times Computation using motion parallax τ, which is the difference in position between ₀ and p ₁ , is included.

In the first image measurement method, as a physical quantity indicating the superposition time, a time between one measurement time of the two measurement times and the superposition time is t _c , and the two measurement times are measured. When the time between them is Δt,
_{n t c} = t _c / _Δt
The normalized time _n t _c expressed by the following is adopted, and the normalized time _n t _c is
_{n t c = {p inf p} 0 p 1 p c}
Alternatively, the relational expression may be obtained by using a relational expression or a relational expression equivalent to the relational expression.

Further, in the first image measurement method, the measurement point, the process for obtaining the polar corresponding to the position p _c by polar transformation the position p _c in the superimposed time, present in the measuring space By executing a plurality of measurement points and obtaining the intersections between the polar lines when the plurality of polar lines obtained by these processes are drawn in the polar drawing space, it corresponds to the plurality of polar lines intersecting at the intersections. An orientation of a measurement plane including a plurality of measurement points and / or a physical quantity indicating a superposition time at which the measurement plane overlaps the observation point may be obtained.

Further, the first image measurement method, said by measurement points appearing on the image when the one having an intensity information, the measurement point, to polar transformation the position p _c in the overlapped time A polar line corresponding to the measurement point is obtained, and each point on the trajectory of the polar line when the obtained polar line is drawn in the polar drawing space corresponds to the intensity of the measurement point corresponding to the polar line. The process of voting a value is executed for a plurality of measurement points existing in the measurement space, and the voting of the maximum point is obtained by obtaining a maximum point at which the value by voting during the execution of these processes becomes a maximum value. The orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines that participated in and / or the physical quantity indicating the superposition time at which the measurement plane overlaps the observation point may be obtained.

Furthermore, the first image measurement method, the measurement by the measurement points appearing on the image when the one having an intensity information, converts the position p _c in the superimposed time of the measurement point electrode The polar line corresponding to the point is obtained, the response intensity corresponding to the motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point is obtained, and the obtained polar line The process of voting the response intensity corresponding to the motion parallax τ of the measurement point corresponding to the polar line to each point on the trajectory of the polar line when This is executed for a plurality of measurement points existing in the process, and by obtaining a maximum point where the value obtained by voting during the execution of these processes becomes a maximum value, it corresponds to a plurality of polar lines participating in the vote of the maximum point. Includes multiple measurement points Orientation of the measurement plane, and / or it may be one that obtains a physical quantity indicative of the superimposed time the measurement plane overlaps the observation point.
a
Furthermore, in the first image measurement method, when the seek position p _c, a physical quantity which indicates the superposition time, the measurement point, the respective measurement position p ₀ in the two measurement _time, p ₁ or they Instead of the two measurement positions p ₀ , p ₁ , the measurement point p ₀ of the measurement point at one of the two measurement times and the two measurement positions of the measurement point at the two measurement times Knowing the motion parallax τ between p ₀ and p ₁ and the position p _{inf of the} measurement point after the infinite time in the movement continuation state, the cross ratio {p _inf p ₀ p ₁ p _c } or by using a plurality ratio equivalent operation may be obtained position p _c in the superimposed time of the measurement point.

Furthermore, the first image measurement method includes:
A first step of setting a physical quantity indicating the superposition time as a parameter;
The physical quantity indicating the superposition time set in the first step and the measurement positions p ₀ , p _{1 at the} two measurement times of the measurement points, or these two measurement positions p ₀ , p ₁ alternative, the measurement point, movement between the two measurement positions in one of the measurement time p ₀ and the measurement point of the measurement time, the two measurement positions p ₀ in the two measurement _time, p ₁ How to and parallax tau, and a position _{p inf} of after infinite time in the moving continuous state, using the double ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent calculation, the of the measuring points a second step of obtaining a position p _c in the overlay time,
The measurement point, and a third step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c in the superimposed time,
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing the value of the parameter in the first step, and then
A plurality of polar lines intersecting at the intersections are obtained by obtaining intersections between the polar lines when the plurality of polar lines obtained while repeating the first to third steps are repeated in the polar drawing space. The fourth step may be performed in which the orientation of a measurement plane including a plurality of measurement points corresponding to and / or a physical quantity indicating the superposition time at which the measurement plane overlaps the observation point is obtained.

In this case, the measurement point appearing on the image has intensity information,
In the third step, the polar line is obtained, and the intensity of the measurement point corresponding to the polar line at each point on the trajectory of the polar line when the obtained polar line is drawn in the polar drawing space. Voting for the value corresponding to
Instead of obtaining the intersection in the fourth step, the voting of the maximum point is obtained by obtaining a maximum point at which the value by voting becomes a maximum value while repeating the first to third steps a plurality of times. Preferably, it is a step of obtaining a physical quantity that indicates the orientation of a measurement plane including a plurality of measurement points corresponding to a plurality of polar lines that participated in and / or an overlapping time at which the measurement plane overlaps the observation point, or ,
Measurement points appearing on the image have intensity information,
A fifth step of setting a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a second parameter;
In the second step, the physical quantity indicating the weight time set in the first step, the measurement position p _{0 at} one measurement time of the two measurement times of the measurement point, and the using the the motion parallax τ set in the fifth step, the position p _inf after infinite time has elapsed in the mobile continuity state of the measurement point, determining the position p _c in the superimposed time of the measurement point And
When the third step obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained polar line in the polar drawing space Voting the response intensity corresponding to the motion parallax τ of the measurement point corresponding to the polar line for each point on the trajectory of the polar line,
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter obtains the maximum point where the value obtained by voting becomes the maximum value while repeating the first, fifth, second and third steps a plurality of times instead of obtaining the intersection. Thus, the orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines participating in the voting of the local maximum point and / or the physical quantity that indicates the superposition time at which the measurement plane overlaps the observation point is obtained. It is also a preferred embodiment that it is a step.

Wherein the third step is preferably a step of obtaining a polar depicted as great circle on a sphere by converting the position p _c poles, further, the third step is, the location depicted as a great circle on a sphere by converting p _c poles, more preferably a step of obtaining a more polar, which is projected inside the circle on the plane. Alternatively, the third step may be a step of obtaining a polar line drawn as a straight line on a plane by subjecting the position _pc to polar conversion.

In addition, the first image measurement method includes:
A first step of setting a position p _inf of the measurement point after an infinite time in the moving continuous state by setting the moving direction v as the first parameter,
A second step of setting a physical quantity indicating the superposition time as a second parameter;
The position p _inf set in the first step, the physical quantity indicating the superposition time set in the second step, and the measurement positions p ₀ , p at the two measurement times of the measurement point ₁ or, alternative to these two measurement positions p _0, p _1, of the measurement point, the two one measurement position p ₀ and the measurement point in the measurement time of the measurement time, in the two measurement time and a motion parallax τ between two measuring positions _p 0, _{p 1} if and using equivalent operations and the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio, the superposition of the measuring points a third step of obtaining a position p _c at time,
The measurement point, and a fourth step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c in the superimposed time,
For the plurality of measurement points in the measurement space, the third step and the fourth step of the first to fourth steps are changed to values of the first parameter and the second parameter, respectively. Are repeated several times while changing in the first step and the second step, respectively,
A plurality of polar lines obtained while repeating the first to fourth steps a plurality of times are respectively converted into corresponding polar line drawing spaces among a plurality of polar line drawing spaces according to the value of the first parameter. A pole corresponding to the true movement direction relative to the observation point of the measurement point based on information on the number of polar lines intersecting at the intersection, where intersections between polar lines at the time of drawing are obtained for each polar line drawing space A measurement including a plurality of measurement points corresponding to a plurality of polar lines intersecting at intersections obtained for the polar drawing space corresponding to the true movement direction while obtaining the true movement direction by selecting a line drawing space The fifth step of obtaining the physical quantity indicating the orientation of the plane and / or the superimposition time at which the measurement plane overlaps the observation point may be executed.

In that case, the measurement point appearing on the image has intensity information,
In the fourth step, the polar line is obtained, and at the respective points on the trajectory of the polar line when the obtained polar line is drawn in the polar drawing space corresponding to the polar line, Voting a value corresponding to the intensity of the corresponding measurement point,
Instead of obtaining the intersection in the fifth step, a maximum point at which the value by voting becomes a maximum value while repeating the first to fourth steps is determined for each polar drawing space. The true movement direction is obtained by selecting the polar drawing space corresponding to the true movement direction based on the local maximum information at the local maximum point, and the polar drawing space corresponding to the true movement direction is obtained. Determining the orientation of a measurement plane including a plurality of measurement points corresponding to a plurality of polar lines participating in voting of the maximum points and / or a physical quantity indicating the superposition time at which the measurement plane overlaps the observation point Preferably, or
Measurement points appearing on the image have intensity information,
A sixth step of setting a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter;
In the third step, the position p _inf set in the first step, the physical quantity indicating the weight time set in the second step, and the two measurement times of the measurement point are set. a measurement position p ₀ in the measurement time of one among the sixth using said motion parallax τ set in step, a step of determining the position p _c in the superimposed time of the measurement points,
When the fourth step obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained polar line in the polar drawing space Voting a response intensity corresponding to the motion parallax τ of a measurement point corresponding to the polar line for each point on the trajectory of the polar line,
For each of a plurality of measurement points in the measurement space, the third step and the fourth step are changed, and the values of the parameters are changed in the first step, the second step, and the sixth step. While repeating several times,
The fifth step executed thereafter is not the determination of the intersection, but the maximum value obtained by voting while repeating the first, second, sixth, third and fourth steps a plurality of times. A point is obtained for each polar drawing space, and the true moving direction is obtained by selecting the polar drawing space corresponding to the true moving direction based on the local maximum information at the local maximum point, and the true moving direction is obtained. The orientation of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the maximum points obtained for the polar drawing space corresponding to the moving direction, and / or the measurement plane as the observation point It is also a preferred aspect that it is a step of obtaining a physical quantity that indicates the overlapping superposition time.

The second image measurement method of the image measurement methods of the present invention is an arbitrary image in the measurement space that appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. P ₀ , p ₁ , and the moving direction of the measurement point relative to the observation point v between the two measurement times v, The movement of the measurement point relative to the observation point in the same direction as the movement direction v between the two measurement times and at the same speed as the movement speed between the two measurement times remains unchanged. the measurement points in the overlay time position p _inf of the measurement points after an infinite time in the movement continues while scheduled shall be _continued, measuring plane including the measurement point in the movement continuation state overlaps the observation point When the position of the p _c, the orientation of the measurement plane and the n _s,
Four positions _p inf of the measurement _points, and p _0, p 1, _p cross ratios determined by _c and the equivalent operations _{{p inf p 0 p 1 p} c} or plurality ratio, orientation _{n s} of the measurement plane Using the inner product ( _ns / v) with the moving direction v, the measurement plane at the measurement time of one of the two measurement times from the observation point azimuth _ns and / or the observation point It is characterized in that a physical quantity indicating the shortest distance is obtained.

Here, the in the second image measuring method, wherein the cross-ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent calculation, the measurement point, the two measurement at two measurement times Instead of the positions p ₀ and p ₁ , the measurement position p ₀ of the measurement point at one of the two measurement times and the two measurement positions p of the measurement point at the two measurement times Computation using motion parallax τ, which is the difference in position between ₀ and p ₁ , is included.

In the second image measurement method, as a physical quantity using the shortest distance as an index, the shortest distance between the observation point and the measurement plane at one of the two measurement times is represented as d _s , The time between one measurement time of the two measurement times and the weight time is t _c , and the relative movement distance of the measurement point with respect to the observation point between the two measurement times is Δx, where the time between the two measurement times is Δt,
_n d _s = d _s / Δx
The standardized shortest distance _n d _s expressed by the following formula is adopted, and the standardized shortest distance _n d _{s
n t c} = t _c / _Δt
Using the normalized time _n t _c expressed by: and the inner product (n _s · v),
_{n d s = n t c (} n s · v)
It is preferable to use the following relational expression.

The second image measurement method is:
A first step of setting a physical quantity indicating the shortest distance as a first parameter;
A second step of setting the inner product ( _ns / v) as a second parameter;
The physical quantity that is set in the first step and that indicates the shortest distance, the inner product ( _ns / v) that is set in the second step, and each measurement position at the two measurement times of the measurement point p _0, p ₁ or the two measuring positions p _0, replaces the p _1, of the measurement point, the measurement position p ₀ and the measurement point at one measurement time of said two measurement time, the 2 From the motion parallax between the two measurement positions p ₀ and p _{1 at} one measurement time and the position p _inf after the infinite time has elapsed in the movement continuation state, the cross ratio {p _inf p ₀ p ₁ p _c } or using a plurality ratio equivalent operation, a third step of obtaining a position p _c in the superimposed time of the measurement points,
A fourth step of obtaining the polar corresponding to the position p _c by converting the measurement point, the position p _c in the superimposed time pole,
Said a point of pole line, and the moving direction v and angle _{^{r = cos -1 (n s ·}} v)
And a fifth step for obtaining a point comprising
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are set as the respective values of the first parameter and the second parameter. By repeating a plurality of times while changing in the first step and the second step, the value of the first parameter relating to one measurement point in the fifth step is the same, and the second step A curve connecting a plurality of points obtained by a plurality of executions with different values of the parameter is obtained for each of the plurality of measurement points for each value of the first parameter;
A plurality of curves corresponding to a plurality of curves intersecting at the intersection are obtained by obtaining intersections between the curves when the plurality of curves obtained while repeating the first to fifth steps are drawn in the curve drawing space. orientation n _s measurement plane including a measurement _point, and / or, a sixth step of obtaining a physical quantity that indicates the shortest distance to the measuring plane in the measurement time of one of the two measurement time from the observation point It may be executed.

In this case, the measurement point appearing on the image has intensity information,
The fifth step is a step of obtaining the point and voting a value corresponding to the intensity of the measurement point corresponding to the point to the point in the curve drawing space corresponding to the point;
Instead of obtaining the intersection in the sixth step, the voting of the maximum point is obtained by obtaining the maximum point at which the value by voting becomes a maximum value while repeating the first to fifth steps a plurality of times. Azimuth n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in and / or the shortest distance from the observation point to the measurement plane at one of the two measurement times Preferably, this is a step of obtaining a physical quantity that indicates
Measurement points appearing on the image have intensity information,
A seventh step of setting a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter;
In the third step, the physical quantity indexed in the shortest distance set in the first step, the inner product ( _ns / v) set in the second step, and the measurement point, The measurement position p _{0 at} one of the two measurement times, the motion parallax τ set in the seventh step, and the position p of the measurement point after the infinite time has elapsed in the movement continuation state by using the _inf, of the measurement point, a step of determining the position p _c in the superimposed time,
The fifth step obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and measures the measurement point corresponding to the point on the polar line Voting a response intensity corresponding to the motion parallax τ to a point in the curve drawing space corresponding to the point on the polar line,
For a plurality of measurement points in the measurement space, the third to fifth steps are repeated a plurality of times while changing each value of each parameter in the first, second, and seventh steps,
Instead of obtaining the intersection, the sixth step to be executed thereafter is a maximum value obtained by voting while repeating the first, second, seventh and third to fifth steps a plurality of times. Is obtained from the azimuth n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the vote of the maximum point and / or the two measurement times from the observation point. It is also a preferred aspect that it is a step of obtaining a physical quantity that indicates the shortest distance to the measurement plane at one of the measurement times.

  Here, the fifth step in the second image measurement method obtains a curve drawn on the spherical surface as a curve connecting a plurality of points related to one measurement point obtained by repeating the fifth step. Preferably, the fifth step is drawn on a spherical surface as a curve connecting a plurality of points related to one measurement point obtained by repeating the fifth step, and further on a plane. More preferably, it is a step of obtaining a curve projected inside the circle.

In addition, the second image measurement method includes:
A first step of setting a position p _inf of the measurement point after an infinite time in the moving continuous state by setting the moving direction v as the first parameter,
A second step of setting a physical quantity indicative of the shortest distance as a second parameter;
A third step of setting the inner product ( _ns / v) as a third parameter;
The position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step, the physical quantity indicating the shortest distance set in the second step, and the first The inner product ( _ns · v) set in step 3 and the measurement positions p ₀ , p _{1 at the} two measurement times of the measurement point, or these two measurement positions p ₀ , p ₁ , The measurement position p ₀ of the measurement point at one of the two measurement times and the motion parallax τ between the two measurement positions p ₀ and p _{1 of the} measurement point at the two measurement times from using the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent calculation, a fourth step of obtaining a position _{p c} in the superimposed time of the measurement points,
A fifth step of obtaining a polar corresponding to the position p _c by converting the measurement point, the position p _c in the superimposed time pole,
Said a point of pole line, and the moving direction v and angle _{^{r = cos -1 (n s ·}} v)
And a sixth step for obtaining a point comprising
For the plurality of measurement points in the measurement space, the fourth to sixth steps of the first to sixth steps, the respective values of the first to third parameters, respectively, By repeating a plurality of times while changing in each step from 1 to 3, the value of the first parameter and the value of the second parameter are the same for one measurement point in the sixth step. Are the same, and a curve connecting a plurality of points obtained by a plurality of executions with different values of the third parameter is represented by each value of the first parameter and each value of the second parameter. For each of the combinations, the plurality of measurement points is obtained, and then
When a plurality of curves obtained while repeating the first to sixth steps a plurality of times are drawn in the corresponding curve drawing spaces among the plurality of curve drawing spaces according to the value of the first parameter. Obtaining an intersection between the curves for each curve drawing space, and selecting a curve drawing space corresponding to the true movement direction of the measurement point relative to the observation point based on information on the number of curves intersected at the intersection. To determine the true movement direction, and the orientation n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection obtained with respect to the curve drawing space corresponding to the true movement direction, and / or The seventh step of obtaining a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times is performed. May I, in that case,
Measurement points appearing on the image have intensity information,
In the sixth step, the point is obtained, and a value corresponding to the intensity of the measurement point corresponding to the point is voted for the point in the curve drawing space corresponding to the point where the curve including the point is drawn. Step,
Instead of determining the intersection in the seventh step, a maximum point at which the value obtained by voting becomes a maximum value while repeating the first to sixth steps is determined for each curve drawing space. The true movement direction is obtained by selecting the curve drawing space corresponding to the true movement direction based on the local maximum information at the point, and the maximum obtained for the curve drawing space corresponding to the true movement direction a plurality of orientations n _s measurement plane including a measurement point corresponding to a plurality of curves participated in the vote of _points, and / or from the observation point to the measurement plane at one measurement time of said two measurement time Preferably, this is a step for obtaining a physical quantity that indicates the shortest distance of
Measurement points appearing on the image have intensity information,
An eighth step of setting a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a fourth parameter;
In the fourth step, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step and the shortest distance set in the second step. The physical quantity to be indexed, the inner product ( _ns · v) set in the third step, the measurement position p _{0 of the} measurement point at one of the two measurement times, and the eighth using said motion parallax τ set in step, of the measurement point, a step of determining the position p _c in the superimposed time,
The sixth step obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and measures the measurement point corresponding to the point on the polar line Voting a response intensity corresponding to the motion parallax τ to a point in the curve drawing space corresponding to the point on the polar line,
For a plurality of measurement points in the measurement space, the steps from the fourth to the sixth are repeated a plurality of times while changing each value of each parameter in the first, second, third and eighth steps,
The value obtained by voting during the repetition of the first, second, third, eighth, and fourth to sixth steps a plurality of times, instead of the seventh step executed thereafter obtaining the intersection. Finds a local maximum point for each curve drawing space, and obtains the true moving direction by selecting a curve drawing space corresponding to the true moving direction based on the local maximum information at the local maximum point. Azimuth n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the voting of local maximum points obtained for the curve drawing space corresponding to the true moving direction, and / or from the observation point It is also a preferred aspect that it is a step of obtaining a physical quantity indicating the shortest distance to the measurement plane at one of the two measurement times.

Further, the third image measurement method of the image measurement methods of the present invention is an arbitrary one in the measurement space that appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. P ₀ , p ₁ , respectively, at the two different measurement times of the measurement point, and v, the movement direction of the measurement point between the two measurement times relative to the observation point, v, The movement of the measurement point relative to the observation point in the same direction as the movement direction v between the two measurement times and at the same speed as the movement speed between the two measurement times remains unchanged. When the position of the measurement point after elapse of infinite time in the movement continuation state scheduled to be continued is _pinf ,
Using a single ratio (p _inf p ₀ p ₁ ) determined by the three positions p _inf , p ₀ , and p ₁ of the measurement point (p _inf p ₀ p ₁ ) or an operation equivalent to the single ratio, and / or It is also a preferable aspect to obtain a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times.

Here, in the third image measurement method, the single ratio (p _inf p ₀ p ₁ ) or an operation equivalent to the single ratio includes two measurement positions of the measurement point at the two measurement times. instead of p _0, p _1, wherein the measurement point, the the measurement position p ₀ in one of the measurement time of the two measurement time, the measurement point, the two measurement positions p in the two measurement time ₀ , P ₁ and a calculation using motion parallax τ, which is a difference in position between the two.

The third image measurement method employs each position projected on the spherical surface as each position p _inf , p ₀ , and p ₁ of the measurement point, and uses the observation point as a physical quantity that indicates the shortest distance. D _{s is} the shortest distance from the measurement plane at one of the two measurement times, and the relative movement distance of the measurement point between the two measurement times is the observation point. Δx
_n d _s = d _s / Δx
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the normalized shortest distance _n d _s as a parameter;
Using the normalized shortest distance _n d _s set in the first step and a simple ratio (p _inf p ₀ p ₁ ) or an operation equivalent to the simple ratio,
^{_{R = cos -1 (n d s}} / (p inf p 0 p 1))
A second step for obtaining a radius R defined by the relational expression or a relational expression equivalent to the relational expression,
A third step of obtaining a small circle with a radius R centered on the measurement position of the measurement point at one of the two measurement times;
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing the parameters in the first step, and then
A plurality of small circles that intersect at the intersections by obtaining intersections between the small circles when the plurality of small circles obtained while repeating the first to third steps a plurality of times are drawn in the small circle drawing space. the azimuth n _so the measurement plane including a plurality of measurement points _{corresponding,} and / or may be one that performs a fourth step of determining a normalized shortest distance _n d _s0 for the measurement plane.

In this case, the measurement point appearing on the image has intensity information, and the third step obtains the small circle and draws the obtained small circle in the small circle drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the local maximum by obtaining a local maximum at which the value obtained by voting becomes a local maximum while repeating the first to third steps a plurality of times orientation n _so the measurement _plane, and / or is preferably a step of obtaining a normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
A fifth step of setting a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a second parameter;
In the second step, the standardized shortest distance _n d _s set in the first step, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state, and the 2 of the measurement point Using the measurement position p _{0 at} one of the two measurement times and the motion parallax τ set in the fifth step to determine the radius R;
In the third step, the small circle corresponding to the measurement point is obtained, the response intensity corresponding to the motion parallax τ of the measurement point is obtained, and the obtained small circle is drawn in the small circle drawing space. Voting a response intensity corresponding to the motion parallax τ of each measurement point corresponding to the small circle for each point on the trajectory of the small circle when
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter is not the determination of the intersection, but the maximum point at which the value by voting becomes a maximum value while repeating the first, fifth, second and third steps a plurality of times. By obtaining the azimuth n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles participating in the voting of the local maximum point, and / or the normalized shortest distance _n d _s0 for the measurement plane is obtained. It is a preferable aspect that it is a step,
It is also preferable that the third step is a step of obtaining a small circle having a radius R on the spherical surface, and further obtaining a small circle obtained by projecting the small circle on the spherical surface into a circle on the plane. It is an aspect.

Further, the third image measurement method employs each position projected on the spherical surface as each position p _inf , p ₀ , p ₁ of the measurement point, and the physical quantity indicating the shortest distance is used as the physical quantity. D _{s is} the shortest distance between the observation point and the measurement plane at one of the two measurement times, and the relative movement of the measurement point between the two measurement times with respect to the observation point When the distance is Δx,
_n d _s = d _s / Δx
In adopting the normalized shortest distance _n d _s represented,
A first step of setting a position p _inf of the measurement point after an infinite time in the moving continuous state by setting the moving direction v as the first parameter,
A second step of setting the normalized shortest distance _n d _s as a second parameter;
The position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step, the normalized shortest distance _n d _s set in the second step, and a simple ratio ( p _inf p ₀ p ₁ ) or an operation equivalent to the simple ratio,
^{_{R = cos -1 (n d s}} / (p inf p 0 p 1))
A third step for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A fourth step of obtaining a small circle having a radius R around the measurement position of the measurement point at one of the two measurement times;
For the plurality of measurement points in the measurement space, the third step and the fourth step are performed, and the values of the first parameter and the second parameter are respectively set to the first step and the second step. Repeat multiple times while changing in steps, then
A plurality of small circles obtained while repeating the first to fourth steps a plurality of times are respectively transferred to corresponding small circle drawing spaces among the plurality of small circle drawing spaces according to the value of the first parameter. A small intersection corresponding to the true movement direction of the measurement point relative to the observation point is obtained based on information on the number of small circles intersecting at the intersection by obtaining intersections between the small circles at the time of drawing for each small circle drawing space. Measurement including a plurality of measurement points corresponding to a plurality of small circles intersecting at intersections obtained for a small circle drawing space corresponding to the true movement direction while obtaining the true movement direction by selecting a circle drawing space plane orientation n _so, and / or may be one that performs a fifth step of obtaining a normalized shortest distance _n d _s0 for the measurement plane,
In that case, the measurement point appearing on the image has intensity information,
In the fourth step, the small circle is obtained, and each small point on the locus of the small circle when the obtained small circle is drawn in the small circle drawing space corresponding to the small circle is added to the small circle. Voting a value corresponding to the intensity of the corresponding measurement point,
Instead of obtaining the intersection in the fifth step, a maximum point at which the value by voting becomes a maximum value while repeating the first to fourth steps is determined for each small circle drawing space. The true movement direction is obtained by selecting the small circle drawing space corresponding to the true movement direction based on the local maximum information at the local maximum point, and the small circle drawing space corresponding to the true movement direction is obtained. This is a step of obtaining the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles participating in the vote of the given maximum point and / or the normalized shortest distance _n d _s0 for the measurement plane. Preferably, or
Measurement points appearing on the image have intensity information,
A sixth step of setting a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter;
The second step includes the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step, and the normalized shortest distance _n set in the third step. The radius R is obtained using d _s , the measurement position p _{0 at} one of the two measurement times of the measurement point, and the motion parallax τ set in the fifth step. Step,
The fourth step obtains the small circle corresponding to the measurement point, obtains the response strength of the measurement point corresponding to the motion parallax τ, and determines the obtained small circle as a small circle corresponding to the small circle. Voting the response intensity corresponding to the motion parallax τ of each measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in a circle drawing space;
For a plurality of measurement points in the measurement space, the third step and the fourth step are repeated a plurality of times while changing each value of each parameter in the first, second and sixth steps,
Subsequent to the fifth step being executed, instead of obtaining the intersection, the value obtained by voting during the repetition of the first, second, sixth, third and fourth steps becomes a maximum value. A local maximum point is obtained for each small circle drawing space, and the true moving direction is obtained by selecting a small circular drawing space corresponding to the true moving direction based on information on the local maximum value at the local maximum point. Azimuth n _{so of} a measurement plane including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the maximum points obtained for the small circle drawing space corresponding to the moving direction of the circle, and / or It is also a preferred aspect that it is a step of _{obtaining the} normalized shortest distance _n d _s0 .

Furthermore, a fourth image measurement method of the image measurement methods of the present invention is an arbitrary image in the measurement space that appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. P ₀ , p ₁ , and the movement direction v between the two measurement times relative to the observation point are the same as the measurement positions at two measurement times different from each other. P _inf is the position of the measurement point after elapse of infinite time in the planned movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is. When
Using a single ratio (p _inf p ₀ p ₁ ) determined by the three positions p _inf , p ₀ , and p ₁ of the measurement point, or an operation equivalent to the single ratio,
A physical quantity indicating a distance between the observation point and the measurement point at one of the two measurement times is obtained.

Here, in the fourth image measurement method, for the calculation of the single ratio (p _inf p ₀ p ₁ ) or equivalent to the single ratio, two measurement positions p _{0 of the} measurement point at the two measurement times are used. , in place of p _1, wherein the measurement point, the the measurement position p ₀ in one of the measurement time of the two measurement time, the measurement point, the two measurement positions p in the two measurement time _0, p Computation using motion parallax τ, which is the difference in position between _one, is included.

In the fourth image measurement method, as a physical quantity indicating the distance, a distance between the observation point and the measurement point at one of the two measurement times is set as d ₀ , and the two measurements are performed. When the movement distance of the measurement point between the times with respect to the observation point is Δx,
_n d ₀ = d ₀ / Δx
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _inf p ₀ p ₁ )
Or a relational expression equivalent to the relational expression.

A fifth image measurement method of the image measurement methods of the present invention is an arbitrary image in the measurement space that appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. The movement of the measurement point in the same direction as the movement direction relative to the observation point between two different measurement times and at the same speed as the movement speed between the two measurement times is continued as it is. in a mobile continuous state of events and shall, wherein the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, the coordinates of the voting space defined by the azimuth n _s of the measurement plane parameters A first step to set as
In the first step, the measurement position p ₀ of the measurement point at one of the two measurement times, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state, set, from the coordinates of the voting space, the measurement point, a second step of obtaining the two measurement positions p _0, p ₁ if motion parallax is the difference in the position of and τ in the two measurement time ,
A third step of determining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

The sixth image measurement method of the image measurement methods of the present invention is an arbitrary image in the measurement space that appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. By setting, as a first parameter, a moving direction v of the measuring point relative to the observation point between two different measurement times, the measuring point is moved in the same direction as the moving direction, and A first step of setting a position p _{inf of the} measurement point after elapse of an infinite time in a scheduled movement continuation state, in which the movement at the same speed as the movement speed between the two measurement times is continued;
In the mobile continuity state, and the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, defined by the azimuth n _s of the measurement plane, voting space corresponding to the first parameter A second step of setting the coordinates in as a second parameter;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the voting set in the second step A third step of obtaining a motion parallax τ that is a difference between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from coordinates in space;
A fourth step of obtaining response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

The seventh image measurement method of the image measurement methods of the present invention is:
A predetermined observation point in the measurement space for observing the predetermined measurement space, and an arbitrary measurement point in the measurement space that appears in an image when the measurement space is viewed from the observation point. to set a physical quantity that indicates the shortest distance between the measurement plane at one measurement time of the two different measurement times, the coordinates of the voting space defined by the azimuth n _s of the measuring plane as a parameter A first step;
In the same direction as the relative movement direction of the measurement point relative to the observation point between the measurement point p ₀ at one of the two measurement times and the measurement point between the two measurement times. And the position p _inf after the infinite time has elapsed in the planned movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is, and is set in the first step. A second step of obtaining a motion parallax τ that is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the coordinates in the voting space;
A third step of determining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

The eighth image measurement method of the image measurement methods of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first step of setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
Wherein from the observation point, comprising said measuring points, and the physical quantity that indicates the shortest distance to the measurement plane at the measurement time of one of the two measurement time is defined by the azimuth n _s of the measurement plane, the A second step of setting the coordinates in the voting space according to the first parameter as a second parameter;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the voting set in the second step A third step of obtaining a motion parallax τ that is a difference between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from coordinates in space;
A fourth step of obtaining response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

The ninth image measurement method of the image measurement methods of the present invention is:
Two measurement positions p ₀ , at two different measurement times of arbitrary measurement points in the measurement space, which appear in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. a first step of setting, as a parameter, motion parallax τ, which is a difference in position between p ₁ ;
In the same direction as the relative movement direction of the measurement point relative to the observation point between the measurement point p ₀ at one of the two measurement times and the measurement point between the two measurement times. And the position p _inf after elapse of an infinite time of the measurement point in the movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is, and the first Is determined by a physical quantity that indicates a superimposition time at which the measurement plane including the measurement point overlaps the observation point and the orientation n _{s of the} measurement plane in the movement continuation state. A second step for determining coordinates in a voting space;
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the first step is obtained at the measurement point. A third step;
Voting the response strength determined in the third step to the coordinates determined in the second step in the voting space; and
For the plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while sequentially changing the parameter values in the first step. It is characterized by doing.

The tenth image measurement method of the image measurement methods of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first step of setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
A second step of setting, as a second parameter, motion parallax τ, which is a difference in position between the two measurement positions p ₀ and p ₁ at the two measurement times, of the measurement point;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the motion parallax τ set in the second step and a, the in the mobile continuity state, and the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, defined by the azimuth n _s of the measurement plane, according to the first parameter A third step for determining coordinates in the voting space;
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the second step is obtained at the measurement point. A fourth step;
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

The eleventh image measurement method of the image measurement methods of the present invention is:
Two measurement positions p ₀ , at two different measurement times of arbitrary measurement points in the measurement space, which appear in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. a first step of setting, as a parameter, motion parallax τ, which is a difference in position between p ₁ ;
The same direction as the relative movement direction of the measurement point with respect to the observation point between the two measurement times and the measurement position p ₀ at one of the two measurement times. And a position p _inf after the infinite time of the measurement point in a state where the movement continues at the same speed as the movement speed between the two measurement times and the scheduled movement continuation state, A physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point, based on the motion parallax τ set in step 1 When, a second step of obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the first step is obtained at the measurement point. A third step;
Voting the response strength determined in the third step to the coordinates determined in the second step in the voting space; and
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

The twelfth image measurement method of the image measurement methods of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first step of setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
A second step of setting, as a second parameter, motion parallax τ, which is a difference between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the motion parallax τ set in the second step And a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point, and the direction n _{s of the} measurement plane A third step for obtaining coordinates in the voting space according to the first parameter, defined by:
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the second step is obtained at the measurement point. A fourth step;
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

Furthermore, the thirteenth image measurement method of the image measurement methods of the present invention is:
Based on two images when the measurement space is viewed at two different measurement times from a predetermined observation point in the predetermined measurement space, any two measurement points in the measurement space A first step of obtaining a response intensity corresponding to motion parallax, which is a difference between positions of two measurement positions at one measurement time;
The response intensity obtained in the first step is set in the same direction as the movement direction of the measurement point relative to the observation point between the two measurement times and between the two measurement times. A physical quantity that indicates a superposition time at which a measurement plane including the measurement point overlaps the observation point in a planned movement continuation state in which the movement at the same speed as the movement speed is continued, and an orientation of the measurement plane A second step of voting on coordinates corresponding to the measurement points and the motion parallax within a defined voting space;
The first step and the second step are executed a plurality of times for a plurality of measurement points in the measurement space.

  Here, in the thirteenth image measurement method, a measurement plane including a plurality of measurement points participating in the voting of the maximum point by obtaining a maximum point in the voting space where the value by the vote becomes a maximum value. And / or a third step for obtaining a physical quantity that indicates a superimposition time at which the measurement plane overlaps the observation point.

The fourteenth image measurement method of the image measurement methods of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A first step of setting a general moving direction as a parameter;
Based on two images when the measurement space is viewed from the observation point at the two measurement times, the movement is a difference between the two measurement positions at the two measurement times of the measurement point. A second step for obtaining a response intensity corresponding to the parallax;
The response intensity obtained in the second step is set in the same direction as the movement direction of the measurement point relative to the observation point between the two measurement times, and between the two measurement times. The physical quantity indicating the superposition time when the measurement plane including the measurement point overlaps the observation point in the planned movement continuation state in which the movement at the same speed as the movement speed is continued as it is and the direction of the measurement plane A third step of voting on coordinates corresponding to the measurement point and the motion parallax in a voting space defined by the parameters set in the first step,
Executing the second and third steps of the first to third steps for a plurality of measurement points in the measurement space a plurality of times while changing parameter values in the first step. It is characterized by.

  Here, in the fourteenth image measurement method, a local maximum point at which the value obtained by voting becomes a local maximum value is obtained for each voting space, and the measurement point is relative to the observation point based on information on the local maximum value at the local maximum point. A plurality of measurement points participating in the voting of the maximum points obtained for the voting space corresponding to the true movement direction while obtaining the true movement direction by selecting a voting space corresponding to the true movement direction. And / or a fourth step of obtaining a physical quantity that indicates a superposition time at which the measurement plane overlaps the observation point.

The fifteenth image measurement method of the image measurement methods of the present invention is:
Based on two images when the measurement space is viewed at two different measurement times from a predetermined observation point in the predetermined measurement space, the two measurements at arbitrary measurement points in the measurement space A first step of obtaining a response intensity corresponding to a motion parallax that is a difference between positions of two measurement positions at a time;
The response intensity obtained in the first step is a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point. Voting to coordinates corresponding to the measurement point and the motion parallax τ in a voting space defined by the orientation of the measurement plane,
The first step and the second step are executed a plurality of times for a plurality of measurement points in the measurement space.

  Here, in the image measurement method, the azimuth of a measurement plane including a plurality of measurement points participating in the voting of the maximum point by obtaining a maximum point in the voting space where the value by the voting becomes a maximum value, And / or a third step of obtaining a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times.

The sixteenth image measurement method of the image measurement methods of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A first step of setting a general moving direction as a parameter;
Motion parallax, which is the difference between the two measurement positions at the two measurement times of the measurement point, based on the two images when the measurement space is viewed from the observation point at the two measurement times. A second step for obtaining a response intensity corresponding to
The response intensity obtained in the second step is a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point. A third step of voting on coordinates corresponding to the measurement points and the motion parallax in a voting space according to the parameters set in the first step, defined by the orientation of the measurement plane; Have
Executing the second and third steps of the first to third steps for a plurality of measurement points in the measurement space a plurality of times while changing parameter values in the first step. It is characterized by.

  Here, in the sixteenth image measurement method, a local maximum point at which the value by voting becomes a local maximum is obtained for each voting space, and the relative measurement point relative to the observation point is based on the local maximum information at the local maximum point. A plurality of measurement points participating in the voting of the maximum points obtained for the voting space corresponding to the true movement direction while obtaining the true movement direction by selecting a voting space corresponding to the true movement direction. And / or a fourth step of obtaining a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times. May be.

The seventeenth image measurement method of the image measurement methods of the present invention is
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , respectively, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the visual axis direction v connecting the two observation points is represented by p _axis . wherein one of the observation points of the point, when on the viewing plane extending parallel to the measurement plane including the measurement point, the position of intersection between the straight line and the p _c,
The four positions _{p axis,} using p _R, p L, the cross ratio _{{p axis p R p L p} c} or plurality ratio equivalent operations determined by _{p c,} the orientation of the measurement plane, and / or A physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points is obtained.

Here, in the seventeenth image measurement method, the cross ratio {p _axis p _R p _L p _c } or an operation equivalent to the cross ratio is obtained by observing the measurement point from the two observation points. Instead of the two measurement positions p _R and p _L , the measurement position p _R when the measurement point is observed from one of the two observation points and the two measurement points. Calculations using binocular parallax σ, which is the difference in position between the measurement positions p _R and p _L when observed from the observation point, are included.

In the seventeenth image measurement method, one of the measurement plane and the two observation points is used as a physical quantity that indicates a distance in the visual axis direction between the measurement plane and the two observation points. When the distance in the visual axis direction between the observation points is d _c and the distance between the two observation points is Δx _LR ,
_{n d c = d c / Δx} LR
In adopting the normalized distance _n d _c represented, the normalized distance _n d _c,
n d _c = {p _axis p _R p _L p _c }
Alternatively, the relational expression may be obtained by using a relational expression or a relational expression equivalent to the relational expression.

Further, in the seventeenth image measuring method, a process for obtaining a polar corresponding to the position p _c by converting the position p _c of the intersection on the observation plane electrode, the plurality present in the measuring space By executing the measurement points and obtaining the intersections between the polar lines when the plurality of polar lines obtained by these processes are drawn in the polar drawing space, a plurality of polar lines intersecting at the intersections are obtained. A physical quantity that indicates the azimuth of the measurement plane including the measurement point and / or the distance in the visual axis direction between the measurement plane and one of the two observation points may be obtained.

Further, the image measuring method of the 17, when the measurement point appearing on an image is one that has information of intensity, the measurement point by converting a position p _c of intersection on said viewing plane electrode And a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line when the calculated polar line is drawn in the polar line drawing space. A voting process is executed for a plurality of measurement points existing in the measurement space, and a maximal point at which the value obtained by voting during the execution of these processes becomes a maximal value is obtained, thereby participating in the voting of the maximal point The orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines and / or the distance in the visual axis direction between the measurement plane and one of the two observation points The physical quantity to be indexed may be obtained .

Further, in the seventeenth image measurement method, the measurement point appearing on the image has intensity information, and the position p _c of the intersection point on the observation plane is subjected to polar conversion to obtain the measurement point. The corresponding polar line is obtained, and the response intensity corresponding to the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points is obtained and obtained. A process of voting a response intensity corresponding to the binocular parallax σ of each measurement point corresponding to the polar line for each point on the polar line trajectory when the polar line is drawn in the polar drawing space. , Executing a plurality of measurement points existing in the measurement space, and obtaining a maximum point at which a value obtained by voting during the execution of these processes becomes a maximum value, thereby obtaining a plurality of extreme points participating in the vote of the maximum point. Measurement plane with multiple measurement points corresponding to the line Orientation, and / or it may be one that obtains a physical quantity indicative of the distance of the visual axis direction between one of the observation point of the measuring plane and the two observation points.

Furthermore, in the 17th image measuring method, when the seek position p _c, a physical quantity indicative of the distance of the visual axis direction between one of the observation point of the measuring plane and the two observation points The measurement points p _R and p _L when observed from the two observation points, or the two observation points of the measurement points instead of the two measurement positions p _R and p _L. And binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points, and the measurement position p _R when observed from one of the observation points. Knowing the position p _axis of the infinite point of the measurement point, and using the _cross ratio {p _axi p _R p _L p _c } or an operation equivalent to the cross ratio, the intersection point on the observation plane it may be determined position p _c of.

Furthermore, in the seventeenth image measurement method, a first step of setting, as a parameter, a physical quantity that indicates a distance in the visual axis direction between the measurement plane and one of the two observation points. When,
A physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points set in the first step, and the two measurement points. Each measurement position p _R , p _L when observed from the observation point, or these measurement points instead of these two measurement positions p _R , p _L were observed from one of the two observation points a binocular parallax σ between the measurement position p _R and the measurement point, the two two measurement positions p _R when observed from the observation _point, and if p _L of the time, the position p _axis of the point at infinity from a second step of using the double ratio _{{p axis p R p L p} c} or plurality ratio equivalent calculation to determine the position _{p c} of the intersection on the observation plane,
And a third step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c of the intersection on the observation plane,
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing the value of the parameter in the first step, and then
A plurality of polar lines intersecting at the intersections by obtaining intersections between the polar lines when the plurality of polar lines obtained while repeating the first to third steps a plurality of times are drawn in the polar drawing space. A physical quantity that indicates the orientation of the measurement plane including a plurality of measurement points corresponding to and / or the distance in the visual axis direction between the measurement plane and one of the two observation points is obtained. The fourth step may be executed.

In this case, the measurement point appearing on the image has intensity information, and the third step obtains the polar line and draws the obtained polar line in the polar drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the local maximum points by obtaining a local maximum point at which the value by voting becomes a maximum value while repeating the first to third steps a plurality of times Preferably, it is a step of obtaining a physical quantity indicating the orientation of the measurement plane and / or the distance in the visual axis direction between the measurement plane and one of the two observation points, or
Measurement points appearing on the image have intensity information,
The measurement point, a fifth step of setting two measurement positions p _R when observed from the two observation _points, the binocular parallax σ between and if p _L as the second parameter,
The second step is a physical quantity that indicates the distance in the visual axis direction between the measurement plane and one of the two observation points set in the first step; The measurement position p _R when observed from one of the two observation points of the measurement point, the binocular parallax σ set in the fifth step, and the position p _{axis of the} infinity point with bets, a step of obtaining an intersection p _c on the observation plane,
The third step obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained polar line in the polar drawing space Voting a response intensity corresponding to the binocular parallax σ of a measurement point corresponding to the polar line for each point on the trajectory of the polar line when
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter obtains the maximum point where the value obtained by voting becomes the maximum value while repeating the first, fifth, second and third steps a plurality of times instead of obtaining the intersection. Thus, the orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines participating in the vote of the local maximum point, and / or one observation point of the measurement plane and the two observation points It is also a preferred form that it is a step of obtaining a physical quantity indicating the distance in the visual axis direction between the two.

Here, in the image measuring method of the first 17, the third step is preferably a step of obtaining a polar depicted as great circle on a sphere by converting the position p _c poles, more the third step is, the drawn as a great circle on a sphere by converting the position p _c poles, more preferably a further step of obtaining the polar which is projected inside the circle on the plane. Alternatively, the third step may be a step of obtaining a polar line drawn as a straight line on a plane by subjecting the position _pc to polar conversion.

The seventeenth image measurement method is
A first step of setting the position p _{axis of the} infinity point by setting the direction v of the visual axis as a first parameter;
A second step of setting, as a second parameter, a physical quantity that indicates a distance in the visual axis direction between the measurement plane and one of the two observation points;
When observing from the position p _axis set in the first step, the physical quantity indexed in the visual axis direction set in the second step, and the two observation points of the measurement point each measurement position p _{R of,} p _L or the two measuring positions p _R, replaces the p _L, of the measurement point, the measurement position when observed from one of the observation point of said two observation point p _R And the binocular parallax σ between the two measurement positions p _R and p _L when the measurement points are observed from the two observation points, or the cross ratio {p _axis p _R p _L p _c } or using said plurality ratio equivalent operation, a third step of determining the position p _c of the intersection on the observation plane of the measurement point,
The measurement point, and a fourth step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c of the intersection on the observation plane,
For the plurality of measurement points in the measurement space, the third step and the fourth step of the first to fourth steps are changed to values of the first parameter and the second parameter, respectively. Are repeated several times while changing in the first step and the second step, respectively,
A plurality of polar lines obtained while repeating the first to fourth steps a plurality of times are respectively converted into corresponding polar line drawing spaces among a plurality of polar line drawing spaces according to the value of the first parameter. An intersection between polar lines at the time of drawing is obtained for each polar drawing space, and the true line drawing space corresponding to the true visual axis direction is selected based on information on the number of polar lines intersecting at the intersection. And / or the orientation of a measurement plane including a plurality of measurement points corresponding to a plurality of polar lines intersecting at the intersection obtained for the polar drawing space corresponding to the true visual axis direction, and / or The fifth step of obtaining a physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points may be executed.

In this case, the measurement point appearing on the image has intensity information, and the fourth step obtains the polar line, and the obtained polar line corresponds to the polar line. Voting a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line when drawn in the line drawing space, and the fifth step is a step of voting the intersection. Instead of obtaining the maximum point where the value obtained by voting during the repetition of the first to fourth steps a plurality of times is determined for each polar drawing space, and based on the information on the maximum value at the maximum point. The true visual axis direction is obtained by selecting the polar drawing space corresponding to the true visual axis direction, and the maximum point obtained for the polar drawing space corresponding to the true visual axis direction is voted Duplicate corresponding to multiple participating polar lines A physical quantity indicating the orientation of the measurement plane including the measurement point and / or the distance in the visual axis direction between the measurement plane and one of the two observation points. Is preferred, or
Measurement points appearing on the image have intensity information,
A sixth step of setting a binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a third parameter;
The third step is between the position _paxis set in the first step and the measurement plane and one of the two observation points set in the second step. A physical quantity indicating a distance in the visual axis direction, a measurement position p _R when the measurement point is observed from one of the two observation points, and the fifth step. wherein by using the binocular parallax sigma, a step of obtaining an intersection p _c on the observation plane,
In the fourth step, a polar line corresponding to the measurement point is obtained, a response intensity corresponding to the binocular parallax σ of the measurement point is obtained, and the obtained polar line is drawn in the polar drawing space. Voting a response intensity corresponding to the binocular parallax σ of a measurement point corresponding to the polar line for each point on the trajectory of the polar line when
For each of a plurality of measurement points in the measurement space, the third step and the fourth step are changed, and the values of the parameters are changed in the first step, the second step, and the sixth step. While repeating several times,
The fifth step executed thereafter is not the determination of the intersection, but the maximum value obtained by voting while repeating the first, second, sixth, third and fourth steps a plurality of times. A point is obtained for each polar drawing space, and the true visual axis direction is obtained by selecting a polar drawing space corresponding to the true visual axis direction based on the local maximum information at the local maximum point, and The orientation of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the maximum points obtained for the polar drawing space corresponding to the true visual axis direction, and / or the measurement plane and the aforementioned It is also a preferred embodiment that it is a step of obtaining a physical quantity indicating the distance in the visual axis direction between one of the two observation points.

An eighteenth image measurement method of the image measurement methods of the present invention is an image in the measurement space that appears in an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. P _R , p _L , the visual axis direction connecting the two observation points, v, the measurement points when the measurement positions of the arbitrary observation points when observed from the two observation points are respectively included, The position of an infinite point on a straight line extending in the same direction as the visual axis direction v is p _axis , including one observation point of the two observation points, and on an observation plane extending in parallel with the measurement plane including the measurement point Where the position of the intersection with the straight line is p _c and the orientation of the measurement plane is n _s ,
The four positions _{p axis, p R, p L} , cross ratios determined by _{p c {p axis p R p} L p c} or plurality ratio equivalent operations and the visual axis and the orientation _{n s} of the measurement plane by using the inner product of the direction v of the (n s _· v), the azimuth n _s measurement _plane, and / or the shortest between one observation point of the said measurement plane the two observation points It is characterized in that a physical quantity indicating a distance is obtained.

Here, in the eighteenth image measurement method, the cross ratio {p _axis p _R p _L p _c } or an operation equivalent to the cross ratio is obtained by observing the measurement point from the two observation points. Instead of the two measurement positions p _R and p _L , the measurement position p _R when the measurement point is observed from one of the two observation points and the two measurement points. Computation using binocular parallax σ, which is the difference in position between the two measurement positions p _R and p _L when observed from the observation point, is included.

In the eighteenth image measurement method, as a physical quantity using the shortest distance as an index, the shortest distance between the measurement plane and one of the two observation points is d _s , and the measurement plane and the measurement point when the distance of the visual axis direction between one of the observation points of the two observation points were d _c, the distance between the each other two observation points and [Delta] x _LR,
n d _s = d _s / Δx _LR
The standardized shortest distance _n d _s expressed by the following formula is adopted, and the standardized shortest distance _n d _{s
n d c = d c / Δx} LR
Using the normalized distance _n d _c expressed by the above and the inner product (n _s · v),
_n d _s = _n d _c (n _s · v)
It is preferable to use the following relational expression.

The eighteenth image measurement method is
A first step of setting a physical quantity indicating the shortest distance as a first parameter;
A second step of setting the inner product ( _ns / v) as a second parameter;
When observing from each of the two observation points of the measurement point, the physical quantity indicating the shortest distance set in the first step, the inner product ( _ns / v) set in the second step each measurement position p _{R of,} p _L or the two measuring positions p _R, replaces the p _L, of the measurement point, the measurement position when observed from one of the observation point of said two observation point p _R And the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points and the position p _{axis of the} infinity point of the measurement point. with _{p axis p R p L p c} } or plurality ratio equivalent operation, a third step of determining the position _{p c} of the intersection on the observation plane,
A fourth step of obtaining the polar corresponding to the position p _c by polar transformation the position p _c of the intersection on the observation plane,
The point on the polar line, and the direction v and angle of the visual axis r = cos ⁻¹ (n _s · v)
And a fifth step for obtaining a point comprising
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are set as the respective values of the first parameter and the second parameter. By repeating a plurality of times while changing in the first step and the second step, the value of the first parameter relating to one measurement point in the fifth step is the same, and the second step A curve connecting a plurality of points obtained by a plurality of executions with different values of the parameter is obtained for each of the plurality of measurement points for each value of the first parameter;
A plurality of curves corresponding to a plurality of curves intersecting at the intersection are obtained by obtaining intersections between the curves when the plurality of curves obtained while repeating the first to fifth steps a plurality of times are drawn in the curve drawing space. A sixth step of obtaining a physical quantity indicating the orientation n _{s of the} measurement plane including the measurement point and / or the shortest distance between the measurement plane and one of the two observation points is executed. May be,
In that case, the measurement point appearing on the image has intensity information, and the fifth step obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to the point, wherein the sixth step repeats the first to fifth steps a plurality of times instead of obtaining the intersection by obtaining the maximum point value by vote becomes the maximum value, azimuth n _s measurement plane including a plurality of measurement points corresponding to a plurality of curves participated in the vote of ultra large _point, and / or, and the measurement plane Preferably, it is a step of obtaining a physical quantity that indicates the shortest distance between one of the two observation points, or
Measurement points appearing on the image have intensity information,
A seventh step of setting a binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a third parameter;
In the third step, the physical quantity indexed in the shortest distance set in the first step, the inner product ( _ns / v) set in the second step, and the measurement point, Using the measurement position p _R when observing from one of the two observation points, the binocular parallax σ set in the seventh step, and the position p _{axis of the} infinity point a step of determining the position p _c of the intersection on the observation plane,
The fifth step obtains the point on the polar line corresponding to the measurement point, obtains the response intensity of the measurement point corresponding to the binocular parallax σ, and measures the point corresponding to the point on the polar line Voting a response intensity of a point corresponding to the binocular disparity σ to a point in the curve drawing space corresponding to the point on the polar line,
For a plurality of measurement points in the measurement space, the third to fifth steps are repeated a plurality of times while changing each value of each parameter in the first, second, and seventh steps,
Instead of obtaining the intersection, the sixth step to be executed thereafter is a maximum value obtained by voting while repeating the first, second, seventh and third to fifth steps a plurality of times. Is obtained by obtaining the azimuth n _{s of the} measurement plane including a plurality of measurement points corresponding to the plurality of curves that participated in the vote of the maximum point and / or the measurement plane and the two observation points. It is also a preferable aspect that it is a step of obtaining a physical quantity indicating the shortest distance between one of the observation points.

  In this case, it is preferable that the fifth step is a step of obtaining a curve drawn on a spherical surface as a curve connecting a plurality of points related to one measurement point obtained by repeating the fifth step. Further, the fifth step is drawn on a spherical surface as a curve connecting a plurality of points related to one measurement point, which is obtained by repeating the fifth step, and is further projected into a circle on the plane. More preferably, the obtained curve is obtained.

The eighteenth image measurement method is as follows.
A first step of setting the position p _{axis of the} infinity point by setting the visual axis direction v as a first parameter;
A second step of setting a physical quantity indicative of the shortest distance as a second parameter;
A third step of setting the inner product ( _ns / v) as a third parameter;
The position p _{axis of the} infinity point set in the first step, the physical quantity indicating the shortest distance set in the second step, and the inner product set in the third step ( n _s · v) and each of the measurement points p _R and p _L when observed from each of the two observation points of the measurement point, or the measurement points instead of these two measurement positions p _R and p _L , The measurement position p _R when observed from one of the two observation points and the two measurement positions p _R and p _L when observed from the two observation points. and a binocular parallax sigma, a fourth step of using the double ratio _{{p axis p R p L p} c} or plurality ratio equivalent calculation to determine the position _{p c} of the intersection on the observation plane,
A fifth step of obtaining a polar corresponding to the position p _c by polar transformation the position p _c of the intersection on the observation plane,
The point on the polar line, and the direction v and angle of the visual axis r = cos ⁻¹ (n _s · v)
And a sixth step for obtaining a point comprising
For the plurality of measurement points in the measurement space, the fourth to sixth steps of the first to sixth steps, the respective values of the first to third parameters, respectively, By repeating a plurality of times while changing in each step from 1 to 3, the value of the first parameter and the value of the second parameter are the same for one measurement point in the sixth step. Are the same, and a curve connecting a plurality of points obtained by a plurality of executions with different values of the third parameter is represented by each value of the first parameter and each value of the second parameter. For each of the combinations, the plurality of measurement points is obtained, and then
When a plurality of curves obtained while repeating the first to sixth steps a plurality of times are drawn in the corresponding curve drawing spaces among the plurality of curve drawing spaces according to the value of the first parameter. Obtaining the true visual axis direction by selecting the curve drawing space corresponding to the true visual axis direction based on the information on the number of curves intersecting at the intersection. Azimuth n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection obtained with respect to the curve drawing space corresponding to the true visual axis direction, and / or the measurement plane and the two The seventh step of obtaining a physical quantity indicating the shortest distance between one of the observation points may be performed,
In this case, the measurement point appearing on the image has intensity information, and the sixth step obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to the point in which a curve including the point is drawn, wherein the seventh step is to replace the first to sixth in place of obtaining the intersection A maximum point where the value obtained by voting during a plurality of steps is maximized is obtained for each curve drawing space, and a curve drawing space corresponding to the true visual axis direction is selected based on information on the maximum value at the maximum point. And a measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the voting of the maximum points obtained for the curve drawing space corresponding to the true visual axis direction. Direction _ns And / or is preferably a step of obtaining a physical quantity indicating the shortest distance between the measurement plane and one of the two observation points, or
Measurement points appearing on the image have intensity information,
An eighth step of setting binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a fourth parameter;
The fourth step includes the position p _{axis of the} infinity point set in the first step, the physical quantity indicating the shortest distance set in the second step, and the third step. The inner product ( _ns · v) set in step, the measurement position p _R when the measurement point is observed from one of the two observation points, and the eighth step. wherein by using the binocular parallax σ was a step of determining a position p _c of the intersection on the observation plane,
The sixth step obtains the point on the polar line corresponding to the measurement point, obtains the response intensity of the measurement point corresponding to the binocular parallax σ, and measures the point corresponding to the point on the polar line Voting a response intensity of a point corresponding to the binocular disparity σ to a point in the curve drawing space corresponding to the point on the polar line,
For a plurality of measurement points in the measurement space, the steps from the fourth to the sixth are repeated a plurality of times while changing each value of each parameter in the first, second, third and eighth steps,
Instead of obtaining the intersection, the seventh step executed thereafter is a value obtained by voting while repeating the first, second, third, eighth, and fourth to sixth steps a plurality of times. Is determined for each curve drawing space by selecting a curve drawing space corresponding to the true visual axis direction based on information on the maximum value at the local maximum point. And azimuth n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the voting of the maximum points obtained for the curve drawing space corresponding to the true visual axis direction, and / or It is also a preferred aspect that it is a step of obtaining a physical quantity that indicates the shortest distance between the measurement plane and one of the two observation points.

In addition, a nineteenth image measurement method of the image measurement methods of the present invention is an image in the measurement space that appears in an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. Each of the measurement points when observed from each of the two observation points p _R and p _L , the visual axis direction connecting the two observation points v, and the measurement point, When the position of an infinite point on a straight line extending in the same direction as the two visual axes is p _axis ,
Using a single ratio determined by the three positions p _axis , p _R , and p _L (p _axis p _R p _L ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the measurement plane A physical quantity indicating the shortest distance between one of the two observation points is obtained.

Here, in the nineteenth image measurement method, the unit ratio (p _axis p _R p _L ) or an operation equivalent to the unit ratio is obtained when the measurement point is observed from the two observation points. Instead of the two measurement positions p _R and p _L , the measurement position p _R when the measurement point is observed from one of the two observation points, and the two observations of the measurement point Calculations using binocular parallax σ, which is the difference in position between the two measurement positions p _R and p _L when observed from a point, are included.

In the nineteenth image measurement method, each position projected on a spherical surface is adopted as each of the positions p _axis , p _R , and p _L , and the measurement plane and the 2 are used as physical quantities indicating the shortest distance. When the shortest distance between one of the two observation points is d _s and the distance between the two observation points is Δx _LR ,
_n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the normalized shortest distance _n d _s as a parameter;
Using the normalized shortest distance _n d _s set in the first step and a simple ratio (p _axis p _R p _L ) or an operation equivalent to the simple ratio,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A second step for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A third step of obtaining a small circle with a radius R centered on the measurement position of the measurement point when observed from one of the two observation points;
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing the parameters in the first step, and then
A plurality of small circles intersecting at the intersections by obtaining intersections between the small circles when the plurality of small circles obtained while repeating the first to third steps a plurality of times are drawn in the small circle drawing space. The fourth step of obtaining the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to and / or the normalized shortest distance _n d _s0 for the measurement plane may be performed,
In this case, the measurement point appearing on the image has intensity information, and the third step obtains the small circle and draws the obtained small circle in the small circle drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the local maximum by obtaining a local maximum at which the value obtained by voting becomes a local maximum while repeating the first to third steps a plurality of times orientation n _so the measurement _plane, and / or is preferably a step of obtaining a normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
A fifth step of setting, as a second parameter, binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement points;
In the second step, the normalized shortest distance _n d _s set in the first step, the position p _{axis of the} infinity point, and one of the two observation points of the measurement point Using the measurement position p ₀ when observed from the observation point and the binocular parallax σ set in the fifth step, to determine the radius R;
The third step obtains the small circle corresponding to the measurement point, obtains the response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained small circle in the small circle drawing space Voting the response intensity corresponding to the binocular parallax of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter is not the determination of the intersection, but the maximum point at which the value by voting becomes a maximum value while repeating the first, fifth, second and third steps a plurality of times. By obtaining the azimuth n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles participating in the voting of the local maximum point, and / or the normalized shortest distance _n d _s0 for the measurement plane is obtained. It is also a preferred embodiment that it is a step.

  Preferably, the third step is a step of obtaining a small circle having a radius R on the spherical surface, and further obtaining a small circle obtained by projecting the small circle on the spherical surface into a circle on the plane. .

The nineteenth image measurement method employs each position projected on a spherical surface as each position p _axis , p _R , and p _L , and uses the measurement plane as a physical quantity indicating the shortest distance. When the shortest distance between one of the two observation points is d _s and the distance between the two observation points is Δx _LR ,
_n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the position p _{axis of the} infinity point by setting the visual axis direction v as a first parameter;
A second step of setting the normalized shortest distance _n d _s as a second parameter;
The position p _{axis of the} infinity point set in the first step, the normalized shortest distance _n d _s set in the second step, and a simple ratio (p _axi p _R p _L ) or the Using simple ratio and equivalent operations,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A third step for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A fourth step of obtaining a small circle having a radius R around the measurement position of the measurement point when observed from one of the two observation points;
For the plurality of measurement points in the measurement space, the third step and the fourth step are performed, and the values of the first parameter and the second parameter are respectively set to the first step and the second step. Repeat multiple times while changing in steps, then
A plurality of small circles obtained while repeating the first to fourth steps a plurality of times are respectively transferred to corresponding small circle drawing spaces among the plurality of small circle drawing spaces according to the value of the first parameter. The intersection between the small circles at the time of drawing is obtained for each small circle drawing space, and the true circle is selected by selecting the small circle drawing space corresponding to the true visual axis direction based on the information on the number of small circles intersecting at the intersection. And the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to a plurality of small circles intersecting at the intersection obtained for the small circle drawing space corresponding to the true visual axis direction, and / or Alternatively, the fifth step of obtaining the normalized shortest distance _n d _s0 for the measurement plane may be performed.
In this case, the measurement point appearing on the image has intensity information, and the fourth step obtains the small circle, and the obtained small circle corresponds to the small circle. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in a circle drawing space, and the fifth step is for locating the intersection. Instead of obtaining, a local maximum point at which the value by voting during the first to fourth steps is repeated a plurality of times is determined for each small circle drawing space, and based on the local maximum information at the local maximum point. The true visual axis direction is obtained by selecting the small circle drawing space corresponding to the true visual axis direction, and the maximum point obtained for the small circle drawing space corresponding to the true visual axis direction is voted The compound corresponding to the participating small circles Orientation n _so the measurement plane including a measurement _point, and / or is preferably a step of obtaining a normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
A sixth step of setting a binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a third parameter;
In the second step, the position p _{axis of the} infinity point set in the first step, the normalized shortest distance _n d _s set in the second step, and the measurement point The step of obtaining the radius R using the measurement position p _R when observed from one of the two observation points and the binocular parallax σ set in the fifth step. ,
The fourth step obtains the small circle corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and corresponds the obtained small circle to the small circle. Voting the response intensity corresponding to the binocular parallax σ of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in the small circle drawing space;
For a plurality of measurement points in the measurement space, the third step and the fourth step are repeated a plurality of times while changing each value of each parameter in the first, second and sixth steps,
Subsequent to the fifth step being executed, instead of obtaining the intersection, the value obtained by voting during the repetition of the first, second, sixth, third and fourth steps becomes a maximum value. A local maximum point is determined for each small circle drawing space, and the true visual axis direction is determined by selecting a small circular drawing space corresponding to the true visual axis direction based on local maximum information at the local maximum point. Direction n _{so of} a measurement plane including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of local maximum points obtained for the small circle drawing space corresponding to the true visual axis direction, and / or the measurement It is also a preferred aspect that it is a step of _{obtaining the} normalized shortest distance _n d _s0 for the plane.

Furthermore, a twentieth image measurement method of the image measurement methods of the present invention is an image in the measurement space that appears in an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. The measurement positions of any of the measurement points when observed from each of the two observation points are p _R , p _L , and include the measurement points, respectively, and are the same as the visual axis direction v connecting the two observation points. When the position of the infinity point on the straight line extending in the direction is p _axis ,
The three positions _{p axis,} using p R, _{p L,} single-ratio determined by the _{(p axis p R p L)} or single-ratio and equivalent operation, and the measuring point, one of the two observation points It is characterized in that a physical quantity indicating the distance between one observation point is obtained.

Here, in the twentieth image measurement method, the single ratio (p _axis p _R p _L ) or an operation equivalent to the single ratio is calculated as 2 when the measurement point is observed from each of the two observation points. Instead of the two measurement positions p _R and p _L , the measurement position p _R when the measurement point is observed from one of the two observation points, and the two observation points of the measurement point Calculation using the binocular parallax σ which is the difference in position between the two measurement positions p _R and p _L when observed from FIG.
In the twentieth image measurement method, as a physical quantity indicating the distance, the distance between the measurement point and one of the two observation points is d ₀ , and the distance between the two observation points is when the distance was set to Δx _LR,
n d ₀ = d ₀ / Δx _LR
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _axis p _R p _L )
Or a relational expression equivalent to the relational expression.

In addition, a twenty-first image measurement method of the image measurement methods of the present invention is an image in the measurement space that appears in an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. A physical quantity that indicates a distance in a visual axis direction connecting the two observation points between a measurement plane including an arbitrary measurement point and one of the two observation points; a first step of setting the coordinates of the voting space defined by the azimuth n _s as a parameter,
A measurement position p _R when the measurement point is observed from one of the two observation points, and an infinite point on a straight line including the measurement point and extending in the same direction as the visual axis direction. From the position p _axis and the coordinates in the voting space set in the first step, between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points. A second step of obtaining a binocular parallax σ which is a position difference;
A third step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

A twenty-second image measurement method of the image measurement methods of the present invention observes the inside of a predetermined measurement space, and sets the visual axis direction v connecting two predetermined observation points in the measurement space to the first. By setting it as a parameter, on the straight line that includes any measurement point in the measurement space that appears in the image when the measurement space is viewed from the two observation points, and extends in the same direction as the visual axis direction A first step of setting a position p _{axis of the} infinity point;
A physical quantity indicative of the distance of the visual axis direction between one of the observation point of said the measurement plane including a measurement point the two observation points, defined by the azimuth n _s of the measurement plane, the A second step of setting the coordinates in the voting space according to the first parameter as a second parameter;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. Further, from the coordinates in the voting space, a third binocular parallax σ that is a difference between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is obtained. And the steps
A fourth step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

A twenty-third image measurement method of the image measurement methods of the present invention observes the inside of a predetermined measurement space, one observation point of the two predetermined observation points in the measurement space, and the two in the One observation point and a physical quantity that indicates the shortest distance between the measurement plane including the arbitrary measurement point within the measurement space appeared in the image when viewing the said measurement space, the azimuth n _s of the measurement plane A first step of setting coordinates in a prescribed voting space as parameters;
A measurement position p _R when the measurement point is observed from one of the two observation points and the visual axis direction that includes the measurement point and connects the two observation points. Two measurement positions p of the measurement point observed from the two observation points, based on the position p _{axis of the} infinity point on the straight line and the coordinates in the voting space set in the first step. _A second step of obtaining a binocular parallax σ which is a difference in position between _{R 1} and p _L ;
A third step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

According to a twenty-fourth image measurement method of the image measurement methods of the present invention, a visual axis direction v linking two predetermined observation points in the measurement space is observed in the first measurement space. By setting as a parameter, an infinite line on a straight line that includes any measurement point in the measurement space that appears in the image when the measurement space is viewed from the two observation points and extends in the same direction as the visual axis direction. A first step of setting a far point position p _axis ;
And one of the observation point of said two observation points, the physical quantity that indicates the shortest distance to the measurement plane including a measurement point is defined by the azimuth n _s of the measurement plane, said first parameter A second step of setting the corresponding coordinates in the voting space as a second parameter;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. Further, from the coordinates in the voting space, a third binocular parallax σ that is a difference between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is obtained. And the steps
A fourth step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

A twenty-fifth image measurement method of the image measurement methods of the present invention is a measurement space that appears in two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. A first step of setting, as a parameter, binocular parallax σ which is a difference in position between two measurement positions p _R and p _L at any measurement point
A measurement position p _R when the measurement point is observed from one of the two observation points, and the visual axis direction that includes the measurement point and connects the two observation points. From the position p _{axis of the} infinity point on the straight line and the binocular parallax σ set in the first step, the measurement plane including the measurement point and one observation point of the two observation points a physical quantity which indicates the, the distance of the visual axis direction between, a second step of obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
A third step of obtaining a response intensity corresponding to the binocular parallax σ set in the first step based on two images when the measurement space is viewed from the two observation points. When,
A fourth step of voting the response strength determined in the third step to the coordinates determined in the second step in the voting space;
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

According to a twenty-sixth image measurement method of the image measurement methods of the present invention, the visual axis direction v linking two predetermined observation points in the measurement space is observed in the first measurement space. By setting as a parameter, an infinite line on a straight line that includes any measurement point in the measurement space that appears in the image when the measurement space is viewed from the two observation points and extends in the same direction as the visual axis direction. A first step of setting a far point position p _axis ;
A second step of setting, as a second parameter, binocular parallax σ which is a difference between two measurement positions p _R and p _L when the measurement point is observed from the two observation points;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. From the binocular parallax σ, the physical quantity indicating the distance in the visual axis direction between the measurement plane including the measurement point and one of the two observation points, and the orientation of the measurement plane is defined by a n _s, a third step of obtaining the coordinates of the voting space corresponding to the first parameter,
Fourth step of obtaining response intensity corresponding to the binocular parallax σ set in the second step based on two images when the measurement space is viewed from the two observation points. When,
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

A twenty-seventh image measurement method of the image measurement methods of the present invention is a measurement space that appears in two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. A first step of setting, as a parameter, binocular parallax σ which is a difference in position between two measurement positions p _R and p _L at any measurement point
A measurement position p _R when the measurement point is observed from one of the two observation points and the visual axis direction that includes the measurement point and connects the two observation points. Measurement including one observation point of the two observation points and the measurement point based on the position p _{axis of the} infinity point on the straight line and the binocular parallax σ set in the first step a physical quantity which indicates the shortest distance between the plane and a second step of obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
A third step of obtaining a response intensity corresponding to the binocular parallax σ set in the first step based on two images when the measurement space is viewed from the two observation points. When,
Voting the response strength determined in the third step to the coordinates determined in the second step in the voting space; and
For a plurality of measurement points in the measurement space, the second to fourth steps of the first to fourth steps are executed a plurality of times while changing parameter values in the first step. It is characterized by that.

Furthermore, the twenty-eighth image measurement method of the image measurement methods of the present invention observes the inside of a predetermined measurement space, and sets the visual axis direction v connecting the two predetermined observation points in the measurement space to the first. By setting as a parameter, the arbitrary measurement points in the measurement space appearing in the image when the measurement space is viewed from the two observation points, and on a straight line extending in the same direction as the visual axis direction A first step of setting a position p _{axis of the} infinity point;
A second step of setting, as a second parameter, binocular parallax σ which is a difference between two measurement positions p _R and p _L when the measurement point is observed from the two observation points;
The measurement position p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. Based on the binocular parallax σ, the physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the direction n _{s of the} measurement plane A third step for obtaining coordinates in the voting space according to the first parameter, defined by:
Fourth step of obtaining response intensity corresponding to the binocular parallax σ set in the second step based on two images when the measurement space is viewed from the two observation points. When,
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. It is characterized by being executed multiple times.

The twenty-ninth image measurement method of the image measurement methods of the present invention is based on two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. A first step of obtaining a response intensity corresponding to binocular parallax, which is a difference between positions of two measurement positions when observing from any of the two observation points at an arbitrary measurement point in the measurement space;
The visual axis direction connecting the two observation points between the measurement plane including the measurement points and one observation point of the two observation points with the response intensity obtained in the first step. A second step of voting to coordinates corresponding to the measurement point and the binocular parallax in a voting space defined by a physical quantity that indicates the distance of the measurement plane and the orientation of the measurement plane;
The first step and the second step are executed a plurality of times for a plurality of measurement points in the measurement space.

  Here, in the twenty-ninth image measurement method, a measurement plane including a plurality of measurement points participating in the voting of the maximum point by obtaining a maximum point in the voting space where the value by the voting becomes a maximum value. And / or a third step of obtaining a physical quantity indicating a distance in the visual axis direction between the measurement plane and one of the two observation points. Also good.

A thirtieth image measurement method of the image measurement methods of the present invention is a method for setting, as a parameter, a visual axis direction connecting two predetermined observation points in the measurement space for observing the predetermined measurement space. 1 step,
Based on two images when the inside of the measurement space is viewed from the two observation points, a position between two measurement positions of the arbitrary measurement point in the measurement space when observed from the two observation points A second step of obtaining a response intensity corresponding to binocular parallax that is a difference between
The response intensity obtained in the second step is a physical quantity indicating the distance in the visual axis direction between the measurement plane including the measurement point and one of the two observation points. A third step of voting on coordinates corresponding to the measurement point and the binocular parallax in a voting space defined by the orientation of the measurement plane and according to the parameter set in the first step And
Executing the second and third steps of the first to third steps for a plurality of measurement points in the measurement space a plurality of times while changing parameter values in the first step. It is characterized by.

  Here, in the thirtieth image measurement method, a maximum point at which the value by voting becomes a maximum value is obtained for each voting space, and the measurement point is relative to the observation point based on information on the maximum value at the maximum point. The voting space corresponding to the true visual axis direction is selected to obtain the true visual axis direction, and a plurality of participants who participated in the voting of the maximum points obtained for the voting space corresponding to the true visual axis direction 4th which calculates | requires the physical quantity which indexes the direction of the measurement plane containing this measurement point and / or the distance of the said visual axis direction between this measurement plane and one observation point of said two observation points It may have a step.

The thirty-first image measurement method of the image measurement methods of the present invention is based on two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. A first step of obtaining a response intensity corresponding to a binocular parallax σ that is a difference between positions of two measurement positions when observing from two observation points at an arbitrary measurement point in space;
The response intensity obtained in the first step is calculated by using a physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the measurement plane Voting to coordinates corresponding to the measurement point and the binocular parallax σ in a voting space defined by an orientation,
The first step and the second step are executed a plurality of times for a plurality of measurement points in the measurement space.

Here, in the thirty-first image measuring method, by obtaining a local maximum point in the voting space where the value by the voting becomes a local maximum value, a measurement plane including a plurality of measuring points participating in the local maximum voting is obtained. There may be provided a third step of obtaining a physical quantity that indicates the shortest distance between the azimuth _ns and / or one of the two observation points and the measurement plane.

Furthermore, a thirty-second image measurement method of the image measurement methods of the present invention observes the inside of a predetermined measurement space, and sets the visual axis direction connecting two predetermined observation points in the measurement space as a parameter. A first step;
Based on two images when the inside of the measurement space is viewed from the two observation points, a position between two measurement positions of the arbitrary measurement point in the measurement space when observed from the two observation points A second step of obtaining a response intensity corresponding to binocular parallax that is a difference between
The response intensity obtained in the second step is obtained by using a physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the measurement plane. And a third step of voting on coordinates corresponding to the measurement point and the binocular parallax in the voting space according to the parameter set in the first step, defined by the azimuth,
Executing the second and third steps of the first to third steps for a plurality of measurement points in the measurement space a plurality of times while changing parameter values in the first step. It is characterized by.

  Here, in the thirty-second image measurement method, a local maximum point at which the value obtained by voting becomes a local maximum value is obtained for each voting space, and the measurement point is relative to the observation point based on information on the local maximum value at the local maximum point. The true visual axis direction is determined by selecting a voting space corresponding to the true visual axis direction, and a plurality of voting points determined for the voting space corresponding to the true visual axis direction A fourth step of obtaining a physical quantity indicating the azimuth of the measurement plane including the measurement point and / or the shortest distance between one of the two observation points and the measurement plane; Also good.

  The image measurement method of the present invention is also expressed as follows.

(1) of an arbitrary point in the image, the position of the current time p _0, p ₁ the position of the next _time, an infinite time position after lapse p _inf, and "plane camera center the point rests the position of the time "crossing when the _{p c} a standard up using a cross ratio determined by these four locations _{{p inf p 0 p 1 p} c}, across the three-dimensional orientation _{n s} and the plane of the plane A method for obtaining the conversion time _n t _c . Here, the normalization time is a time obtained by normalizing the time t _c until crossing the plane with the time difference Δt from the present time to the next time, and is represented by the following equation.

_n t _c = t _c / Δt (a)
(2) A method of obtaining the normalized time _n t _c until crossing the plane using the following relational expression.

_{n t c = {p inf p} 0 p 1 p c} (b)
(2-1) for a plurality of points in the image to polar transformation the position p _c for "time plane in terms of their rests crosses camera center" (or dual conversion), the polar resultant method of determining the three-dimensional orientation n _s of the plane as the intersection.

(2-2) How to know the "above the position _{p 0, p 1, p inf} at three time" normalized time _n t _c and calculates the position _{p c} using Equation (b). This method will be referred to as a cross-ratio conversion from using cross ratio _{{p inf p 0 p 1 p} c}.

(3) In (1), for each point in the image position p _c for "time when the point is on the plane crosses the center of the camera" determined by performing a multi-ratio conversion of claims 2-1, then the way they position as the intersection of polar that polar transformation to obtain a three-dimensional orientation n _s plane.

(4) In (3), a method for obtaining the three-dimensional orientation n _s0 of the plane and the normalized time _n t _c0 until crossing the plane by executing the following steps.

Step 1: The normalization time parameter _n t _c is arbitrarily set, and the position _pc is calculated by performing the cross ratio conversion of (2-2) for each point in the image.

Step 2: Next, polar positions of these positions are converted to draw a plurality of corresponding polar lines. The intensity of the polar line is “brightness at position p ₀ in the image”, and the intensity obtained by adding these intensities at a place where a plurality of polar lines intersect.

Step 3: The above steps are performed by changing the normalized time parameter _n t _c, and a parameter value _n t _{c0 at} which the plurality of polar lines drawn in step 2 intersect at one point is obtained. As the parameter value, “normalized time _n t _c0 until crossing the plane” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection.

  (5) The method of drawing the polar line in step 2 as a great circle on the spherical surface in (4).

  (5-1) A method of projecting and drawing the great circle in (5) onto the inside of a circle on a plane.

  (6) A method of drawing the polar line in step 2 as a straight line on a plane in (4).

(7) (3), (4), the moving direction v (i.e., _{p inf} position after lapse infinite time) without knowing, to traverse the three-dimensional orientation _{n s} and planes of the following steps A method for obtaining the normalized time _n t _c .

  Step 1: The moving direction parameter v is arbitrarily set.

Step 2: The direction of the parameter v is set to “position p _inf after elapse of infinite time”.

  Step 3: (3) or (4) is executed.

Step 4: The above steps are performed by changing the parameter v, and the parameter value v _{0 at} which the polar line drawn in Step 3 intersects at one point is obtained. This parameter value is the true moving direction v ₀ , and the plane direction _ns and the normalized time _n t _c until crossing the plane are obtained as the coordinates of the intersection.

(8) of an arbitrary point in the image, the position of the current time p _0, p ₁ the position of the next _time, an infinite time position after lapse p _inf, and "plane camera center the point rests when the p _c the position at time "across, using the relationship and relation of the formula (c) of the formula (b), normalized shortest distance to the 3-dimensional orientation n _s and the plane of the plane _n d _s How to ask. Here, the standardized shortest distance is a distance obtained by normalizing the shortest distance d _s to the plane by the distance Δx that the camera (or plane) has moved from the present time to the next time, and is expressed by Expression (d). . Further, v in the following expression is the moving direction of the camera or the plane.

_{n d s = n t c (} n s · v) (c)
_n d _s = d _s / Δx (d)
(9) In (8), a method for determining the normalized shortest distance _n d _s0 to 3D orientation _{n s0} and planes of running the following steps.

Step 1: The normalized shortest distance parameter _n d _s is arbitrarily set.

Step 2: arbitrarily set the angle r between the direction of movement v and the three-dimensional orientation _{n s} - i.e., the inner product of v and _{n s} a _{(n s · v) cos (} r) and optionally set - and, The normalized time parameter _n t _c is calculated as _n d _s / cos (r) using equation (c).

Step 3: For each point in the image, the position ratio _pc is calculated by performing the cross ratio conversion of claim 2-1.

  Step 4: Next, a point that forms an angle r with the moving direction v is obtained on the polar line obtained by converting the positions.

Step 5: Calculate the points in Step 4 while changing the angle r, and draw a curve composed of these points. The intensity of the curve is the "brightness of the position p ₀ in the image", in a place where a plurality of curves intersect the strength obtained by adding them strength.

Step 6: The above steps are performed by changing the normalized shortest distance parameter _n d _s, and a parameter value _n d _{s0 at} which the plurality of curves drawn in step 5 intersect at one point is obtained. As the parameter value, “normalized shortest distance _n d _s0 to the plane” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection.

(10) A method for obtaining the three-dimensional orientation n _s0 of the plane and the normalized shortest distance _n d _s0 to the plane by executing the following steps in (8).

Step 1: The normalized shortest distance parameter _n d _s is arbitrarily set, and “single ratio determined by three positions p ₀ , p ₁ , p _inf (p _inf p ₀ p ₁ )” is used for each point in the image. Then, the parameter R is calculated by the equation (e).

^{_{R = cos -1 (n d s}} / (p inf p 0 p 1)) (e)
Step 2: Next, a small circle having a radius R is drawn around the “current time position p ₀ ”. The intensity of the small circle is “brightness at position p ₀ in the image”, and the intensity obtained by adding these intensities at a place where a plurality of small circles intersect.

Step 3: The above steps are performed by changing the normalized shortest distance parameter _n d _s, and a parameter value _n d _{s0 at} which a plurality of small circles drawn in Step 2 intersect at one point is obtained. As the parameter value, “normalized shortest distance _n d _s0 to the plane” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection.

  (11) A method of drawing the curve of step 5 in (9) and the small circle of step 2 in claim 12 on a spherical surface.

  (11-1) A method of projecting and drawing the curve or small circle in (11) onto a circle on a plane.

  (12) A method of projecting and drawing the curve of step 5 in (9) and the small circle of step 2 in claim 12 on a plane.

(13) In (8), (9), and (10), even if the moving direction v (that is, the position p _inf after elapse of infinite time) is not known, the three-dimensional orientation _ns and the plane method of determining the normalized shortest distance _n d _s up.

  Step 1: The moving direction parameter v is arbitrarily set.

Step 2: The direction of the parameter v is set to “position p _inf after elapse of infinite time”.

  Step 3: (8) or (9) or (10) is executed.

Step 4: The above steps are performed by changing the parameter v, and the parameter value v _{0 at} which the curve (or small circle) drawn in Step 3 intersects at one point is obtained. This parameter value is true in the moving direction v _0, as the coordinates of the intersection, the normalized shortest distance _n d _s up orientation n _s and or the plane of the plane is determined.

(14) For an arbitrary point in the image, a unit ratio determined by these three positions, where p ₀ is the current time position, p _{1 is} the position of the next time, and p _inf is the position after the infinite time has elapsed. A method of obtaining a normalized distance _n d ₀ from the center of the camera to the position of a point in the space using (p _inf p ₀ p ₁ ) according to equation (f). Here, the standardized distance is a distance obtained by normalizing the distance d ₀ to the point by the distance Δx that the camera (or point) has moved from the present time to the next time, and is represented by the equation (g).

_n d ₀ = (p _inf p ₀ p ₁ ) (f)
_n d ₀ = d ₀ / Δx (g)
(15) A method of calculating the parameter R by the following equation using “standardized point distance _n d ₀ ” in Step 1 of (10).

R = cos ⁻¹ ( _n d _s / _n d ₀ ) (h)
(16) A method using the “difference (motion parallax) between the current position and the next time position” instead of the next time position in (1) to (15).

(17) In (1) to (7), the current time position p ₀ , the next time position p ₁ , the position p _inf after infinite time, the moving direction v, and the standardized time until crossing the plane the _n t _c, respectively, the position _{p R} in the right camera image, the position _{p L} in the left camera image, the position _{p axis} on visual axis connecting the left and right cameras, visual axis direction _{a xis,} and "visual axis direction in the plane Is replaced with the standardized distance _n d _c ″ until crossing the plane, and the three-dimensional orientation n _s of the plane and the standardized distance _n d _c until crossing the plane are obtained from the stereo image. Here, the normalized distance is the distance d _c to cross the plane visual axis direction, the distance normalized by the distance [Delta] x _LR between the right and left cameras. Here, the right camera and the left camera may be exchanged.

(18) (8) to (13), in (15), the position _{p 0} at the present time, the position _{p 1} at the next time, and infinite time position after lapse _{p inf,} movement direction v and up plane standard, the reduction shortest distance _n d _s, respectively, the position _{p R} in the right camera image, the position _{p L} in the left camera image, the position _{p axis} on visual axis connecting the left and right cameras, visual axis direction _{a xis,} and "stereo image The standardized shortest distance _n d _{s, stero} ”to the plane in FIG. 4 is used to obtain the three-dimensional orientation n _s of the plane and the standardized shortest distance _n d _s to the plane from the stereo image. Here, the normalized distance _n d _{s, stero} is a distance obtained _{by normalizing} the shortest distance d _s to the plane by the distance Δx _LR between the left and right cameras. Here, the right camera and the left camera may be exchanged.

(19) In (14), the position _{p 0} at the present time, the position _{p 1} at the next time, and the position _{p inf} after infinite time has elapsed, respectively, the position _{p R} in the right camera image of the left camera image A method of obtaining a “normalized distance _n d ₀ to the position of a point in space” from a stereo image by replacing the position p _L with “a position p _axis on the visual axis connecting the left and right cameras”. Here, the standardized distance is a distance obtained by normalizing the distance d ₀ to the point with the distance Δx _LR between the left and right cameras. Here, the right camera and the left camera may be exchanged.

(20) A method of using the “difference (binocular parallax) between the left camera image position and the right camera image position” instead of the position p _L in the left camera image in (17) to (19).

  (21) The method according to any one of (1) to (20), wherein an image taken by a flat camera is used as an input image.

  (22) The method according to any one of (1) to (20), wherein an image captured by a spherical camera is used as an input image.

  (23) The method according to claim 30, wherein a “difference (motion parallax) between a current position and a next time position (motion parallax)” is obtained from a planar camera image, and then the motion parallax is projected onto a spherical surface.

  (24) The method according to claim 43, wherein a “difference (binocular parallax) between a left camera image position and a right camera image position” is obtained from a planar camera image, and then the binocular parallax is projected on a spherical surface.

(25) Based on “three-dimensional azimuth n _{s of} plane” and “normalized time _n t _c until crossing” measured by the method relating to (3) among (21), (22), and (23) A method for controlling robots, hobby equipment, and mobile machines such as automobiles and airplanes.

(26) “3-dimensional orientation n _{s of} plane” measured by the method related to (8) (or (3)) of (21), (22), and (23) and “normalized shortest distance _n to the plane” A method of “depth-separating multiple objects and environments that appear to overlap in an image” based on d _s (or normalized time _n t _c until crossing a plane) ”.

(27) Among the (21), (22), and (23), the “three-dimensional orientation n _{s of the} plane” measured by the method related to (18) or (17) and the “normalized shortest distance _n d _s to the plane” (or visual axis direction normalized distance _{n d} c to plane) _"based on the" method of depth separating "multiple objects and environments seen overlap in the image.
(I) Forward method (I-1) Normalization time In (4) of (16), a method for voting the response intensity obtained by the method (or device) for detecting motion parallax.
(I-2) Normalization time + v unknown Method of voting the response intensity obtained by the method (or device) for detecting motion parallax in (7) and (4) of (16).
(I-3) Standardized shortest distance In (10) of (16), a method for voting the response intensity obtained by the method (or device) for detecting motion parallax.
(I-4) Normalized shortest distance + v unknown In (13) and (10) of (16), a method for voting the response intensity obtained by the method (or device) for detecting motion parallax.
(I-5) Stereo + standardized distance In (17) and (4) of (20), a method of voting the response intensity obtained by the method (or device) for detecting binocular parallax.
(I-6) Stereo + standardized distance + a _axis unknown In (17) and (7) and (4) of (20), the response intensity obtained by the method (or device) for detecting binocular parallax is How to vote.
(I-7) Stereo + standardized shortest distance A method for voting the response intensity obtained by the method (or device) for detecting binocular parallax in (18) and (10) of (20).
(I-8) Stereo + Normalized shortest distance + a _axis unknown Response intensity obtained by the method (or apparatus) for detecting binocular parallax in (18), (13) and (10) of (20) How to vote.
(I-9) Standardized shortest distance In (9) of (16), a method for voting the response intensity obtained by the method (or device) for detecting motion parallax.
(I-10) Normalized shortest distance + v unknown In (13) and (9) of (16), a method for voting the response intensity obtained by the method (or device) for detecting motion parallax.
(I-11) Stereo + standardized shortest distance In (18) and (9) of (20), a method for voting the response intensity obtained by the method (or device) for detecting binocular parallax.
(I-12) Stereo + Normalized shortest distance + a _xis unknown Response intensity obtained by the method (or apparatus) for detecting binocular parallax in (18), (13) and (9) of (20) How to vote.
(II) Reverse Direction Method (II-1) Normalization Time Step 1: Consider the number i of an arbitrary pixel _i p ₀ in the image at the current time.

Step 2: The three (two degrees of freedom of a plane orientation vector _{n s,} l freedom of normalized time _n t _c) freedom SEQ consider any element _{(n sj,} n _{t cj)} number j in.

Step 3: Calculate “motion parallax _ij τ at pixel _i p ₀ ” corresponding to the numbers i and j.

Step 4: The response intensity at the motion parallax _ij τ is calculated from the input image by a method (or apparatus) for detecting motion parallax.

Step 5: Vote for the response strength to the elements (n _sj , _n t _cj ) in the 3- _DOF array.

  Step 6: The above is repeated for a predetermined range of pixels i and j.

The element having the maximum response in the voted three-degree-of-freedom array is detected, and the “plane orientation and standard time” can be detected from the address (n _s0 , _n t _c0 ).
(II-2) Standard time + v unknown In (II-1), even if the moving direction V (that is, the position p _inf after elapse of infinite time) is not known, the three-dimensional orientation _ns and the plane Of obtaining the normalized time _n t _c until crossing

  Step 1: The moving direction parameter v is arbitrarily set.

  Step 2: (II-1) is executed.

  Step 3: The above steps are performed by changing the parameter v.

When the parameter value V ₀ that gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true movement direction V ₀ can be detected as the parameter value. Further, it is possible to detect the three degrees of freedom sequence (Step 2) address that responds maximum in _{(n s0, n t c0)} "plane orientation and the normalized time" from.
(II-3) Normalized shortest distance Step 1: Consider the number i of an arbitrary pixel _i p ₀ in the image at the current time.

Step 2: In addition, consider the number j of the three degrees of freedom sequences (two degrees of freedom of a plane orientation vector _{n s,} l freedom of normalized shortest distance _n d _s) any element _{(n sj,} n _{d sj)} in .

Step 3: Calculate “motion parallax _ij τ at pixel _i p ₀ ” corresponding to the numbers i and j.

Step 4: The response intensity at the motion parallax _ij τ is calculated from the input image by a method (or apparatus) for detecting motion parallax.

Step 5: Vote the response strength to the elements (n _sj , _n d _sj ) in the 3- _DOF array.

  Step 6: The above is repeated for a predetermined range of pixels i and j.

The element having the maximum response is detected in the voted three-degree-of-freedom array, and the “plane orientation and standardized shortest distance” can be detected from the address (n _s0 , _n d _s0 ).
(II-4) Normalized shortest distance + v unknown In (II-3), even if the moving direction v (that is, the position p _inf after elapse of infinite time) is not known, the three-dimensional orientation n _s of the plane in the following steps And a method for obtaining the normalized shortest distance _n d _s .

  Step 1: The moving direction parameter v is arbitrarily set.

  Step 2: (II-3) is executed.

  Step 3: The above steps are performed by changing the parameter v.

When the parameter value v ₀ that gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true movement direction v ₀ can be detected as the parameter value. In addition, the “plane orientation and normalized shortest distance” can be detected from the address (n _s0 , _n d _s0 ) that responds to the maximum in the three-degree-of-freedom array (step 2).
(II-5) Stereo + standardized distance Step 1: Consider the number i of an arbitrary pixel _i p _R in the right camera image.

Step 2: In addition, consider three degrees of freedom sequences (two degrees of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized distance _n d _c) any element _{(n sj,} n _{d cj)} in the number j.

Step 3: Calculate “binocular parallax _ij σ at pixel _i p _R ” corresponding to the numbers i and j.

Step 4: The response intensity at the binocular parallax _ij σ is calculated from the input image by the method (or apparatus) for detecting the binocular parallax.

Step 5: Vote for the response strength to the elements (n _sj , _n d _cj ) in the 3- _DOF array.

  Step 6: The above is repeated for a predetermined range of pixels i and elements j.

Detecting elements for the maximum response of the 3 DOF sequences found above, it can detect the address (n _s0, n _{d c0)} from "plane orientation and the normalized distance".
(II-6) Stereo + standardized distance + a _xis unknown In (II-5), even if the visual axis direction a _xis (that is, the position p _axis on the visual axis connecting the left and right cameras) is not known, the following steps are performed. 3D orientation _{n s} and a method for determining the normalized distance _n d _c plane.

Step 1: The visual axis direction parameter a _xis is arbitrarily set.

  Step 2: (II-5) is executed.

Step 3: The above steps are performed by changing the parameter _axis .

When the parameter value _axis0 that _gives the maximum response is obtained from the three-degree-of-freedom array voted above, the true visual axis direction _axis0 can be detected as the parameter value. Further, it is possible to detect the three degrees of freedom sequence (Step 2) address that responds maximum in _{(n s0, n d c0)} from "plane orientation and the normalized distance".
(II-7) Stereo + standardized shortest distance Step 1: Consider number i of an arbitrary pixel _i p _R in the right camera image.

Step 2: In addition, consider 3 (two degrees of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized shortest distance _n d _s) freedom SEQ any element _{(n sj,} n _{d sj)} in the number j .

Step 3: “Binocular disparity _ij σ at pixel _i p _R corresponding to the numbers i and j is calculated.

Step 4: The response intensity at the binocular parallax _ij σ is calculated from the input image by a method (or apparatus) for detecting binocular parallax.

Step 5: Vote the response strength to the elements (n _sj , _n d _sj ) in the 3- _DOF array.

  Step 6: The above is repeated for a predetermined range of pixels i and elements j.

The element having the maximum response is detected in the voted three-degree-of-freedom array, and the “plane orientation and standardized shortest distance” can be detected from the address (n _s0 , _n d _s0 ).
(II-8) Stereo + standardized shortest distance + a _xis unknown (II-7), the following steps can be performed without knowing the visual axis direction a _xis (that is, the position p _axis on the visual axis connecting the left and right cameras). in 3-dimensional orientation _{n s} and a method for determining the normalized shortest distance _n d _s plane.

Step 1: The visual axis direction parameter a _xis is arbitrarily set.

  Step 2: (II-7) is executed.

Step 3: The above steps are performed by changing the parameter _axis .

When the parameter value _axis0 that _gives the maximum response is obtained from the three-degree-of-freedom array voted above, the true visual axis direction _axis0 can be detected as the parameter value. In addition, the “plane orientation and normalized shortest distance” can be detected from the address (n _s0 , _n d _s0 ) that responds to the maximum in the three-degree-of-freedom array (step 2).

(III) Compound Method (III-1) Normalization Time Step 1: Consider number i of an arbitrary pixel _i p ₀ in the image at the current time.

Step 2: Consider an arbitrary motion parallax _k τ number k.

Step 3: those corresponding to the numbers i and k "3 DOF sequence (two degrees of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized time _n t _c) in the element group _{{(n sj, n} t _cj )} ".

Step 4: The response intensity at the motion parallax _k τ is calculated from the input image by a method (or device) for detecting motion parallax.

Step 5: Vote the response strength for the above element group {(n _sj , _n t _cj )}.

  Step 6: The above is repeated for a predetermined range of i and k.

The element having the maximum response is detected in the voted three-degree-of-freedom array, and the “plane orientation and normalized time” can be detected from the address (n _s0 , _n t _c0 ).
(III-2) Normalization time + v unknown In (III-1), even if the moving direction v (that is, the position p _inf after elapse of infinite time) is not known, the three-dimensional orientation _ns and the plane Of obtaining the normalized time _n t _c until crossing

  Step 1: The moving direction parameter v is arbitrarily set.

  Step 2: (III-1) is executed.

  Step 3: The above steps are performed by changing the parameter v.

When the parameter value v ₀ that gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true movement direction v ₀ can be detected as the parameter value. Further, it is possible to detect the three degrees of freedom sequence (Step 2) address that responds maximum in _{(n s0, n t c0)} "plane orientation and the normalized time" from.
(III-3) Normalized shortest distance Step 1: Consider the number i of an arbitrary pixel _i p ₀ in the image at the current time.

Step 2: Consider an arbitrary motion parallax _k τ number k.

Step 3: (2 degrees of freedom, one degree of freedom of the normalized shortest distance _n d _s of planar orientation vector _{n s)} in element group thereof number i and corresponding to the k "3 DOF sequence _{{(n sj, n} d _sj )} ”.

Step 4: The response intensity at the motion parallax _k τ is calculated from the input image by a method (or device) for detecting motion parallax.

Step 5: Vote the response strength for the above element group {(n _sj , _n d _sj )}.

  Step 6: The above is repeated for a predetermined range of i and k.

The element having the maximum response is detected in the voted three-degree-of-freedom array, and the “plane orientation and standardized shortest distance” can be detected from the address (n _s0 , _n d _s0 ).
(III-4) Normalized shortest distance + v unknown In (III-3), even if the moving direction v (that is, the position p _inf after elapse of infinite time) is not known, the three-dimensional orientation n _s of the plane in the following steps And a method for obtaining the normalized shortest distance _n d _s .

  Step 1: The moving direction parameter v is arbitrarily set.

  Step 2: (III-2) is executed.

  Step 3: The above steps are performed by changing the parameter v.

When the parameter value v ₀ that gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true movement direction v ₀ can be detected as the parameter value. In addition, the “plane orientation and normalized shortest distance” can be detected from the address (n _s0 , _n d _s0 ) that responds to the maximum in the three-degree-of-freedom array (step 2).
(III-5) Stereo + standardized distance Step 1: Consider number i of an arbitrary pixel _i p _R in the right camera image.

Step 2: Consider the number k of any binocular parallax _k sigma.

Step 3: they correspond to the numbers i and k "3 (2 degree of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized distance _n d _c) freedom array in the element group _{{(n sj, n} d _cj )} ".

Step 4: a method for detecting a binocular parallax (or device) by, calculating the response intensity in the binocular disparity _k sigma from the input image.

Step 5: Vote the response strength for the above element group {(n _sj , _n d _cj )}.

  Step 6: The above is repeated for a predetermined range of i and k.

Detecting elements for the maximum response of the 3 DOF sequences found above, it can detect the address (n _s0, n _{d c0)} from "plane orientation and the normalized distance".
(III-6) Stereo + standardized distance + a _axis unknown In (III-5), even if the visual axis direction a _xis (that is, the position p _axis on the visual axis connecting the left and right cameras) is not known, the following steps are performed. 3D orientation _{n s} and a method for determining the normalized distance _n d _c plane.

Step 1: The visual axis direction parameter a _xis is arbitrarily set.

  Step 2: (III-5) is executed.

Step 3: The above steps are performed by changing the parameter _axis .

When the parameter value _axis0 that _gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true visual axis direction _axis00 can be detected as the parameter value. Further, it is possible to detect the three degrees of freedom sequence (Step 2) address that responds maximum in _{(n s0, n d c0)} from "plane orientation and the normalized distance".
(III-7) Stereo + standardized shortest distance Step 1: Consider number i of an arbitrary pixel _i p _k in the right camera image.

Step 2: Consider the number k of any binocular parallax _k sigma.

Step 3: (2 degrees of freedom, one degree of freedom of the normalized shortest distance _n d _s of planar orientation vector _{n s)} in element group thereof number i and corresponding to the k "3 DOF sequence _{{(n sj, n} d _sj )}.

  Step 4: The response intensity at the binocular parallax kσ is calculated from the input image by the method (or apparatus) for detecting the binocular parallax.

Step 5: Vote the response strength for the above element group {(n _sj , _n d _sj )}.

  Step 6: The above is repeated for a predetermined range of i and k.

The element having the maximum response is detected in the voted three-degree-of-freedom array, and the “plane orientation and standardized shortest distance” can be detected from the address (n _s0 , _n d _s0 ).
(III-8) Stereo + standardized shortest distance + a _axis unknown (III-7) Even if the visual axis direction a _xis (that is, the position p _axis on the visual axis connecting the left and right cameras) is not known, the following steps in 3-dimensional orientation _{n s} and a method for determining the normalized shortest distance _n d _s plane.

Step 1: The visual axis direction parameter a _xis is arbitrarily set.

  Step 2: (III-7) is executed.

Step 3: The above steps are performed by changing the parameter _axis .

When the parameter value _axis0 that _gives the maximum response is obtained from the three-degree-of-freedom array voted above, the true visual axis direction _axis0 can be detected as the parameter value. In addition, the “plane orientation and normalized shortest distance” can be detected from the address (n _s0 , _n d _s0 ) that responds to the maximum in the three-degree-of-freedom array (step 2).

To (IV) Generalized (IV-1) normalized time of 3 degrees of freedom sequences (two degrees of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized time _n t _c), a method of detecting a motion parallax (or A method for voting the response strength obtained by the device. By detecting the element having the maximum response in the array, it is possible to detect the “plane orientation and normalization time” from the address (n _s0 , _n t _c0 ).

In all of the present invention, “voting the response intensity obtained by the method (or apparatus) for detecting parallax” may be “voting an amount related to the luminance of the input image”. Further, the “method (or device) for detecting parallax” may be a “method (or device) for detecting the speed in an image”.
(IV-2) Normalization time + v unknown Step 1: The moving direction parameter v is arbitrarily set.

  Step 2: (IV-1) is executed.

  Step 3: The above steps are performed by changing the parameter v.

When the parameter value v ₀ that gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true movement direction v ₀ can be detected as the parameter value. Further, it is possible to detect the three degrees of freedom sequence (Step 2) address that responds maximum in _{(n s0, n t c0)} "plane orientation and the normalized time" from.
(IV-3) normalized shortest distance 3 degrees of freedom sequences (two degrees of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized shortest distance _n d _s), the method of detecting a motion parallax (or device) A method of voting the response strength obtained by By detecting the element having the maximum response in the array, it is possible to detect the “plane orientation and normalized shortest distance” from the address (n _s0 , _n d _s0 ).
(IV-4) Normalized shortest distance + v unknown Step 1: The moving direction parameter v is arbitrarily set.

  Step 2: (IV-3) is executed.

  Step 3: The above steps are performed by changing the parameter v.

When the parameter value v ₀ that gives the maximum response in the three-degree-of-freedom array voted above is obtained, the true movement direction v ₀ can be detected as the parameter value. In addition, the “plane orientation and normalized shortest distance” can be detected from the address (n _s0 , _n d _s0 ) that responds to the maximum in the three-degree-of-freedom array (step 2).
(Two degrees of freedom of a plane orientation vector _{n s,} 1 degree of freedom normalized distance _{n d c) (IV-5} ) Stereo + normalized distance 3 degrees of freedom array, a method of detecting a binocular parallax (or device The method of voting the response intensity obtained by By detecting the element having the maximum response in the array, the “plane orientation and normalized distance” can be detected from the address (n _s0 , _n d _c0 ).
(IV-6) Stereo + standardized distance + a _xis unknown Step 1: The visual axis direction parameter a _xis is arbitrarily set.

  Step 2: (IV-5) is executed.

Step 3: The above steps are performed by changing the parameter _axis .

When the parameter value _axis0 that _gives the maximum response is obtained from the three-degree-of-freedom array voted above, the true visual axis direction _axis0 can be detected as the parameter value. Further, it is possible to detect the three degrees of freedom sequence (Step 2) address that responds maximum in _{(n s0, n d c0)} from "plane orientation and the normalized distance".
(IV-7) to the stereo + normalized shortest distance 3 degrees of freedom sequences (two degrees of freedom of a plane orientation vector n _s, 1 degree of freedom normalized shortest distance _n d _s), a method for detecting a binocular parallax (or A method for voting the response strength obtained by the device. By detecting the element having the maximum response in the array, it is possible to detect the “plane orientation and normalized shortest distance” from the address (n _s0 , _n d _s0 ).
(IV-8) Stereo + standardized shortest distance + a _xis unknown Step 1: The visual axis direction parameter a _xis is arbitrarily set.

  Step 2: (IV-7) is executed.

Step 3: The above steps are performed by changing the parameter _axis .

When the parameter value _axis0 that _gives the maximum response is obtained from the three-degree-of-freedom array voted above, the true visual axis direction _axis0 can be detected as the parameter value. In addition, the “plane orientation and normalized shortest distance” can be detected from the address (n _s0 , _n d _s0 ) that responds to the maximum in the three-degree-of-freedom array (step 2).

The first image measurement device of the image measurement device of the present invention is:
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , relative to the observation point, in the same direction as the movement direction v between the two measurement times, and the same as the movement speed between the two measurement times The position of the measurement point after infinite time has elapsed in the scheduled movement continuation state where the movement at the speed is continued as is, p _inf , and the measurement plane including the measurement point in the movement continuation state is superimposed on the observation point when the position of the measurement point in time was p _c,
The four positions _p inf measurement _point, p _0, p 1, using the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations determined by _{p c,} the orientation of the measurement plane, And / or a calculation unit that obtains a physical quantity that indicates a superposition time at which the measurement plane overlaps the observation point.

Here, in the above-described first image measurement device, said executed in the arithmetic unit, the compound ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations, the measurement point, the two measurement instead of the position p _0, p ₁ at two measurement time, the measurement point, the measurement position p ₀ at one measurement time of said two measurement time, the measurement point, the two Calculations using motion parallax τ, which is the difference in position between the two measurement positions p ₀ and p ₁ at the measurement time, are included.

In the first image measurement device, the arithmetic unit calculates a time between one measurement time of the two measurement times and the superposition time as a physical quantity indicating the superposition time by t _c , 2 When the time between two measurement times is Δt,
_{n t c} = t _c / _Δt
The normalized time _n t _c expressed by the following is adopted, and the normalized time _n t _c is
_{n t c = {p inf p} 0 p 1 p c}
Or a relational expression equivalent to the relational expression.

In the first image measurement device, the calculation unit includes:
A parameter changing unit that changes the value of the parameter with the physical quantity indicating the superposition time as a parameter;
Instead of the physical quantity indexed by the superimposition time set by the parameter changing unit and the measurement positions p ₀ and p _{1 at the} two measurement times of the measurement points, or these two measurement positions p ₀ and p ₁ . The motion parallax between the measurement position p ₀ of the measurement point at one of the two measurement times and the two measurement positions p ₀ and p _{1 of the} measurement point at the two measurement times. and tau, wherein a and a position _{p inf} of after infinite time in a mobile continuous state, using the equivalent operation with the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio, the superposition of the measuring points a compound ratio conversion unit for determining the position p _c at time,
A polar transformation unit for obtaining the polar corresponding to the measurement point by converting the measurement point, the position p _c in the superimposed time pole,
The plurality of measurement points obtained in the measurement space while each calculation in the multi-ratio conversion unit and the multi-ratio conversion unit is repeated a plurality of times while the parameter value is changed by the parameter setting unit. Of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines intersecting at the intersection, and / or And a detection unit that obtains a physical quantity indicating the superposition time at which the measurement plane overlaps the observation point.

  In that case, when the measurement point appearing on the image has intensity information, the polar conversion unit obtains the polar line and draws the obtained polar line in the polar drawing space. A value corresponding to the intensity of the measurement point corresponding to the polar line is voted for each point on the trajectory of the polar line, and the detection unit replaces the cross ratio conversion instead of obtaining the intersection. Voting of the maximum point by obtaining a maximum point where the value by voting becomes a maximum value while each of the operations in the unit and the maximum conversion unit is repeated a plurality of times while the value of the parameter is changed by the parameter changing unit It is preferable to obtain the orientation of a measurement plane including a plurality of measurement points corresponding to a plurality of polar lines that participated in and / or a physical quantity that indicates a superposition time at which the measurement plane overlaps the observation point.

In the first image measurement apparatus, the measurement point appearing on the image has intensity information,
The arithmetic unit changes the value of the second parameter using the motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a second parameter. With a parameter changer,
The multi-ratio conversion unit sets the physical quantity that is set by the parameter changing unit to index the superposition time, the measurement position p _{0 at} one of the two measurement times of the measurement point, and the first from said motion parallax τ set by the second parameter change unit, and the position p _inf after infinite time has elapsed in the mobile continuity state of the measurement point, and requests the position p _c in the superimposed time of the measurement point Yes,
The polar converter obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained polar line in the polar drawing space Voting the response intensity corresponding to the motion parallax τ of the measurement point corresponding to the polar line for each point on the trajectory of the polar line,
Instead of obtaining the intersection, the detection unit performs each calculation in the cross ratio conversion unit and the pole conversion unit on the plurality of measurement points in the measurement space in the parameter change unit and the second parameter change unit. Including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the maximum points by obtaining a maximum point where the value by voting during a plurality of repetitions while changing each value of each parameter becomes a maximum value It is also a preferred aspect to obtain a physical quantity that indicates the orientation of the measurement plane and / or the superposition time at which the measurement plane overlaps the observation point.

In the first image measurement device, the calculation unit includes:
A first parameter changing unit that changes the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state by changing the value of the first parameter using the movement direction v as a first parameter;
A second parameter changing unit that changes the value of the second parameter using the physical quantity indicating the superposition time as a second parameter;
The position p _inf set by the first parameter changing unit, the physical quantity indexed by the superposition time set by the second parameter changing unit, and each measurement position at the two measurement times of the measurement point p _0, p ₁ or the two measuring positions p _0, replaces the p _1, of the measurement point, the measurement position p ₀ and the measurement point at one measurement time of said two measurement time, the 2 one of the measuring two measuring positions _p at time 0, the _{p 1} if the motion parallax τ between, using the double ratio _{{p inf} p ₀ p 1 _{p c}} or plurality ratio equivalent calculation, the measurement a cross-ratio conversion unit for determining the position p _c in the superimposed time point,
A polar transformation unit for obtaining the polar corresponding to the measurement point by converting the measurement point, the position p _c in the superimposed time pole,
With respect to a plurality of measurement points in the measurement space, each calculation in the cross ratio conversion unit and the polar conversion unit is performed by using the first parameter and the second parameter as the first parameter and the second parameter, respectively. The plurality of polar lines obtained while being repeated a plurality of times while being changed by the two parameter changing units are respectively converted into corresponding polar line drawing spaces among the plurality of polar line drawing spaces according to the value of the first parameter. A pole corresponding to the true movement direction relative to the observation point of the measurement point based on information on the number of polar lines intersecting at the intersection, where intersections between polar lines at the time of drawing are obtained for each polar line drawing space The true movement direction is determined by selecting a line drawing space, and a plurality of polar lines that intersect at the intersections obtained for the polar drawing space corresponding to the true movement direction Azimuth measurement plane including a measurement point, and / or is preferably one that includes a detector for determining a physical quantity indicative of the superimposed time the measurement plane overlaps the observation point,
In that case, the measurement point appearing on the image has intensity information, and the polar converter obtains the polar line, and the obtained polar line corresponds to the polar line. A value corresponding to the intensity of the measurement point corresponding to the polar line is voted for each point on the trajectory of the polar line when drawn in the drawing space, and the detection unit determines the intersection. Instead, a value obtained by voting becomes a maximum value while each calculation in the cross ratio conversion unit and the maximum conversion unit is repeated a plurality of times while each value of the first parameter and the second parameter is sequentially changed. A local maximum point is obtained for each polar line drawing space, and the true moving direction is obtained by selecting a polar line drawing space corresponding to the true moving direction based on information on the local maximum value at the local maximum point. Corresponding to the direction of movement The direction of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines that participated in the voting of the maximum points obtained for the polar drawing space, and / or the superimposition time at which the measurement plane intersects the observation points is used as an index Preferably to determine the physical quantity to be
Measurement points appearing on the image have intensity information,
The arithmetic unit changes a value of the third parameter using a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter. With a parameter changer,
The multi-ratio conversion unit includes a position p _inf set by the first parameter changing unit, a physical quantity indicating the superposition time set by the second parameter changing unit, and the 2 of the measurement points. one of the measurement position p ₀ in the measurement time of one of the measurement time, the third with said motion parallax τ set by the parameter changing unit of determining the position p _c in the superimposed time of the measurement point Is,
The polar converter obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained polar line in the polar drawing space Voting the response intensity corresponding to the motion parallax τ of the measurement point corresponding to the polar line for each point on the trajectory of the polar line,
Instead of obtaining the intersection by the detection unit, the first, second, and third parameters are calculated in the cross ratio conversion unit and the pole conversion unit for a plurality of measurement points in the measurement space. In the changing unit, each value of each parameter is changed a plurality of times, and a maximum point at which the value obtained by voting becomes a maximum value is obtained for each polar drawing space, and the true value is calculated based on the maximum value information at the maximum point. The polar line drawing space corresponding to the moving direction is selected to obtain the true moving direction, and a plurality of polar lines participating in voting for the maximum points obtained for the polar line drawing space corresponding to the true moving direction It is also a preferred aspect to obtain the orientation of the measurement plane that includes a plurality of measurement points corresponding to and / or a physical quantity that indicates the superposition time at which the measurement plane overlaps the observation point.

The second image measuring device of the image measuring device of the present invention is
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , v is the moving direction between the two measurement times relative to the observation point, v, and the two measurement times are relative to the observation point. In the same direction as the moving direction v between and at the same speed as the moving speed between the two measurement times, the measurement point after the infinite time has elapsed in the scheduled movement continuation state P _inf , the position of the measurement point at the time of superimposition where the measurement plane including the measurement point overlaps the observation point in the movement continuation state is p _c , and the orientation of the measurement plane is n _s ,
Four positions _p inf of the measurement _points, and p _0, p 1, _p cross ratios determined by _c and the equivalent operations _{{p inf p 0 p 1 p} c} or plurality ratio, orientation _{n s} of the measurement plane Using the inner product ( _ns / v) with the moving direction v, the measurement plane at the one measurement time of the two measurement times from the observation point azimuth n _s and / or the observation point And an arithmetic unit that obtains a physical quantity that indicates the shortest distance up to.

Here, in the above-described second image measuring apparatus, the executed in the arithmetic unit, the compound ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations, the measurement point, the two measurement instead of the position p _0, p ₁ at two measurement time, the measurement point, the measurement position p ₀ at one measurement time of said two measurement time, the measurement point, the two Computation using motion parallax τ, which is the difference in position between the two measurement positions p ₀ and p ₁ at the measurement time, is included.

In the second image measurement device, the calculation unit uses the shortest distance as an index to measure the shortest distance between the observation point and the measurement plane at one of the two measurement times. The distance is d _s , the time between the one measurement time of the two measurement times and the weight time is t _c , and the measurement point between the two measurement times is relative to the observation point When the typical movement distance is Δx and the time between the two measurement times is Δt,
_n d _s = d _s / Δx
The standardized shortest distance _n d _s expressed by the following formula is adopted, and the standardized shortest distance _n d _{s
n t c} = t _c / _Δt
Using the normalized time _n t _c expressed by: and the inner product (n _s · v),
_{n d s = n t c (} n s · v)
It may be obtained using the following relational expression.

In the second image measurement device, the calculation unit includes:
A first parameter changing unit that changes the value of the first parameter using the physical quantity indicating the shortest distance as a first parameter;
A second parameter changing unit that changes the value of the second parameter using the inner product ( _ns / v) as a second parameter;
The physical quantity indicating the shortest distance set by the first parameter changing unit, the inner product ( _ns / v) set by the second parameter changing unit, and the two measurement times of the measurement points each measurement position p _0, p _1, or in, the two measuring positions p _0, replaces the p _1, of the measurement point, the two measurement positions in one of the measurement time p ₀ and the measurement point of the measurement time Of the motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times and the position p _inf after the infinite time has elapsed in the movement continuation state, the cross ratio {p _inf p ₀ with p 1 _{p c}} or plurality ratio equivalent calculation, the cross-ratio conversion unit for determining the position p _c in the superimposed time of the measurement points,
A polar transformation unit for obtaining the polar corresponding to the position p _c by converting the measurement point, the position p _c in the superimposed time pole,
Said a point of pole line, and the moving direction v and angle _{^{r = cos -1 (n s ·}} v)
And a point calculation unit for obtaining a point forming
With respect to a plurality of measurement points in the measurement space, each calculation in the cross ratio conversion unit, the pole conversion unit, and the point calculation unit is performed by using each value of the first parameter and the second parameter, respectively. The value of the first parameter for one measurement point in the point calculation unit is the same by being repeated a plurality of times while being changed by the one parameter changing unit and the second parameter changing unit, and Performing a calculation after a curve connecting a plurality of points obtained by a plurality of calculations with different values of the second parameter is obtained for the plurality of measurement points for each value of the first parameter;
By calculating intersections between curves when drawing a plurality of curves obtained while repeating each calculation in the cross ratio conversion unit, the polar conversion unit, and the point calculation unit a plurality of times in a curve drawing space, shortest orientation n _s measurement plane including a plurality of measurement points corresponding to a plurality of curves which intersect in the _{intersection,} and / or from the observation point to the measurement plane at one measurement time of said two measurement time It may be provided with a detection unit for obtaining a physical quantity indicating the distance,
In this case, the measurement point appearing on the image has intensity information, and the point calculation unit obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to a point, and instead of obtaining the intersection, each of the detection unit in the cross ratio conversion unit, the pole conversion unit, and the point calculation unit By obtaining a local maximum point at which the value obtained by voting during the calculation is repeated a plurality of times, the azimuth n _{s of the} measurement plane including a plurality of measurement points corresponding to the plurality of curves participating in the voting of the local maximum point, And / or preferably to obtain a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times, or
Measurement points appearing on the image have intensity information,
A third parameter change in which the computing unit changes the third parameter using the motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter. Part
The cross ratio conversion unit sets the physical quantity indicating the shortest distance set by the first parameter changing unit, the inner product ( _ns / v) set by the second parameter changing unit, and the measurement The measurement position p _{0 of the} point at one of the two measurement times, the motion parallax τ set by the third parameter, and the infinite time lapse of the measurement point in the movement continuation state using the position p _inf after, of the measurement points, which determine the position p _c in the superimposed time,
The calculation unit obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and determines the measurement point corresponding to the point on the polar line, Voting a response intensity corresponding to the motion parallax τ to a point in the curve drawing space corresponding to the point on the polar line,
Instead of obtaining the intersection, the detection unit performs each calculation in the cross ratio conversion unit, the polar conversion unit, and the calculation unit for the plurality of measurement points in the measurement space. In addition, the third parameter changing unit corresponds to a plurality of curves that participated in the voting of the local maximum point by calculating the local maximum point at which the value obtained by the voting during the multiple repetitions while changing each value of each parameter. A physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times, and / or the direction n _{s of the} measurement plane including a plurality of measurement points. It is also a preferred embodiment.

In the second image measurement device, the calculation unit includes:
A first parameter changing unit that changes the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state by changing the value of the first parameter using the movement direction v as a first parameter;
A second parameter changing unit that changes the value of the second parameter using the physical quantity indicating the shortest distance as a second parameter;
A third parameter changing unit that changes the value of the third parameter using the inner product ( _ns / v) as a third parameter;
A physical quantity indicative of the shortest distance set by the second parameter calculation unit and the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set by the first parameter calculation unit And the inner product ( _ns · v) set by the third parameter calculation unit and the measurement positions p ₀ , p _{1 at the} two measurement times of the measurement points, or these two measurement positions p ₀ , Instead of p ₁ , the measurement point p ₀ of the measurement point at one of the two measurement times and the two measurement positions p ₀ and p _{1 of the} measurement point at the two measurement times and a motion parallax τ between, using the double ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations, cross-ratio conversion unit for determining the position p _c in the superimposed time of the measurement point When,
A polar transformation unit for obtaining the polar corresponding to the position p _c by converting the measurement point, the position p _c in the superimposed time pole,
Said a point of pole line, and the moving direction v and angle _{^{r = cos -1 (n s ·}} v)
And a point calculation unit for obtaining a point forming
With respect to a plurality of measurement points in the measurement space, each calculation in the cross-ratio conversion unit, the pole conversion unit, and the point calculation unit is performed using each of the first to third parameter values from the first to third measurement points. By being repeated a plurality of times while being changed by each parameter changing unit up to the third, the value of the first parameter for the one measurement point in the point calculation unit is the same and the value of the second parameter Are the same, and a curve connecting a plurality of points obtained by a plurality of computations with different values of the third parameter is a value of each of the first parameter and each value of the second parameter. The calculation is performed after the plurality of measurement points are obtained for each combination of
Among a plurality of curve drawing spaces according to the value of the first parameter, a plurality of curves obtained while each calculation in the cross ratio conversion unit, the polar conversion unit, and the point calculation unit is repeated a plurality of times The intersection of curves when drawn in the corresponding curve drawing space is obtained for each curve drawing space, and the true movement of the measurement point relative to the observation point based on the information on the number of curves intersected at the intersection The true movement direction is obtained by selecting a curve drawing space corresponding to the direction, and a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection obtained for the curve drawing space corresponding to the true movement direction are obtained. orientation n _s measurement plane _containing, and / or a physical quantity that indicates the shortest distance to the measurement plane at one measurement time of said two measurement time from the observation point May be a structure having a Mel detection unit,
In that case, the measurement point appearing on the image has intensity information,
The point calculation unit obtains the point and votes a value corresponding to the intensity of the measurement point corresponding to the point to a point in the curve drawing space corresponding to the point where a curve including the point is drawn In place of obtaining the intersection, the detection unit becomes a maximum value by voting while each calculation in the cross ratio conversion unit, the polar conversion unit, and the point calculation unit is repeated a plurality of times. A maximum point is obtained for each curve drawing space, and the true movement direction is obtained by selecting a curve drawing space corresponding to the true movement direction based on information on the maximum value at the maximum point, and the true movement The direction n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the voting of the maximum points obtained for the curve drawing space corresponding to the direction, and / or the two measurement times from the observation points Out of Is preferably seeks a physical quantity that indicates the shortest distance to the measuring plane in a square of the measurement time, or,
Measurement points appearing on the image have intensity information,
The arithmetic unit changes a value of the fourth parameter using a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a fourth parameter. With a parameter changer,
The multi-ratio conversion unit is set by the first parameter changing unit, and the measurement point position p _inf after the infinite time has elapsed in the movement continuation state, and the second parameter changing unit is set, The measurement position p at one measurement time of the two measurement times of the physical quantity indicating the shortest distance, the inner product ( _ns / v) set by the third parameter changing unit, and the measurement point _0, by using the a motion parallax τ set by the fourth parameter changing unit, of the measurement points, which determine the position p _c in the superimposed time,
The point calculation unit obtains the point on the polar line corresponding to the measurement point, obtains a response intensity of the measurement point corresponding to the motion parallax τ, and determines the measurement point corresponding to the point on the polar line. Voting a response intensity corresponding to the motion parallax τ to a point in the curve drawing space corresponding to the point on the polar line,
Instead of obtaining the intersection, the detection unit performs each calculation in the cross ratio conversion unit, the polar conversion unit, and the calculation unit for the plurality of measurement points in the measurement space. In each of the third and fourth parameter changing units, a maximum point at which the value obtained by voting during a plurality of repetitions is changed while changing each value of each parameter is determined for each curve drawing space, and the maximum value at the maximum point is determined. The true movement direction is obtained by selecting the curve drawing space corresponding to the true movement direction based on the information, and the maximal point obtained for the curve drawing space corresponding to the true movement direction is participated in the vote. orientation n _s measurement plane including a plurality of measurement points corresponding to a plurality of _curves, and / or until the measurement plane at one measurement time of said two measurement time from the observation point It is a step of obtaining a physical quantity that indicates the shortest distance is also a preferred embodiment.

The third image measuring device of the image measuring device of the present invention is:
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in the image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , v is the moving direction between the two measurement times relative to the observation point, v, and the two measurement times are relative to the observation point. In the same direction as the moving direction v between and at the same speed as the moving speed between the two measurement times, the measurement point after the infinite time has elapsed in the scheduled movement continuation state _Where p _inf is the position of
Using a single ratio (p _inf p ₀ p ₁ ) determined by the three positions p _inf , p ₀ , and p ₁ of the measurement point (p _inf p ₀ p ₁ ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the An arithmetic unit is provided that calculates a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times.

Here, in the calculation performed by the calculation unit of the third image measurement apparatus, the single ratio (p _inf p ₀ p ₁ ) or calculation equivalent to the single ratio is performed by using the two measurement points. Instead of the two measurement positions p ₀ and p ₁ at the measurement time, the measurement position p ₀ of the measurement point at one of the two measurement times and the two measurement times of the measurement point The calculation using the motion parallax τ that is the difference in position between the two measurement positions p ₀ and p ₁ is included.

In the third image measuring apparatus, the arithmetic unit employs each position projected on the spherical surface as each position p _inf , p ₀ , and p ₁ of the measurement point, and a physical quantity that indicates the shortest distance And d _{s is} the shortest distance between the observation point and the measurement plane at one of the two measurement times, and the measurement point between the two measurement times is relative to the observation point. When the typical movement distance is Δx,
_n d _s = d _s / Δx
In adopting the normalized shortest distance _n d _s represented,
A parameter changing unit that changes the parameter using the normalized shortest distance _n d _s as a parameter;
Using the normalized shortest distance _n d _s set in the parameter changing unit and a simple ratio (p _inf p ₀ p ₁ ) or an operation equivalent to the simple ratio,
^{_{R = cos -1 (n d s}} / (p inf p 0 p 1))
A parameter calculation unit for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A small circle calculation unit for obtaining a small circle with a radius R centered on the measurement position of the measurement point at one of the two measurement times;
A plurality of small circles obtained while each calculation in the parameter calculation unit and the small circle calculation unit is repeated a plurality of times while the parameter is changed by the parameter change unit for a plurality of measurement points in the measurement space By determining the intersections between each other, the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to the small circles when the plurality of small circles intersecting at the intersections are drawn in the small circle drawing space, and / or the measurement And a detector for _{obtaining a} normalized shortest distance _n d _s0 about the plane,
In this case, the measurement point appearing on the image has intensity information, and the small circle calculation unit obtains the small circle and draws the obtained small circle in the small circle drawing space. A value corresponding to the intensity of the measurement point corresponding to the small circle is voted for each point on the trajectory of the small circle, and the parameter calculation is performed instead of obtaining the intersection. A plurality of measurements corresponding to a plurality of small circles that participated in the voting of the local maximum point by obtaining a local maximum point at which the value by voting becomes a local maximum value while each calculation in the unit and the small circle arithmetic unit is repeated a plurality of times orientation n _so the measurement plane including the _point, and / or, it is preferable to determine the normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
The arithmetic unit changes the value of the second parameter using the motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a second parameter. With a parameter changer,
The parameter calculation unit includes the standardized shortest distance _n d _s set by the parameter changing unit, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state, and the two measurements of the measurement point The radius R is obtained using the measurement position p ₀ at one measurement time of the time and the motion parallax τ set by the second parameter changing unit,
The small circle conversion unit obtains the small circle corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained small circle in the small circle drawing space. Voting the response intensity corresponding to the motion parallax τ of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when
Instead of the detection unit obtaining the intersection, the parameter calculation unit and the second parameter change are performed for each calculation in the parameter calculation unit and the small circle conversion unit for a plurality of measurement points in the measurement space. A plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the local maximum point by obtaining a local maximum point at which the value by voting becomes a local maximum value while changing each value of each parameter in the section It is also a preferable aspect to obtain the azimuth n _{so of the} measurement plane including, and / or the normalized shortest distance _n d _s0 for the measurement plane.

In the third image measurement apparatus, the calculation unit employs each position projected on the spherical surface as each position p _inf , p ₀ , and p ₁ of the measurement point, and uses the shortest distance as an index. As the physical quantity to be calculated, the shortest distance between the observation point and the measurement plane at one of the two measurement times is d _s , and the observation point of the measurement point between the two measurement times When the relative movement distance with respect to is Δx,
_n d _s = d _s / Δx
In adopting the normalized shortest distance _n d _s represented,
A first parameter changing unit that changes the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state by changing the value of the first parameter using the movement direction v as a first parameter;
A second parameter changing unit that changes the value of the second parameter using the normalized shortest distance _n d _s as a second parameter;
The measurement point position p _inf set after the infinite time in the movement continuation state set by the first parameter change unit, and the normalized shortest distance _n d _s set by the second parameter change unit , Using a simple ratio (p _inf p ₀ p ₁ ) or an operation equivalent to the simple ratio,
^{_{R = cos -1 (n d s}} / (p inf p 0 p 1))
A parameter calculation unit for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A small circle calculation unit for obtaining a small circle with a radius R centered on the measurement position of the measurement point at one of the two measurement times;
For each of the plurality of measurement points in the measurement space, each calculation in the parameter calculation unit and the small calculation unit is performed by changing each value of the first parameter and the second parameter to the first parameter change unit and the A plurality of small circles obtained while being repeated a plurality of times while being changed by the second parameter changing unit are each corresponding to a small circle drawing space among a plurality of small circle drawing spaces corresponding to the value of the first parameter. Corresponding to the true movement direction of the measurement point relative to the observation point based on information on the number of small circles intersecting at the intersection. The true movement direction is obtained by selecting a small circle drawing space, and a plurality of small circles intersecting at intersections obtained for the small circle drawing space corresponding to the true movement direction are determined. A plurality of orientations n _so the measurement plane including a measurement _point, and / or may be one that includes a detector for determining the normalized shortest distance _n d _s0 for the measurement plane,
In this case, the measurement point appearing on the image has intensity information, and the small circle calculation unit obtains the small circle, and the obtained small circle corresponds to the small circle. A value corresponding to the intensity of the measurement point corresponding to the small circle is voted for each point on the path of the small circle when drawn in a circle drawing space, and the detection unit obtains the intersection Instead, a maximum point at which a value obtained by voting becomes a maximum value while each calculation in the parameter calculation unit and the small circle calculation unit is repeated is obtained for each small circle drawing space, and is based on information on the maximum value at the maximum point. The true movement direction was determined by selecting the small circle drawing space corresponding to the true movement direction, and the maximal point obtained for the small circle drawing space corresponding to the true movement direction was voted Supports multiple small circles That a plurality of azimuth n _so the measurement plane including a measurement _point, and / or, preferably and requests the normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
The arithmetic unit changes a value of the third parameter using a motion parallax τ between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter. With a parameter changer,
The parameter change unit is set by the first parameter change unit, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state, and the normalization set by the second parameter change unit Using the shortest distance _n d _s , the measurement position p _{0 at} one of the two measurement times of the measurement point, and the motion parallax τ set by the third parameter changing unit, Determining the radius R;
The small circle calculation unit obtains the small circle corresponding to the measurement point, obtains a response intensity of the measurement point corresponding to the motion parallax τ, and determines the obtained small circle as a small circle corresponding to the small circle. Voting the response intensity corresponding to the motion parallax τ of each measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in a circle drawing space;
Instead of the detection unit obtaining the intersection, the first, second, and third parameters are calculated in the parameter calculation unit and the small circle calculation unit for a plurality of measurement points in the measurement space. The maximum value at which the value obtained by voting during a plurality of repetitions is changed while changing each value of each parameter in the changing unit is determined for each small circle drawing space, and the true value is calculated based on the information on the maximum value at the maximum point. A plurality of small circles that participated in voting for the maximum points obtained for the small circle drawing space corresponding to the true moving direction while obtaining the true moving direction by selecting the small circle drawing space corresponding to the moving direction It is also a preferred embodiment to obtain the azimuth n _{so of the} measurement plane including a plurality of measurement points corresponding to and / or the normalized shortest distance _n d _s0 for the measurement plane.

Furthermore, the 4th image measuring device of the image measuring devices of this invention is:
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , relative to the observation point, in the same direction as the movement direction v between the two measurement times, and the same as the movement speed between the two measurement times When the position of the measurement point after elapse of an infinite time in a state where the movement at the speed is continued as it is and when the movement is continuously planned is p _inf ,
Using the single ratio (p _inf p ₀ p ₁ ) determined by the three positions p _inf , p ₀ , and p ₁ of the measurement point (p _inf p ₀ p ₁ ) or an operation equivalent to the single ratio, the observation point and the two measurement times An arithmetic unit that obtains a physical quantity indicating the distance from the measurement point at one of the measurement times is provided.

Here, for the calculation performed by the calculation unit of the fourth image measurement device, the single ratio (p _inf p ₀ p ₁ ) or equivalent to the single ratio, the two measurement points may be calculated using the two measurement points. Instead of the two measurement positions p ₀ and p ₁ at the measurement time, the measurement position p ₀ of the measurement point at one of the two measurement times and the two measurement times of the measurement point The calculation using the motion parallax τ which is the difference in position between the two measurement positions p ₀ and p ₁ in FIG.

In the fourth image measurement device, the arithmetic unit calculates a distance between the observation point and the measurement point at one of the two measurement times as a physical quantity indicating the distance, d _0. When the movement distance of the measurement point between the two measurement times with respect to the observation point is Δx,
_n d ₀ = d ₀ / Δx
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _inf p ₀ p ₁ )
Or a relational expression equivalent to the relational expression.

The fifth image measuring device of the image measuring device of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A measurement plane including the measurement point in a movement continuation state in which the movement in the same direction as the general movement direction and the movement at the same speed as the movement speed between the two measurement times is continued as it is. a physical quantity indicative of the superimposed time overlapping the observation point, a parameter setting unit that sets as a parameter the coordinates of voting space defined by the azimuth n _s of the measurement plane,
The measurement position p ₀ of the measurement point at one of the two measurement times, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state, and the parameter setting unit A motion parallax calculation unit that obtains a motion parallax τ that is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the coordinates in the voting space,
A response intensity calculation unit for obtaining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the parameter setting unit in the voting space;
For each of a plurality of measurement points in the measurement space, each process in the motion parallax calculation unit, the response intensity calculation unit, and the voting unit is executed a plurality of times while changing a parameter value in the parameter setting unit. Features.

The sixth image measuring device of the image measuring device of the present invention is
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting a specific moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first parameter setting unit that sets the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
In the mobile continuity state, and the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, defined by the azimuth n _s of the measurement plane, voting space corresponding to the first parameter A second parameter setting unit for setting the coordinates in the second parameter,
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set by the first parameter setting unit, and the second parameter setting unit are set. A motion parallax calculation unit for obtaining a motion parallax τ which is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the coordinates in the voting space;
A response intensity calculation unit for obtaining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fifth voting unit for voting the response strength obtained by the response strength calculating unit to the coordinates set by the second parameter setting unit in a voting space corresponding to the first parameter;
For each of the plurality of measurement points in the measurement space, each process in the motion parallax calculation unit, the response intensity calculation unit, and the voting unit is changed, and each value of each parameter is changed in the first and second parameter setting units. It is characterized by being executed a plurality of times.

The seventh image measuring device of the image measuring device of the present invention is
A predetermined observation point in the measurement space that observes the measurement space, and an arbitrary measurement point in the measurement space that appears in an image when the measurement space is viewed from the observation point. to set a physical quantity that indicates the shortest distance between the measurement plane at one measurement time of the two different measurement times, the coordinates of the voting space defined by the azimuth n _s of the measuring plane as a parameter A parameter setting section;
In the same direction as the relative movement direction of the measurement point relative to the observation point between the measurement point p ₀ at one of the two measurement times and the measurement point between the two measurement times. And the position p _inf after the infinite time has elapsed in the planned movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is, and the parameter setting unit sets A motion parallax calculation unit that obtains a motion parallax τ that is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the coordinates in the voting space;
A response intensity calculation unit for obtaining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the parameter setting unit in the voting space;
Each process in the motion parallax calculation unit, the response intensity calculation unit, and the voting unit in the measurement space is executed a plurality of times while changing parameter values in the parameter setting unit.

Further, the eighth image measuring device of the image measuring device of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting a specific moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first parameter setting unit that sets the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
Wherein from the observation point, comprising said measuring points, and the physical quantity that indicates the shortest distance to the measurement plane at the measurement time of one of the two measurement time is defined by the azimuth n _s of the measurement plane, the A second parameter setting unit that sets the coordinates in the voting space according to the first parameter as a second parameter;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set by the first parameter setting unit, and the second parameter setting unit are set. A motion parallax calculation unit for obtaining a motion parallax τ which is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the coordinates in the voting space;
A response intensity calculation unit for obtaining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the second parameter setting unit in a voting space corresponding to the first parameter;
For each of the plurality of measurement points in the measurement space, each process in the motion parallax calculation unit, the response intensity calculation unit, and the voting unit is changed, and each value of each parameter is changed in the first and second parameter setting units. It is characterized by being executed a plurality of times.

The ninth image measuring device of the image measuring device of the present invention is
Two measurement positions p ₀ , at two measurement times different from each other at arbitrary measurement points in the measurement space that appear in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. a parameter setting unit that sets a motion parallax τ that is a difference in position between p ₁ as a parameter;
In the same direction as the relative movement direction of the measurement point relative to the observation point between the measurement point p ₀ at one of the two measurement times and the measurement point between the two measurement times. In addition, the position p _inf after the infinite time has elapsed in the movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is and the parameter setting The movement parallax τ set by the unit is defined by a physical quantity that indicates a superposition time at which the measurement plane including the measurement point overlaps the observation point in the movement continuation state, and the direction n _{s of the} measurement plane A coordinate calculation unit for obtaining coordinates in the voting space;
Based on two images when the measurement space is viewed from the observation point at the two measurement times, a response intensity corresponding to the motion parallax τ set in the first step is obtained at the measurement point. A response strength calculator,
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates obtained by the coordinate computing unit in the voting space;
For the plurality of measurement points in the measurement space, the process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is executed a plurality of times while changing the parameter value in the parameter setting unit. To do.

The tenth image measuring device of the image measuring device of the present invention is
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting a specific moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first parameter setting unit that sets the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
A second parameter setting unit that sets, as a second parameter, a motion parallax τ that is a difference in position between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set by the first parameter setting unit, and the second parameter setting unit are set. and a motion parallax τ was, in the mobile continuity state, and the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, defined by the azimuth n _s of the measurement plane, said first A coordinate calculation unit for obtaining coordinates in the voting space according to the parameters of
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the second step is obtained at the measurement point. A response strength calculator,
A voting unit for voting the response strength obtained by the response strength computing unit in the voting space according to the first parameter to the coordinates obtained by the coordinate computing unit;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is changed, and each value of each parameter is changed in the first and second parameter setting units. It is characterized by being executed multiple times.

The eleventh image measuring device of the image measuring device of the present invention is
Two measurement positions p ₀ , at two different measurement times of arbitrary measurement points in the measurement space that appear in the image when the inside of the measurement space is viewed from a predetermined observation point in the predetermined measurement space. a parameter setting unit that sets a motion parallax τ that is a difference in position between p ₁ as a parameter;
The same direction as the relative movement direction of the measurement point with respect to the observation point between the two measurement times and the measurement position p ₀ at one of the two measurement times. And the position p _inf after the infinite time of the measurement point in a state where the movement at the same speed as the movement speed between the two measurement times is continued as it is, and the parameter Based on the motion parallax set by the setting unit, a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times including the measurement point; , a coordinate calculation unit for obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
Based on two images when the measurement space is viewed from the observation point at the two measurement times, a response intensity corresponding to the motion parallax τ set in the first step is obtained at the measurement point. A response strength calculator,
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates obtained by the coordinate computing unit in the voting space;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit. And

Furthermore, the twelfth image measuring device of the image measuring device of the present invention is
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first parameter setting unit for setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
A second parameter setting unit that sets, as a second parameter, a motion parallax τ that is a difference between positions of the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set by the first parameter setting unit, and the second parameter setting unit are set. A physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point, based on the motion parallax τ, and the measurement plane of it is defined by the azimuth n _s, and a coordinate calculation unit for obtaining the coordinates of the voting space corresponding to the first parameter,
Based on two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set by the second parameter setting unit at the measurement point A response intensity calculation unit for obtaining
A voting unit for voting the response strength obtained by the response strength computing unit in the voting space according to the first parameter to the coordinates obtained by the coordinate computing unit;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is changed, and each value of each parameter is changed in the first and second parameter setting units. It is characterized by being executed multiple times.

The thirteenth image measuring device of the image measuring device of the present invention is
Based on two images when the measurement space is viewed at two different measurement times from a predetermined observation point in the predetermined measurement space, any two measurement points in the measurement space A response intensity calculation unit for obtaining a response intensity corresponding to motion parallax, which is a difference between positions of two measurement positions at one measurement time;
The response intensity obtained by the response intensity calculator is set in the same direction as the movement direction of the measurement point relative to the observation point between the two measurement times and between the two measurement times. A physical quantity that indicates a superposition time at which a measurement plane including the measurement point overlaps the observation point in a planned movement continuation state in which the movement at the same speed as the movement speed is continued, and an orientation of the measurement plane A voting unit for voting on coordinates corresponding to the measurement points and the motion parallax within a prescribed voting space;
Each process in the response intensity calculation unit and the voting unit is executed a plurality of times for a plurality of measurement points in the measurement space.

  Here, the thirteenth image measuring apparatus includes a measurement plane including a plurality of measurement points participating in the voting of the local maximum by obtaining a local maximum in the voting space where the value by the voting becomes a local maximum. And / or a detection unit that obtains a physical quantity that indicates a superimposition time at which the measurement plane overlaps the observation point.

The fourteenth image measuring device of the image measuring device of the present invention is
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A parameter setting unit for setting a general moving direction as a parameter;
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the movement that is the difference between the two measurement positions at the two measurement times of the measurement point A response intensity calculation unit for obtaining a response intensity corresponding to the parallax;
The response intensity obtained by the response intensity calculator is set in the same direction as the movement direction of the measurement point relative to the observation point between the two measurement times and between the two measurement times. A physical quantity that indicates a superposition time at which a measurement plane including the measurement point overlaps the observation point in a planned movement continuation state in which the movement at the same speed as the movement speed is continued, and an orientation of the measurement plane A voting unit for voting on coordinates corresponding to the measurement points and the motion parallax in a voting space defined according to the parameters set in the first step;
For each of a plurality of measurement points in the measurement space, each process in the response intensity calculation unit and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit.

  Here, the fourteenth image measuring device obtains a local maximum point at which the value obtained by voting becomes a local maximum value for each voting space, and makes the relative measurement point relative to the observation point based on information on the local maximum value at the local maximum point. A plurality of measurement points participating in the voting of the maximum points obtained for the voting space corresponding to the true movement direction while obtaining the true movement direction by selecting a voting space corresponding to the true movement direction. May be provided with a detection unit that obtains a physical quantity indicating the azimuth of a measurement plane that includes and / or a superposition time at which the measurement plane overlaps the observation point.

The fifteenth image measuring device of the image measuring device of the present invention is
Based on two images when the measurement space is viewed at two different measurement times from a predetermined observation point in the predetermined measurement space, the two measurements at arbitrary measurement points in the measurement space A response intensity calculation unit for obtaining a response intensity corresponding to a motion parallax that is a difference between positions of two measurement positions at time;
The response intensity obtained by the response intensity calculator is a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point. A voting unit for voting on coordinates corresponding to the measurement points and the motion parallax in a voting space defined by the orientation of the measurement plane,
For each of a plurality of measurement points in the measurement space, each process in the response intensity calculation unit and the voting unit is executed a plurality of times.

  Here, the fifteenth image measuring apparatus includes a measurement plane including a plurality of measurement points participating in the voting of the local maximum point by obtaining a local maximum point in the voting space where the value by the voting becomes a local maximum value. And / or a detector that obtains a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times.

Furthermore, the sixteenth image measurement device of the image measurement devices of the present invention is
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A parameter setting unit for setting a general moving direction as a parameter;
Motion parallax, which is the difference between the two measurement positions at the two measurement times of the measurement point, based on the two images when the measurement space is viewed from the observation point at the two measurement times. A response intensity calculation unit for obtaining a response intensity corresponding to
The response intensity obtained by the response intensity calculator is a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point. A voting unit for voting on coordinates corresponding to the measurement points and the motion parallax in a voting space defined by the orientation of the measurement plane and according to the parameter set in the first step. ,
For each of a plurality of measurement points in the measurement space, each process in the response intensity calculation unit and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit.

  Here, the sixteenth image measuring device obtains a local maximum point at which the value obtained by voting becomes a local maximum value for each voting space, and makes the measurement point relative to the observation point based on information on the local maximum value at the local maximum point. A plurality of measurements that participated in the voting of the maximum points obtained for the polar drawing space corresponding to the true movement direction while obtaining the true movement direction by selecting a voting space corresponding to the true movement direction A detector that obtains a physical quantity indicating the azimuth of the measurement plane including the point and / or the shortest distance from the observation point to the measurement plane at one of the two measurement times. May be.

The seventeenth image measurement device of the image measurement devices of the present invention is
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , respectively, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the visual axis direction v connecting the two observation points is represented by p _axis . wherein one of the observation points of the point, when on the viewing plane extending parallel to the measurement plane including the measurement point, the position of intersection between the straight line and the p _c,
The four positions _{p axis,} using p _R, p L, the cross ratio _{{p axis p R p L p} c} or plurality ratio equivalent operations determined by _{p c,} the orientation of the measurement plane, and / or And an arithmetic unit for obtaining a physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points.

Here, in the seventeenth image measuring device, the _cross ratio {p _axis p _R p _L p _c } or a calculation equivalent to the cross ratio is executed in the calculation unit. Instead of the two measurement positions p _R and p _L when observed from two observation points, the measurement position p _R when the measurement point is observed from one of the two observation points; Computation using the binocular parallax σ which is the difference in position between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is included.

In the seventeenth image measurement apparatus, the calculation unit includes, as a physical quantity that indicates a distance in the visual axis direction between the measurement plane and the two observation points, of the measurement plane and the two observation points. When the distance in the visual axis direction between the one observation point is d _c and the distance between the two observation points is Δx _LR ,
_{n d c = d c / Δx} LR
In adopting the normalized distance _n d _c represented, the normalized distance _n d _c,
n d _c = {p _axis p _R p _L p _c }
Or a relational expression equivalent to the relational expression.

In the seventeenth image measurement apparatus, the calculation unit includes:
A parameter changing unit that changes a value of the parameter using a physical quantity indicating a distance in the visual axis direction between the measurement plane and one of the two observation points as a parameter;
The two observations of the physical quantity that is set by the parameter changing unit and indicates the distance in the visual axis direction between the measurement plane and one of the two observation points, and the measurement point When observing from one of the two observation points instead of the two measurement positions p _R and p _L or the two measurement positions p _R and p _L when observed from the point the measurement position p _R and the measurement point, two measurement positions p _R when observed from the two observation _points, and binocular parallax σ between and if p _L, and a position p _axis of the point at infinity , by using the double ratio _{{p axis p R p L p} c} or plurality ratio equivalent calculation, the cross-ratio conversion unit for determining the position _{p c} of the intersection on the observation plane,
A polar transformation unit for obtaining the polar corresponding to the measurement point by polar transformation the position p _c of the intersection on the observation plane,
For a plurality of measurement points in the measurement space, each calculation in the multi-ratio conversion unit and the pole conversion unit is performed while being repeated a plurality of times while the parameter value is changed by the parameter change unit. By obtaining the intersection of the polar lines when the polar lines are drawn in the polar drawing space, the orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines intersecting at the intersection and / or the measurement It may be provided with a detection unit for obtaining a physical quantity that indicates a distance in the visual axis direction between a plane and one of the two observation points,
In that case, the measurement point appearing on the image has intensity information, and the polar converter obtains the polar line and draws the obtained polar line in the polar drawing space. A value corresponding to the intensity of the measurement point corresponding to the polar line is voted for each point on the polar line trajectory, and the fourth step replaces the intersection with the compound point. By calculating each of the operations in the ratio conversion unit and the maximum conversion unit a plurality of times while changing the value of the parameter by the parameter conversion unit, a maximum point where the value by voting becomes a maximum value is obtained. The orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines participating in the voting and / or the visual axis direction between the measurement plane and one of the two observation points The physical quantity to index the distance of Is preferably one, or
Measurement points appearing on the image have intensity information,
The calculation unit uses the binocular parallax σ between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a second parameter value. A second parameter changing unit for changing
The cross ratio conversion unit, which is set by the parameter change unit, is a physical quantity that indicates a distance in the visual axis direction between the measurement plane and one of the two observation points; and the measurement A measurement position p _R when observed from one of the two observation points, the binocular parallax σ set by the second parameter changing unit, and the position p of the infinity point and a _axis, is intended to determine the intersection _{p c} on the observation plane,
When the polar conversion unit obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained polar line in the polar drawing space Voting the response intensity corresponding to the binocular parallax σ of the measurement point corresponding to the polar line for each point on the trajectory of the polar line,
Instead of the detection unit obtaining the intersection, the calculation in the cross ratio conversion unit and the polar conversion unit is performed on the plurality of measurement points in the measurement space by the parameter change unit and the second parameter change. A plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the local maximum points by obtaining a local maximum point at which the value obtained by voting during a plurality of repetitions is changed while changing each value of each parameter in the section It is also preferable to obtain a physical quantity that indicates the orientation of the measurement plane including the and / or the distance in the visual axis direction between the measurement plane and one of the two observation points. It is.

In the seventeenth image measurement apparatus, the calculation unit includes:
A first parameter changing unit that changes the position p _{axis of the} infinity point by changing the value of the visual axis direction v as a first parameter;
A second parameter that changes the value of the second parameter using a physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points as a second parameter. Change part,
The position p _axis set by the first parameter changing unit, the physical quantity set by the second parameter changing unit for indicating the distance in the visual axis direction, and the two observation points of the measurement point, respectively each measurement position p _R, p _L or when observed from these two measurement positions p _R, replaces the p _L, of the measurement point, when observed from one of the observation point of said two observation point From the measurement position p _R and the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points, the cross ratio {p _axis p _R p _L using p _c} or plurality ratio equivalent calculation, the cross-ratio conversion unit for determining the position p _c of the intersection on the observation plane of the measurement point,
A polar transformation unit for obtaining the polar corresponding to the measurement point by converting the measurement point, the position p _c of intersection on said viewing plane electrode,
With respect to a plurality of measurement points in the measurement space, each calculation in the cross-ratio conversion unit and the pole conversion unit is performed by using the first parameter change unit and the first parameter change value and the second parameter value, respectively. Drawing a plurality of polar lines obtained while being changed a plurality of times while being changed by the parameter changing unit in a corresponding polar drawing space among a plurality of polar drawing spaces corresponding to the value of the first parameter The intersection of the polar lines is determined for each polar drawing space, and the true line drawing space is selected by selecting the polar drawing space corresponding to the true visual axis direction based on the information on the number of polar lines intersecting at the intersection. An orientation of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines intersecting at the intersection obtained for the polar drawing space corresponding to the true visual axis direction, and / or obtaining a visual axis direction, and / or , Preferably with the measuring plane and the observation point said at that a detecting section for obtaining a physical quantity indicative of the distance of the visual axis direction between one of the observation points of the two observation points,
In this case, the measurement point appearing on the image has intensity information, and the second pole conversion unit obtains the polar line, and the obtained polar line corresponds to the polar line. A value corresponding to the intensity of the measurement point corresponding to the polar line is voted for each point on the trajectory of the polar line when drawn in the line drawing space, and the detection unit obtains the intersection Instead, the value obtained by voting while each calculation in the cross ratio conversion unit and the maximum conversion unit is repeated a plurality of times while the values of the first parameter and the second parameter are changed becomes a maximum value. A maximum point is obtained for each polar drawing space, and the true visual axis direction is determined by selecting a polar drawing space corresponding to the true visual axis direction based on the local maximum information at the local maximum point, Corresponding to the true visual axis direction The orientation of a measurement plane including a plurality of measurement points corresponding to a plurality of polar lines that participated in the voting of local maximum points obtained for the line drawing space and / or one of the measurement plane and the two observation points It is preferable to obtain a physical quantity that indicates the distance in the visual axis direction between the observation point, or
Measurement points appearing on the image have intensity information,
The calculation unit uses the binocular parallax σ between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a third parameter value of the third parameter. A third parameter changing unit for changing
The cross ratio conversion unit is configured to observe one of the measurement plane and the two observation points set by the position p _axis set by the first parameter change unit and the second parameter change unit. A physical quantity indicating the distance in the visual axis direction between the point, the measurement position p _R of the measurement point when observed from one of the two observation points, and the third parameter the set change unit by using the binocular parallax sigma, is intended to determine the intersection p _c on the observation plane,
When the polar conversion unit obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained polar line in the polar drawing space Voting the response intensity corresponding to the binocular parallax σ of the measurement point corresponding to the polar line for each point on the trajectory of the polar line,
Instead of obtaining the intersection, the detection unit performs the first, second, and third calculations for the plurality of measurement points in the measurement space in the cross ratio conversion unit and the polar conversion unit. The parameter changing unit obtains a maximum point at which the value obtained by voting during a plurality of repetitions while changing each value of each parameter, for each polar drawing space, and calculates the true value based on the maximum value information at the maximum point. The true visual axis direction was obtained by selecting a polar drawing space corresponding to the visual axis direction of the user, and participated in voting for the maximum points obtained for the polar drawing space corresponding to the true visual axis direction An indication of the orientation of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines and / or the distance in the visual axis direction between the measurement plane and one of the two observation points Is to calculate the physical quantity to be And also a preferred embodiment.

The eighteenth image measuring device of the image measuring device of the present invention is
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the visual axis direction connecting the two observation points is v, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the visual axis direction v is p. _axis , the position of the intersection with the straight line on the observation plane including one observation point of the two observation points and extending in parallel with the measurement plane including the measurement point, p _c , and the orientation of the measurement plane where n _s
The four positions _{p axis, p R, p L} , cross ratios determined by _{p c {p axis p R p} L p c} or plurality ratio equivalent operations and the visual axis and the orientation _{n s} of the measurement plane by using the inner product of the direction v of the (n s _· v), the azimuth n _s measurement _plane, and / or the shortest between one observation point of the said measurement plane the two observation points An arithmetic unit for obtaining a physical quantity indicating the distance is provided.

Here, in the eighteenth image measuring apparatus, the calculation point is calculated by performing the calculation of the _cross ratio {p _axis p _R p _L p _c } or the calculation equivalent to the cross ratio by the 2 Instead of the two measurement positions p _R and p _L when observed from one observation point, the measurement position p _R when the measurement point is observed from one of the two observation points; Computation using the binocular parallax σ which is the difference in position between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is included.

The eighteenth image measuring apparatus, as a physical quantity as an index the shortest distance, the shortest distance between one of the observation point of the measuring plane and the two observation points d _s, and the measuring plane when the distance of the visual axis direction between one of the observation points of the two observation points were d _c, the distance between the each other two observation points and [Delta] x _LR,
n d _s = d _s / Δx _LR
The standardized shortest distance _n d _s expressed by the following formula is adopted, and the standardized shortest distance _n d _{s
n d c = d c / Δx} LR
Using the normalized distance _n d _c expressed by the above and the inner product (n _s · v),
_n d _s = _n d _c (n _s · v)
It may be obtained using the following relational expression.

In the eighteenth image measuring apparatus, a first parameter changing unit that changes a value of the first parameter using a physical quantity that indicates the shortest distance as a first parameter by the arithmetic unit;
A second parameter changing unit that changes the value of the second parameter using the inner product ( _ns / v) as a second parameter;
A physical quantity indicating the shortest distance set by the first parameter change unit, an inner product ( _ns / v) set by the second parameter change unit, and the two observation points of the measurement points, respectively each measurement position p _R, p _L or when observed from these two measurement positions p _R, replaces the p _L, of the measurement point, when observed from one of the observation point of said two observation point From the measurement position p _R and the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points and the position p _{axis of the} infinity point, using an equivalent operation with the cross ratio _{{p axis p R p L p} c} or plurality ratio, the cross-ratio conversion unit for determining the position _{p c} of the intersection on the observation plane,
A polar transformation unit for obtaining the polar corresponding to the position p _c by polar transformation the position p _c of the intersection on the observation plane,
The point on the polar line, and the direction v and angle of the visual axis r = cos ⁻¹ (n _s · v)
And a point calculation unit for obtaining a point forming
With respect to a plurality of measurement points in the measurement space, each calculation in the cross-ratio conversion unit, the pole conversion unit, and the point calculation unit is the first value of each of the first parameter and the second parameter. The parameter changing unit and the second parameter changing unit are repeated a plurality of times while being changed, whereby the value of the first parameter relating to one measurement point in the point calculation unit is the same, and the Performing a calculation after a curve connecting a plurality of points obtained by a plurality of calculations with different values of the second parameter is obtained for the plurality of measurement points for each value of the first parameter;
By calculating intersections between curves when drawing a plurality of curves obtained while repeating each calculation in the cross ratio conversion unit, the polar conversion unit, and the point calculation unit a plurality of times in a curve drawing space, An orientation n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection, and / or the shortest distance between the measurement plane and one of the two observation points It may be provided with a detection unit for obtaining a physical quantity to be indexed
In this case, the measurement point appearing on the image has intensity information, and the point calculation unit obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to a point, and instead of obtaining the intersection, each of the detection unit in the cross ratio conversion unit, the pole conversion unit, and the point calculation unit By obtaining a local maximum point at which the value obtained by voting during the calculation is repeated a plurality of times, the azimuth n _{s of the} measurement plane including a plurality of measurement points corresponding to the plurality of curves participating in the voting of the local maximum point And / or it is preferable to obtain a physical quantity indicating the shortest distance between the measurement plane and one of the two observation points, or
Measurement points appearing on the image have intensity information,
The calculation unit uses the binocular parallax σ between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a third parameter value of the third parameter. A third parameter changing unit for changing
The cross ratio conversion unit sets the physical quantity indicating the shortest distance set by the first parameter changing unit, the inner product ( _ns / v) set by the second parameter changing unit, and the measurement Measurement point p _R when observed from one of the two observation points, the binocular parallax σ set by the third parameter changing unit, and the position of the infinity point by using the p _axis, it is intended to determine the position _{p c} of the intersection on the observation plane,
The point calculation unit obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and measures the measurement point corresponding to the point on the polar line Voting a response intensity corresponding to the binocular parallax σ to a point in the curve drawing space corresponding to the point on the polar line,
Instead of obtaining the intersection, the detection unit performs each calculation in the cross ratio conversion unit, the polar conversion unit, and the point calculation unit for the plurality of measurement points in the measurement space. A plurality of curves that participated in the voting of the local maximum points by obtaining the local maximum point at which the value obtained by voting during a plurality of repetitions while changing each value of each parameter in the second and third parameter changing units the azimuth n _s measurement plane including a plurality of measurement points _{corresponding,} and / or, and requests a physical quantity that indicates the shortest distance between one of the observation points of the with the measuring plane two observation point It is also a preferred embodiment.

In the eighteenth image measurement apparatus, the calculation unit includes:
A first parameter changing unit that changes the position p _{axis of the} infinity point by changing the value of the first parameter using the direction v of the visual axis as a first parameter;
A second parameter changing unit that changes the second parameter using the physical quantity indicating the shortest distance as a second parameter;
A third parameter changing unit that changes the value of the third parameter using the inner product ( _ns / v) as a third parameter;
The position p _{axis of the} infinity point set by the second parameter changing unit, the physical quantity indicating the shortest distance set by the third parameter changing unit, and the pole converting unit are set. and the inner product (n s _· v), wherein each measurement position p _R when observed from each of the two observation points of measurement _points, p _L or replaces the two measuring positions p _R, the p _L, the measurement The measurement position p _R when the point is observed from one of the two observation points and the two measurement positions p _R and p _L when the point is observed from the two observation points and a binocular parallax σ between, using the double ratio _{{p axis p R p L p} c} or plurality ratio equivalent operations, cross-ratio conversion for determining the position p _c of intersection on said viewing plane And
A polar transformation unit for obtaining the polar corresponding to the position p _c by polar transformation the position p _c of the intersection on the observation plane,
Said a point of pole line, and the visual axis direction v and angle _{^{r = cos -1 (n s ·}} v)
And a point calculation unit for obtaining a point forming
With respect to a plurality of measurement points in the measurement space, each calculation in the cross-ratio conversion unit, the pole conversion unit, and the point calculation unit is performed using each of the first to third parameter values from the first to third measurement points. By being repeated a plurality of times while being changed by each parameter changing unit up to the third, the value of the first parameter for the one measurement point in the point calculation unit is the same and the value of the second parameter Are the same, and a curve connecting a plurality of points obtained by a plurality of computations with different values of the third parameter is a value of each of the first parameter and each value of the second parameter. The calculation is performed after the plurality of measurement points are obtained for each combination of
Among a plurality of curve drawing spaces according to the value of the first parameter, a plurality of curves obtained while each calculation in the cross ratio conversion unit, the polar conversion unit, and the point calculation unit is repeated a plurality of times The intersection of the curves when drawn in the corresponding curve drawing space is obtained for each curve drawing space, and the curve drawing space corresponding to the true visual axis direction is selected based on the information on the number of curves intersecting at the intersection. Thereby determining the true visual axis direction, and the orientation n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection obtained with respect to the curve drawing space corresponding to the true visual axis direction, And / or a detector for obtaining a physical quantity that indicates the shortest distance between the measurement plane and one of the two observation points.
In this case, the measurement point appearing on the image has intensity information, and the point calculation unit obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting on a point in a curve drawing space corresponding to a point in which a curve including the point is drawn, and instead of obtaining the intersection, the detection unit converts the cross ratio conversion unit, the polar conversion unit, And the true visual axis direction based on information on the local maximum value at each local area, for each curve drawing space, where a local maximum value is obtained by voting while each calculation in the point arithmetic unit is repeated a plurality of times. The true visual axis direction is obtained by selecting a curve drawing space corresponding to the two, and corresponding to a plurality of curves that participated in voting for the maximum points obtained for the curved drawing space corresponding to the true visual axis direction Multiple measurement points Orientation n _s free measurement _plane, and / or, preferably and requests a physical quantity that indicates the shortest distance between one of the observation points of the with the measuring plane two observation points, or,
Measurement points appearing on the image have intensity information,
The arithmetic unit changes the fourth parameter using the binocular parallax σ between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a fourth parameter. And a fourth parameter changing unit
The multi-ratio conversion unit is configured by the first parameter changing unit, the position p _{axis of the} infinity point, and the physical quantity indicating the shortest distance set by the second parameter changing unit; The inner product ( _ns · v) set by the third parameter changing unit, the measurement position p _R when the measurement point is observed from one of the two observation points, the first wherein set in the fourth parameter changing section by using the binocular parallax sigma, it is intended to determine the position p _c of the intersection on the observation plane,
The point calculation unit obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and measures the measurement point corresponding to the point on the polar line Voting a response intensity corresponding to the binocular parallax σ to a point in the curve drawing space corresponding to the point on the polar line,
Instead of the detection unit obtaining the intersection, for each of a plurality of measurement points in the measurement space, each calculation unit in the cross ratio conversion unit, the pole conversion unit, and the point calculation unit is set to the first, A maximum point at which the value obtained by voting during a plurality of repetitions is maximized while changing each value of each parameter in the second, third, and fourth parameter changing units is determined for each curve drawing space, and the maximum at the maximum point is determined. The true visual axis direction is obtained by selecting the curved drawing space corresponding to the true visual axis direction based on the value information, and the maximum obtained for the curved drawing space corresponding to the true visual axis direction orientation n _s measurement plane including a plurality of measurement points corresponding to a plurality of curves participated in the vote of _points, and / or the shortest between one observation point of the said measurement plane the two observation points Physical quantity indicating distance It is also a preferred embodiment is a step of obtaining.

In addition, the nineteenth image measuring device among the image measuring devices of the present invention is:
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the direction of the visual axis connecting the two observation points is v, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the two visual axes. When it is p _axis
Using a single ratio determined by the three positions p _axis , p _R , and p _L (p _axis p _R p _L ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the measurement plane An arithmetic unit for obtaining a physical quantity indicating the shortest distance between one of the two observation points is provided.

Here, for the calculation performed by the calculation unit of the nineteenth image measurement apparatus, the single ratio (p _axis p _R p _L ) or equivalent to the single ratio, the two measurement points may be calculated using the two measurement points. a measurement position p _R when observed from one of the observation points of the observation point, the measurement point, two measurement positions p _R when observed from the two observation _points, the position between and if p _L The calculation using the binocular parallax σ which is the difference is included.

In the nineteenth image measuring apparatus, the calculation unit employs each position projected on a spherical surface as each position p _axis , p _R , and p _{L, and} as a physical quantity indicating the shortest distance, When the shortest distance between the measurement plane and one of the two observation points is d _s , and the distance between the two observation points is Δx _LR ,
_n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A parameter changing unit that changes the parameter using the normalized shortest distance _n d _s as a parameter;
Using the standardized shortest distance _n d _s set in the parameter changing unit and a simple ratio (p _axis p _R p _L ) or an operation equivalent to the simple ratio,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A parameter calculation unit for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A small circle calculation unit for obtaining a small circle having a radius R centered on the measurement position of the measurement point when observed from one of the two observation points;
For a plurality of measurement points in the measurement space, a plurality of small circles obtained while each calculation in the parameter calculation unit and the small calculation unit is repeated a plurality of times while the parameter is changed by the parameter change unit. By obtaining the intersection of small circles when drawn in the small circle drawing space, the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles intersecting at the intersection and / or the measurement plane And a detection unit for _{obtaining a} normalized shortest distance _n d _s0 for
In this case, the measurement point appearing on the image has intensity information, and the small circle calculation unit obtains the small circle and draws the obtained small circle in the small circle drawing space. Voting a value corresponding to the strength of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when
Instead of obtaining the intersection, the detection unit obtains a local maximum point at which a value obtained by voting during each calculation in the parameter computation unit and the small circle computation unit is repeated a plurality of times, thereby obtaining the local maximum. It is preferable to determine the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to a plurality of small circles participating in the point voting and / or the normalized shortest distance _n d _s0 for the measurement plane, or
Measurement points appearing on the image have intensity information,
The calculation unit uses the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a second parameter, and sets the value of the second parameter. A second parameter changing unit for changing,
The parameter calculation unit includes a standardized shortest distance _n d _s set by the parameter changing unit, a position p _{axis of the} infinity point, and one observation point of the two observation points of the measurement point The radius R is obtained using the measurement position p _R when observed from the binocular parallax σ set by the second parameter changing unit,
The small circle conversion unit obtains the small circle corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained small circle in the small circle drawing space Voting the response intensity corresponding to the binocular parallax of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when
For each of a plurality of measurement points in the measurement space, each calculation in the second parameter calculation unit and the small circle conversion unit is changed, and each value of each parameter is changed in the parameter change unit and the second parameter change unit. However, instead of obtaining the intersection, the detection unit obtains a maximum point at which the value obtained by voting during the first, fifth, second, and third steps is repeated a plurality of times instead of obtaining the intersection. By obtaining the azimuth n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles participating in the voting of the local maximum point, and / or the normalized shortest distance _n d _s0 for the measurement plane is obtained. It is also a preferred embodiment that it is a step.

In the nineteenth image measuring apparatus, the calculation unit employs each position projected on a spherical surface as each position p _axis , p _R , and p _L , and uses the shortest distance as a physical quantity. , Where d _{s is} the shortest distance between the measurement plane and one of the two observation points, and Δx _LR is the distance between the two observation points.
_n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A first parameter changing unit that changes the position p _{axis of the} infinity point by changing the value of the first parameter using the visual axis direction v as a first parameter;
A second parameter changing unit that changes the value of the second parameter using the normalized shortest distance _n d _s as a second parameter;
The position p _{axis of the} infinity point set by the first parameter changing unit, the normalized shortest distance _n d _s set by the parameter changing unit, and a single ratio (p _axis p _R p _L ) or Using an operation equivalent to the simple ratio,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A parameter calculation unit for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A small circle calculation unit for obtaining a small circle with a radius R centered on the measurement position of the measurement point when observed from one of the two observation points;
With respect to a plurality of measurement points in the measurement space, each calculation in the parameter calculation unit and the small circle calculation unit is performed by using each value of the first parameter and the second parameter as the first parameter change unit and Drawing a plurality of small circles obtained while being changed a plurality of times while being changed by the second parameter calculation unit, corresponding to small circle drawing in a plurality of small circle drawing spaces according to the value of the first parameter By determining the intersections between the small circles when drawn in space for each small circle drawing space, and selecting the small circle drawing space corresponding to the true visual axis direction based on the information on the number of small circles intersecting at the intersection A measurement plane including a plurality of measurement points corresponding to a plurality of small circles intersecting at intersections obtained for a small circle drawing space corresponding to the true visual axis direction while obtaining a true visual axis direction. n _so, and / or it may be one equipped with a detection unit for determining the normalized shortest distance _n d _s0 for the measurement plane,
In this case, the measurement point appearing on the image has intensity information, and the small circle calculation unit obtains the small circle, and the obtained small circle corresponds to the small circle. A value corresponding to the intensity of the measurement point corresponding to the small circle is voted for each point on the path of the small circle when drawn in a circle drawing space, and the detection unit obtains the intersection Instead of this, information on the local maximum value at each local maximum point is obtained for each small circle drawing space by determining the local maximum point at which the value obtained by voting becomes a local maximum value while each calculation in the parameter arithmetic unit and the small circle arithmetic unit is repeated several times. And determining the true visual axis direction by selecting the small circle drawing space corresponding to the true visual axis direction based on the maximum point obtained for the small circle drawing space corresponding to the true visual axis direction. Multiple small circles that participated in voting A corresponding plurality of orientations n _so the measurement plane including a measurement _point, and / or, preferably and requests the normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
The calculation unit uses the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a third parameter, and sets the value of the third parameter. A third parameter changing unit for changing,
The parameter calculation unit includes the position p _{axis of the} infinity point set by the first parameter changing unit, the normalized shortest distance _n d _s set by the second step, and the measurement point The radius R is obtained by using the measurement position p _R when observed from one of the two observation points and the binocular parallax σ set by the third parameter changing unit. Step,
The small circle calculation unit obtains the small circle corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and corresponds the obtained small circle to the small circle. Voting the response intensity corresponding to the binocular parallax σ of the measurement point corresponding to the small circle for each point on the locus of the small circle when drawn in the small circle drawing space;
Instead of the detection unit obtaining the intersection, the first, second, and third parameters are calculated in the parameter calculation unit and the small circle calculation unit for a plurality of measurement points in the measurement space. The maximum value at which the value obtained by voting during a plurality of repetitions is changed while changing each value of each parameter in the changing unit is determined for each small circle drawing space, and the true value is calculated based on the information on the maximum value at the maximum point. The true visual axis direction is obtained by selecting a small circle drawing space corresponding to the visual axis direction, and a plurality of participants who participated in voting for the maximum points obtained for the small circle drawing space corresponding to the true visual axis direction orientation n _so the measurement plane including a plurality of measurement points corresponding to the small _circle, and / or, it is also a preferred embodiment is intended to determine the normalized shortest distance _n d _s0 for the measurement plane.

Furthermore, the twentieth image measurement device of the image measurement devices of the present invention is
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. When the measurement position is p _R , p _L , the measurement point, and the position of the point at infinity on a straight line extending in the same direction as the visual axis direction v connecting the two observation points is p _axis ,
Using the single ratio (p _axis p _R p _L ) determined by the three positions p _axis , p _R , p _L , or an operation equivalent to the single ratio, the measurement point and the two observation points An arithmetic unit for obtaining a physical quantity indicating the distance to one observation point is provided.

Here, in the calculation of the twentieth image measuring apparatus, which is executed by the calculation unit, the single ratio (p _axis p _R p _L ) or the calculation equivalent to the single ratio includes the two measurement points. Instead of the two measurement positions p _R and p _L when observed from each observation point, the measurement position p _R when the measurement point is observed from one of the two observation points; The calculation includes the binocular parallax σ that is the difference in position between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points.

In the twentieth image measurement apparatus, as a physical quantity indicating the distance, the distance between the measurement point and one of the two observation points is d ₀ , and the distance between the two observation points is when the distance was set to Δx _LR,
n d ₀ = d ₀ / Δx _LR
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _axis p _R p _L )
Or a relational expression equivalent to the relational expression.

The twenty-first image measuring device of the image measuring device of the present invention is
A measurement plane including an arbitrary measurement point in the measurement space that appears in an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space; and one of the two observation points between observation point, a physical quantity indicative of the distance of the visual axis direction connecting the two observation points with each other, the parameter setting for setting the coordinates of the voting space defined by the azimuth n _s of the measuring plane as a parameter And
A measurement position p _R when the measurement point is observed from one of the two observation points, and an infinite point on a straight line including the measurement point and extending in the same direction as the visual axis direction. From the position p _axis and the coordinates in the voting space set by the parameter setting unit, the positions of the two measurement positions p _R and p _L when the measurement point is observed from the two observation points A binocular parallax calculation unit for obtaining a binocular parallax σ which is a difference between
A response intensity calculation unit for obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the measurement space is viewed from the two observation points;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the parameter setting unit in the voting space;
For each of a plurality of measurement points in the measurement space, each process in the binocular parallax calculation unit, the response intensity calculation unit, and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit. It is characterized by.

The twenty-second image measuring device of the image measuring device of the present invention is
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the inside of the measurement space can be viewed from the two observation points. A first parameter setting unit for setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and extending in the same direction as the visual axis direction, which appears in the image when
A physical quantity indicative of the distance of the visual axis direction between one of the observation point of said the measurement plane including a measurement point the two observation points, defined by the azimuth n _s of the measurement plane, the A second parameter setting unit that sets coordinates in the voting space according to the first parameter as a second parameter;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set by the first parameter setting unit, and the second parameter setting Binocular parallax σ which is the difference between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points, based on the coordinates in the voting space set in the unit Binocular parallax calculation unit for obtaining
A response intensity calculation unit for obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the measurement space is viewed from the two observation points;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the second parameter setting unit in a voting space corresponding to the first parameter;
For each of the plurality of measurement points in the measurement space, each process in the binocular parallax calculation unit, the response intensity calculation unit, and the voting unit is performed, and each value of each parameter is set in the first and second parameter setting units. It is characterized by being executed multiple times while changing.

The twenty-third image measuring device of the image measuring device of the present invention is
One observation point of two predetermined observation points in the measurement space for observing the predetermined measurement space, and the measurement space appearing in an image when the measurement space is viewed from the two observation points a parameter setting unit for setting a physical quantity that indicates the shortest distance between the measurement plane including the arbitrary measurement points, the coordinates of the voting space defined by the azimuth n _s of the measuring plane as a parameter of the inner,
A measurement position p _R when the measurement point is observed from one of the two observation points, and the visual axis direction that includes the measurement point and connects the two observation points. Two measurement positions p when the observation point is observed from the two observation points, based on the position p _{axis of the} infinity point on the straight line and the coordinates in the voting space set by the parameter setting unit. _A binocular parallax calculation unit for obtaining a binocular parallax σ which is a difference in position between _{R 1} and p _L ;
A response intensity calculation unit for obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the measurement space is viewed from the two observation points;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the parameter setting unit in the voting space;
For each of a plurality of measurement points in the measurement space, each process in the binocular parallax calculation unit, the response intensity calculation unit, and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit. It is characterized by.

Furthermore, the 24th image measurement device of the image measurement devices of the present invention is:
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the measurement space is viewed from the two observation points. A first parameter setting unit for setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and extending in the same direction as the visual axis direction, which appears in the image of time
And one of the observation point of said two observation points, the physical quantity that indicates the shortest distance to the measurement plane including a measurement point is defined by the azimuth n _s of the measurement plane, said first parameter A second parameter setting unit that sets the coordinates in the corresponding voting space as a second parameter;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set by the first parameter setting unit, and the second parameter setting Binocular parallax σ which is the difference between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points, based on the coordinates in the voting space set in the unit Binocular parallax calculation unit for obtaining
A response intensity calculation unit for obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the measurement space is viewed from the two observation points;
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates set by the second parameter setting unit in a voting space corresponding to the first parameter;
For each of the plurality of measurement points in the measurement space, each process in the binocular parallax calculation unit, the response intensity calculation unit, and the voting unit is performed, and each value of each parameter is set in the first and second parameter setting units. It is characterized by being executed multiple times while changing.

The 25th image measuring device of the image measuring device of the present invention is:
Between two measurement positions p _R and p _{L at} an arbitrary measurement point in the measurement space, which appear in two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space A parameter setting unit that sets binocular parallax σ, which is a difference in position of
A measurement position p _R when the measurement point is observed from one of the two observation points and the visual axis direction that includes the measurement point and connects the two observation points. From the position p _{axis of the} infinity point on the straight line and the binocular parallax σ set by the parameter setting unit, between the measurement plane including the measurement point and one of the two observation points of a physical quantity indicative of the distance of the visual axis direction, and a coordinate calculation unit for obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
Based on two images when the measurement space is viewed from the two observation points, a response intensity calculation unit that obtains a response intensity corresponding to the binocular parallax σ set in the first step of the measurement point When,
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates obtained by the coordinate computing unit in the voting space;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit. And

The twenty-sixth image measuring device of the image measuring device of the present invention is
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the measurement space is viewed from the two observation points. A first parameter setting unit for setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space appearing in the image of time and extending in the same direction as the visual axis direction;
A second parameter setting unit configured to set, as a second parameter, binocular parallax σ that is a difference between two measurement positions p _R and p _L when the measurement point is observed from the two observation points; ,
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set by the first parameter setting unit, and the second parameter setting From the binocular parallax set in the unit, a physical quantity indicating the distance in the visual axis direction between the measurement plane including the measurement point and one of the two observation points; is defined by the azimuth n _s measurement plane, a coordinate calculation unit for obtaining the coordinates of the voting space corresponding to the first parameter,
Based on two images when the measurement space is viewed from the two observation points, a response intensity calculation unit that obtains a response intensity corresponding to the binocular parallax σ set in the second step of the measurement point When,
A voting unit for voting the response strength obtained by the response strength computing unit in the voting space according to the first parameter to the coordinates obtained by the coordinate computing unit;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is changed, and each value of each parameter is changed in the first and second parameter setting units. It is characterized by being executed multiple times.

The twenty-seventh image measuring device of the image measuring device of the present invention is
Between two measurement positions p _R and p _L of arbitrary measurement points in the measurement space, which appear in two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space. A parameter setting unit that sets binocular parallax σ, which is a difference in position between the two, as a parameter;
A measurement position p _R when the measurement point is observed from one of the two observation points and the visual axis direction that includes the measurement point and connects the two observation points. A measurement plane including one observation point of the two observation points and the measurement point based on the position p _{axis of the} infinity point on the straight line and the binocular parallax σ set by the parameter setting unit A coordinate calculation unit for obtaining coordinates in a voting space defined by a physical quantity that indicates the shortest distance between and a direction n _{s of the} measurement plane;
Based on two images when the measurement space is viewed from the two observation points, a response intensity calculation unit that obtains a response intensity corresponding to the binocular parallax σ set in the first step of the measurement point When,
A voting unit for voting the response strength obtained by the response strength computing unit to the coordinates obtained by the coordinate computing unit in the voting space;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit. And

Further, the twenty-eighth image measurement device of the image measurement devices of the present invention is:
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the inside of the measurement space can be viewed from the two observation points. A first parameter setting unit for setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and extending in the same direction as the visual axis direction, which appears in the image when
A second parameter setting unit configured to set, as a second parameter, binocular parallax σ that is a difference between two measurement positions p _R and p _L when the measurement point is observed from the two observation points; ,
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set by the first parameter setting unit, and the second parameter setting A physical quantity that indicates the shortest distance between one of the two observation points and the measurement plane including the measurement point, based on the binocular parallax set in the unit, and the measurement plane of it is defined by the azimuth n _s, and a coordinate calculation unit for obtaining the coordinates of the voting space corresponding to the first parameter,
Response intensity for obtaining response intensity corresponding to binocular parallax σ set by the second parameter setting unit of the measurement point based on two images when the measurement space is viewed from the two observation points An arithmetic unit;
A voting unit for voting the response strength obtained by the response strength computing unit in the voting space according to the first parameter to the coordinates obtained by the coordinate computing unit;
For each of a plurality of measurement points in the measurement space, each process in the coordinate calculation unit, the response strength calculation unit, and the voting unit is changed, and each value of each parameter is changed in the first and second parameter setting units. It is characterized by being executed multiple times.

Moreover, the 29th image measuring device of the image measuring devices of this invention is
When observing an arbitrary measurement point in the measurement space from the two observation points based on two images when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space A response intensity calculation unit for obtaining a response intensity corresponding to binocular parallax, which is a difference between the two measurement positions;
The visual axis direction connecting the two observation points between the measurement plane including the measurement points and one of the two observation points, the response intensity obtained by the response intensity calculation unit A voting unit for voting on coordinates corresponding to the measurement points and the binocular parallax in a voting space defined by a physical quantity that indicates the distance of the measurement plane and the orientation of the measurement plane;
Each process in the response intensity calculation unit and the voting unit is executed a plurality of times for a plurality of measurement points in the measurement space.

  Here, the twenty-ninth image measurement apparatus obtains a maximum point in the voting space where the value obtained by voting becomes a maximum value, thereby including a plurality of measurement points participating in the voting of the maximum point. And / or a detector that obtains a physical quantity that indicates the distance in the visual axis direction between the measurement plane and one of the two observation points. Good.

The thirtieth image measuring device of the image measuring device of the present invention is
A parameter setting unit for setting a visual axis direction connecting two predetermined observation points in the measurement space as a parameter for observing the predetermined measurement space;
Based on two images when the inside of the measurement space is viewed from the two observation points, a position between two measurement positions of the arbitrary measurement point in the measurement space when observed from the two observation points A response intensity calculation unit for obtaining a response intensity corresponding to binocular parallax that is a difference between
The response intensity obtained by the response intensity calculator is a physical quantity that indicates the distance in the visual axis direction between the measurement plane including the measurement point and one of the two observation points. A voting unit for voting on coordinates corresponding to the measurement point and the binocular parallax in a voting space according to the parameter set in the first step defined by the orientation of the measurement plane Prepared,
For each of a plurality of measurement points in the measurement space, each process in the response intensity calculation unit and the voting unit is executed a plurality of times while changing a parameter value in the first parameter setting unit.

  Here, the thirtieth image measurement device obtains a local maximum point at which the value obtained by voting becomes a local maximum value for each voting space, and makes the measurement point relative to the observation point based on information on the local maximum value at the local maximum point. The voting space corresponding to the true visual axis direction is selected to obtain the true visual axis direction, and a plurality of participants who participated in the voting of the maximum points obtained for the voting space corresponding to the true visual axis direction A detection unit for obtaining a physical quantity that indicates the orientation of the measurement plane including the measurement point and / or the distance in the visual axis direction between the measurement plane and one of the two observation points. It may be provided.

The thirty-first image measuring device of the image measuring device of the present invention is
When an arbitrary measurement point in the measurement space is observed from the two observation points based on two images when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space. A response intensity calculation unit for obtaining a response intensity corresponding to binocular parallax, which is a difference between two measurement positions;
The response intensity obtained by the response intensity calculator is a physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the measurement plane A voting unit for voting on coordinates corresponding to the measurement point and the binocular parallax σ in a voting space defined by an orientation,
For each of a plurality of measurement points in the measurement space, each process in the response intensity calculation unit and the voting unit is executed a plurality of times.

  Here, the thirty-first image measuring apparatus includes a measurement plane including a plurality of measurement points participating in the voting of the local maximum point by obtaining a local maximum point in the voting space where the value by the voting becomes a local maximum value. And / or a detector that obtains a physical quantity that indicates the shortest distance between one observation point of the two observation points and the measurement plane.

The thirty-second image measuring device of the image measuring device of the present invention is
A parameter setting unit that observes the inside of a predetermined measurement space, sets a visual axis direction connecting two predetermined observation points in the measurement space as a parameter,
Based on two images when the inside of the measurement space is viewed from the two observation points, a position between two measurement positions of the arbitrary measurement point in the measurement space when observed from the two observation points A response intensity calculation unit for obtaining a response intensity corresponding to binocular parallax that is a difference between
The response intensity obtained by the response intensity calculator is a physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the measurement plane A voting unit for voting on coordinates corresponding to the measurement point and the binocular parallax in the voting space according to the parameter set in the first step, which is defined by the azimuth,
For each of a plurality of measurement points in the measurement space, each process in the response intensity calculation unit and the voting unit is executed a plurality of times while changing parameter values in the parameter setting unit.

  Here, the thirty-second image measuring apparatus obtains a local maximum point at which the value obtained by voting becomes a local maximum value for each voting space, and makes the relative measurement point relative to the observation point based on information on the local maximum value at the local maximum point. The voting space corresponding to the true visual axis direction is selected to obtain the true visual axis direction, and a plurality of participants who participated in the voting of the maximum points obtained for the voting space corresponding to the true visual axis direction A detection unit that obtains a physical quantity that indicates the azimuth of the measurement plane including the measurement point and / or the shortest distance between one of the two observation points and the measurement plane. Also good.

In addition, the image measurement program stored in the first image measurement program storage medium of the present invention is the image measurement program stored in the image measurement program storage medium.
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , relative to the observation point, in the same direction as the movement direction v between the two measurement times, and the same as the movement speed between the two measurement times The position of the measurement point after infinite time has elapsed in the scheduled movement continuation state where the movement at the speed is continued as is, p _inf , and the measurement plane including the measurement point in the movement continuation state is superimposed on the observation point when the position of the measurement point in time was p _c,
The four positions _p inf measurement _point, p _0, p 1, using the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations determined by _{p c,} the orientation of the measurement plane, And / or an image measurement program for obtaining a physical quantity that indicates a superposition time at which the measurement plane overlaps the observation point.

Here, the first image measurement program storage medium a stored image measurement program runs, the compound ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations, the measuring point of the place of the two measuring positions p _0, p ₁ at two measurement time, the measurement point, the measurement position p ₀ at one measurement time of said two measurement time, the measurement point, Computation using motion parallax τ, which is the difference in position between the two measurement positions p ₀ and p ₁ at the two measurement times, is included.

The image measurement program stored in the first image measurement program storage medium uses a time between one measurement time of the two measurement times and the overlap time as a physical quantity indicating the overlap time. _c , when the time between the two measurement times is Δt,
_{n t c} = t _c / _Δt
The normalized time _n t _c expressed by the following is adopted, and the normalized time _n t _c is
_{n t c = {p inf p} 0 p 1 p c}
Or a program obtained using a relational expression equivalent to the relational expression.

The image measurement program stored in the first image measurement program storage medium is:
A first step of setting a physical quantity indicating the superposition time as a parameter;
The physical quantity indicating the superposition time set in the first step and the measurement positions p ₀ , p _{1 at the} two measurement times of the measurement points, or these two measurement positions p ₀ , p ₁ alternative, the measurement point, movement between the two measurement positions in one of the measurement time p ₀ and the measurement point of the measurement time, the two measurement positions p ₀ in the two measurement _time, p ₁ How to and parallax tau, and a position _{p inf} of after infinite time in the moving continuous state, using the double ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent calculation, the of the measuring points a second step of obtaining a position p _c in the overlay time,
The measurement point, and a third step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c in the superimposed time,
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while sequentially changing the parameter values in the first step, and then
A plurality of polar lines intersecting at the intersections by obtaining intersections between the polar lines when the plurality of polar lines obtained while repeating the first to third steps a plurality of times are drawn in the polar drawing space. May be a program that executes the fourth step for obtaining the orientation of the measurement plane including a plurality of measurement points corresponding to and / or the physical quantity that indicates the superposition time at which the measurement plane overlaps the observation point,
In this case, the measurement point appearing on the image has intensity information, and the third step obtains the polar line and draws the obtained polar line in the polar drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the local maximum points by obtaining a local maximum point at which the value by voting becomes a maximum value while repeating the first to third steps a plurality of times It is preferable that it is a step of obtaining a physical quantity that indicates the orientation of the measurement plane and / or the superposition time at which the measurement plane overlaps the observation point, or
Measurement points appearing on the image have intensity information,
The image measurement program includes a fifth step of setting a motion parallax τ between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a second parameter. ,
In the second step, the physical quantity that is set in the first step and that indicates the superposition time, and the measurement position p _{0 at} one of the two measurement times of the measurement point, using the the motion parallax τ set in the fifth step, the position p _inf after infinite time has elapsed in the mobile continuity state of the measurement point, determining the position p _c in the superimposed time of the measurement point And
When the third step obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained polar line in the polar drawing space Voting the response intensity corresponding to the motion parallax τ of the measurement point corresponding to the polar line for each point on the trajectory of the polar line,
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter obtains the maximum point where the value obtained by voting becomes the maximum value while repeating the first, fifth, second and third steps a plurality of times instead of obtaining the intersection. Thus, the orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines participating in the voting of the local maximum point and / or the physical quantity that indicates the superposition time at which the measurement plane overlaps the observation point is obtained. It is also a preferred embodiment that it is a step.

The image measurement program stored in the first image measurement program storage medium is:
A first step of setting a position p _inf of the measurement point after an infinite time in the moving continuous state by setting the moving direction v as the first parameter,
A second step of setting a physical quantity indicating the superposition time as a second parameter;
The position p _inf set in the first step, the physical quantity indicating the superposition time set in the second step, and the measurement positions p ₀ , p at the two measurement times of the measurement point ₁ or, alternative to these two measurement positions p _0, p _1, of the measurement point, the two one measurement position p ₀ and the measurement point in the measurement time of the measurement time, in the two measurement time and a motion parallax τ between two measuring positions _p 0, _{p 1} if and using equivalent operations and the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio, the superposition of the measuring points a third step of obtaining a position p _c at time,
The measurement point, and a fourth step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c in the superimposed time,
For the plurality of measurement points in the measurement space, the third step and the fourth step of the first to fourth steps are changed to values of the first parameter and the second parameter, respectively. Are repeated a plurality of times while sequentially changing in the first step and the second step, respectively,
A plurality of polar lines obtained while repeating the first to fourth steps a plurality of times are respectively transferred to corresponding polar line drawing spaces among a plurality of polar line drawing spaces according to the value of the first parameter. A pole corresponding to the true movement direction relative to the observation point of the measurement point based on information on the number of polar lines intersecting at the intersection, where intersections between polar lines at the time of drawing are obtained for each polar line drawing space A measurement including a plurality of measurement points corresponding to a plurality of polar lines intersecting at intersections obtained for the polar drawing space corresponding to the true movement direction while obtaining the true movement direction by selecting a line drawing space It may be a program for executing a fifth step for obtaining a physical quantity indicating the orientation of the plane and / or the superposition time at which the measurement plane overlaps the observation point,
In this case, the measurement point appearing on the image has intensity information, and the fourth step obtains the polar line, and the obtained polar line corresponds to the polar line. Voting a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line when drawn in the line drawing space, and the fifth step is a step of voting the intersection. Instead of obtaining the maximum point where the value obtained by voting during the repetition of the first to fourth steps a plurality of times is determined for each polar drawing space, and based on the information on the maximum value at the maximum point. The true movement direction is determined by selecting a polar drawing space corresponding to the true movement direction, and a plurality of participants who participated in voting for the maximum points obtained for the polar drawing space corresponding to the true movement direction Double line corresponding to Azimuth measurement plane including a measurement point, and / or, or preferably the measurement plane is a step of obtaining a physical quantity indicative of the superimposed time overlapping the observation point,
Measurement points appearing on the image have intensity information,
The image measurement program includes a sixth step of setting a motion parallax τ between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter. ,
In the third step, the physical quantity indicating the superposition time set in the second step, the position p _inf set in the first step, and the two measurement times of the measurement point a measurement position p ₀ in the measurement time of one among the sixth using said motion parallax τ set in step, a step of determining the position p _c in the superimposed time of the measurement points,
When the fourth step obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and draws the obtained polar line in the polar drawing space Voting a response intensity corresponding to the motion parallax τ of a measurement point corresponding to the polar line for each point on the trajectory of the polar line,
For each of a plurality of measurement points in the measurement space, the third step and the fourth step are changed, and the values of the parameters are changed in the first step, the second step, and the sixth step. While repeating several times,
The fifth step executed thereafter is not the determination of the intersection, but the maximum value obtained by voting while repeating the first, second, sixth, third and fourth steps a plurality of times. A point is obtained for each polar drawing space, and the true moving direction is obtained by selecting the polar drawing space corresponding to the true moving direction based on the local maximum information at the local maximum point, and the true moving direction is obtained. The orientation of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the maximum points obtained for the polar drawing space corresponding to the moving direction, and / or the measurement plane as the observation point It is also a preferred form that it is a step of obtaining a physical quantity that indicates the overlapping superposition time.

Of the image measurement program storage medium of the present invention, the second image measurement program storage medium of the present invention is:
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , v is the moving direction between the two measurement times relative to the observation point, v, and the two measurement times are relative to the observation point. In the same direction as the moving direction v between and at the same speed as the moving speed between the two measurement times, the measurement point after the infinite time has elapsed in the scheduled movement continuation state P _inf , the position of the measurement point at the time of superimposition where the measurement plane including the measurement point overlaps the observation point in the movement continuation state is p _c , and the orientation of the measurement plane is n _s ,
Four positions _p inf of the measurement _points, and p _0, p 1, _p cross ratios determined by _c and the equivalent operations _{{p inf p 0 p 1 p} c} or plurality ratio, orientation _{n s} of the measurement plane Using the inner product ( _ns / v) with the moving direction v, the measurement plane at the measurement time of one of the two measurement times from the observation point azimuth _ns and / or the observation point An image measurement program for obtaining a physical quantity indicating the shortest distance is stored.

Here, the second image measuring program storage medium a stored image measurement program runs, the compound ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent operations, the measuring point of the place of the two measuring positions p _0, p ₁ at two measurement time, the measurement point, the measurement position p ₀ at one measurement time of said two measurement time, the measurement point, Computation using motion parallax τ, which is the difference in position between the two measurement positions p ₀ and p ₁ at the two measurement times, is included.

The image measurement program stored in the second image measurement program storage medium includes the observation point and the measurement plane at one of the two measurement times as a physical quantity having the shortest distance as an index. D _s , the time between one measurement time of the two measurement times and the weight time t _c , and the measurement point between the two measurement times, When the relative movement distance with respect to the observation point is Δx, and the time between the two measurement times is Δt,
_n d _s = d _s / Δx
The standardized shortest distance _n d _s expressed by the following formula is adopted, and the standardized shortest distance _n d _{s
n t c} = t _c / _Δt
Using the normalized time _n t _c expressed by: and the inner product (n _s · v),
_{n d s = n t c (} n s · v)
It may be a program obtained using the relational expression.

The image measurement program stored in the second image measurement program storage medium is:
A first step of setting a physical quantity indicating the shortest distance as a first parameter;
A second step of setting the inner product ( _ns / v) as a second parameter;
The physical quantity that is set in the first step and that indicates the shortest distance, the inner product ( _ns / v) that is set in the second step, and each measurement position at the two measurement times of the measurement point p _0, p ₁ or the two measuring positions p _0, replaces the p _1, of the measurement point, the measurement position p ₀ and the measurement point at one measurement time of said two measurement time, the 2 From the motion parallax between the two measurement positions p ₀ and p _{1 at} one measurement time and the position p _inf after the infinite time has elapsed in the movement continuation state, the cross ratio {p _inf p ₀ p ₁ p _c } or using a plurality ratio equivalent operation, a third step of obtaining a position p _c in the superimposed time of the measurement points,
A fourth step of obtaining the polar corresponding to the position p _c by converting the measurement point, the position p _c in the superimposed time pole,
Said a point of pole line, and the moving direction v and angle _{^{r = cos -1 (n s ·}} v)
And a fifth step for obtaining a point comprising
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are set as the respective values of the first parameter and the second parameter. By repeating a plurality of times while sequentially changing in the first step and the second step, the value of the first parameter for one measurement point in the fifth step is the same, and the first step A curve connecting a plurality of points obtained by a plurality of executions with different values of the parameter 2 is obtained for each of the plurality of measurement points for each value of the first parameter;
A plurality of curves corresponding to a plurality of curves intersecting at the intersection are obtained by obtaining intersections between the curves when the plurality of curves obtained while repeating the first to fifth steps a plurality of times are drawn in the curve drawing space. orientation n _s measurement plane including a measurement _point, and / or, a sixth step of obtaining a physical quantity that indicates the shortest distance to the measuring plane in the measurement time of one of the two measurement time from the observation point It can be a program to be executed,
In that case, the measurement point appearing on the image has intensity information, and the fifth step obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to the point, wherein the sixth step repeats the first to fifth steps a plurality of times instead of obtaining the intersection By obtaining a local maximum point at which the value obtained by voting becomes a local maximum value, the orientation n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves participating in the voting of the local maximum point and / or the observation point Preferably, it is a step of obtaining a physical quantity indicating the shortest distance to the measurement plane at one of the two measurement times, or
Measurement points appearing on the image have intensity information,
The image measurement program includes a seventh step of setting a motion parallax τ between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter. ,
In the third step, the physical quantity indexed in the shortest distance set in the first step, the inner product ( _ns / v) set in the second step, and the measurement point, The measurement position p _{0 at} one of the two measurement times, the motion parallax τ set in the seventh step, and the position p of the measurement point after the infinite time has elapsed in the movement continuation state by using the _inf, of the measurement point, a step of determining the position p _c in the superimposed time,
The fifth step obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and measures the measurement point corresponding to the point on the polar line Voting a response intensity corresponding to the motion parallax τ to a point in the curve drawing space corresponding to the point on the polar line,
Steps 3 to 5 are repeated for a plurality of measurement points in the measurement space a plurality of times while changing each value of each parameter in the first, second and seventh steps, and then executed. Instead of obtaining the intersection, the sixth step is a value obtained by voting while repeating the first, second, seventh and third to fifth steps a plurality of times. By obtaining the local maximum point, the direction n _{s of the} measurement plane including a plurality of measurement points corresponding to the plurality of curves that participated in the vote of the local maximum point, and / or the two measurement times from the observation point It is also a preferred aspect that it is a step of obtaining a physical quantity indicating the shortest distance to the measurement plane at one measurement time.

The image measurement program stored in the second image measurement program storage medium is:
A first step of setting a position p _inf of the measurement point after an infinite time in the moving continuous state by setting the moving direction v as the first parameter,
A second step of setting a physical quantity indicative of the shortest distance as a second parameter;
A third step of setting the inner product ( _ns / v) as a third parameter;
The position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step, the physical quantity indicating the shortest distance set in the second step, and the first The inner product ( _ns · v) set in step 3 and the measurement positions p ₀ , p _{1 at the} two measurement times of the measurement point, or these two measurement positions p ₀ , p ₁ , The measurement position p ₀ of the measurement point at one of the two measurement times and the motion parallax τ between the two measurement positions p ₀ and p _{1 of the} measurement point at the two measurement times from using the cross ratio _{{p inf p 0 p 1 p} c} or plurality ratio equivalent calculation, a fourth step of obtaining a position _{p c} in the superimposed time of the measurement points,
A fifth step of obtaining a polar corresponding to the position p _c by converting the measurement point, the position p _c in the superimposed time pole,
Said a point of pole line, and the moving direction v and angle _{^{r = cos -1 (n s ·}} v)
And a sixth step for obtaining a point comprising
For the plurality of measurement points in the measurement space, the fourth to sixth steps of the first to sixth steps, the respective values of the first to third parameters, respectively, By repeating a plurality of times while sequentially changing each step from 1 to 3, the value of the first parameter for one measurement point in the sixth step is the same and the second parameter Curves connecting a plurality of points that are obtained by a plurality of executions having the same value and different values of the third parameter are represented as values of the first parameter and values of the second parameter. For each of the combinations and the measurement points, and then
When a plurality of curves obtained while repeating the first to sixth steps a plurality of times are drawn in the corresponding curve drawing spaces among the plurality of curve drawing spaces according to the value of the first parameter. Obtaining an intersection between the curves for each curve drawing space, and selecting a curve drawing space corresponding to the true movement direction of the measurement point relative to the observation point based on information on the number of curves intersecting at the intersection. To determine the true movement direction, and the orientation n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection obtained with respect to the curve drawing space corresponding to the true movement direction, and / or A program for executing a seventh step of obtaining a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times It may be a beam,
In this case, the measurement point appearing on the image has intensity information, and the sixth step obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to the point in which a curve including the point is drawn, wherein the seventh step is to replace the first to sixth in place of obtaining the intersection A maximum point at which the value obtained by voting during a plurality of steps is maximized is obtained for each curve drawing space, and a curve drawing space corresponding to the true movement direction is selected based on information on the maximum value at the maximum point. The direction n of the measurement plane including a plurality of measurement points corresponding to a plurality of curves participating in the voting of the maximum points obtained for the curve drawing space corresponding to the true movement direction _s And / or is preferably a step of obtaining a physical quantity that indicates the shortest distance from the observation point to the measurement plane in one of the two measurement times, or
Measurement points appearing on the image have intensity information,
The image measurement program includes an eighth step of setting a motion parallax τ between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a fourth parameter. ,
In the fourth step, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step and the shortest distance set in the second step. The physical quantity to be indexed, the inner product ( _ns · v) set in the third step, the measurement position p _{0 of the} measurement point at one of the two measurement times, and the eighth using said motion parallax τ set in step, of the measurement point, a step of determining the position p _c in the superimposed time,
The sixth step obtains the point on the polar line corresponding to the measurement point, obtains a response intensity corresponding to the motion parallax τ of the measurement point, and measures the measurement point corresponding to the point on the polar line Voting a response intensity corresponding to the motion parallax τ to a point in the curve drawing space corresponding to the point on the polar line,
For a plurality of measurement points in the measurement space, the steps from the fourth to the sixth are repeated a plurality of times while changing each value of each parameter in the first, second, third and eighth steps, The value obtained by voting during the repetition of the first, second, third, eighth, and fourth to sixth steps a plurality of times, instead of the seventh step executed thereafter obtaining the intersection. Finds a local maximum point for each curve drawing space, and obtains the true moving direction by selecting a curve drawing space corresponding to the true moving direction based on the local maximum information at the local maximum point. Azimuth n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the voting of local maximum points obtained for the curve drawing space corresponding to the true moving direction, and / or from the observation point Said It is also preferable is a One step of obtaining a physical quantity that indicates the shortest distance to the measurement plane at one measurement time of the measurement time.

The third image measurement program storage medium of the image measurement program storage medium of the present invention is:
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in the image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , v is the moving direction between the two measurement times relative to the observation point, v, and the two measurement times are relative to the observation point. In the same direction as the moving direction v between and at the same speed as the moving speed between the two measurement times, the measurement point after the infinite time has elapsed in the scheduled movement continuation state _Where p _inf is the position of
Using a single ratio (p _inf p ₀ p ₁ ) determined by the three positions p _inf , p ₀ , and p ₁ of the measurement point (p _inf p ₀ p ₁ ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the An image measurement program for obtaining a physical quantity indicating the shortest distance from an observation point to the measurement plane at one of the two measurement times is stored.

Here, for the calculation performed by the image measurement program stored in the third image measurement program storage medium, the unit ratio (p _inf p ₀ p ₁ ) or the equivalent of the unit ratio, two instead of the measurement position p _0, p ₁ in the two measurement time, the measurement point, the measurement position p ₀ at one measurement time of said two measurement time, the measurement point, the 2 Computation using motion parallax τ, which is a difference in position between two measurement positions p ₀ and p _{1 at} one measurement time, is included.

The image measurement program stored in the third image measurement program storage medium employs each position projected on a spherical surface as each position p _inf , p ₀ , p ₁ of the measurement point, and the shortest distance Is the shortest distance between the observation point and the measurement plane at one of the two measurement times, d _s , and the measurement point between the two measurement times, When the relative movement distance with respect to the observation point is Δx,
_n d _s = d _s / Δx
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the normalized shortest distance _n d _s as a parameter;
Using the normalized shortest distance _n d _s set in the first step and a simple ratio (p _inf p ₀ p ₁ ) or an operation equivalent to the simple ratio,
^{_{R = cos -1 (n d s}} / (p inf p 0 p 1))
A second step for obtaining a radius R defined by the relational expression or a relational expression equivalent to the relational expression,
A third step of obtaining a small circle with a radius R centered on the measurement position of the measurement point at one of the two measurement times;
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while sequentially changing the parameters in the first step, and then
A plurality of small circles intersecting at the intersections by obtaining intersections between the small circles when the plurality of small circles obtained while repeating the first to third steps a plurality of times are drawn in the small circle drawing space. the azimuth n _so the measurement plane including a plurality of measurement points _{corresponding,} and / or may be a program for executing the fourth step of determining a normalized shortest distance _n d _s0 for the measurement plane,
In this case, the measurement point appearing on the image has intensity information, and the third step obtains the small circle and draws the obtained small circle in the small circle drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the local maximum by obtaining a local maximum at which the value obtained by voting becomes a local maximum while repeating the first to third steps a plurality of times Preferably, this is the step of determining the orientation n _{so of the} measurement plane and / or the normalized shortest distance _n d _s0 for the measurement plane, or
Measurement points appearing on the image have intensity information,
The image measurement program includes a fifth step of setting a motion parallax τ between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a second parameter. ,
In the second step, the standardized shortest distance _n d _s set in the first step, the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state, and the 2 of the measurement point Using the measurement position p _{0 at} one of the two measurement times and the motion parallax τ set in the fifth step to determine the radius R;
In the third step, the small circle corresponding to the measurement point is obtained, the response intensity corresponding to the motion parallax τ of the measurement point is obtained, and the obtained small circle is drawn in the small circle drawing space. Voting a response intensity corresponding to the motion parallax τ of each measurement point corresponding to the small circle for each point on the trajectory of the small circle when
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter is not the determination of the intersection, but the maximum point at which the value by voting becomes a maximum value while repeating the first, fifth, second and third steps a plurality of times. By obtaining the azimuth n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles participating in the voting of the local maximum point, and / or the normalized shortest distance _n d _s0 for the measurement plane is obtained. It is also a preferred embodiment that it is a step.

Further, the image measurement program stored in the third image measurement program storage medium adopts each position projected on the spherical surface as each position p _inf , p ₀ , p ₁ of the measurement point, and As a physical quantity indicating the shortest distance, d _{s is} the shortest distance between the observation point and the measurement plane at one of the two measurement times, and the measurement point between the two measurement times. , When the relative movement distance with respect to the observation point is Δx,
_n d _s = d _s / Δx
In adopting the normalized shortest distance _n d _s represented,
A first step of setting a position p _inf of the measurement point after an infinite time in the moving continuous state by setting the moving direction v as the first parameter,
A second step of setting the normalized shortest distance _n d _s as a second parameter;
The position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step, the normalized shortest distance _n d _s set in the second step, and a simple ratio ( p _inf p ₀ p ₁ ) or an operation equivalent to the simple ratio,
^{_{R = cos -1 (n d s}} / (p inf p 0 p 1))
A third step for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A fourth step of obtaining a small circle having a radius R around the measurement position of the measurement point at one of the two measurement times;
For the plurality of measurement points in the measurement space, the third step and the fourth step are performed, and the values of the first parameter and the second parameter are set to the first step and the second step, respectively. Repeated several times while changing sequentially in steps, then
A plurality of small circles obtained while repeating the first to fourth steps a plurality of times are respectively transferred to corresponding small circle drawing spaces among the plurality of small circle drawing spaces according to the value of the first parameter. A small circle corresponding to a true moving direction relative to the observation point of the measurement point based on information on the number of small circles intersecting at the intersection, where intersections between the small circles when drawn are obtained for each small circle drawing space. A measurement plane including a plurality of measurement points corresponding to a plurality of small circles intersecting at intersections obtained for the small circle drawing space corresponding to the true movement direction while obtaining the true movement direction by selecting a drawing space orientation n _so, and / or may be a program for executing a fifth step of obtaining a normalized shortest distance _n d _s0 for the measurement plane,
In this case, the measurement point appearing on the image has intensity information, and the fourth step obtains the small circle, and the obtained small circle corresponds to the small circle. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in a circle drawing space, and the fifth step is for locating the intersection. Instead of obtaining the small circle maximum when the maximum point at which the value obtained by voting during the repetition of the first to fourth steps is a maximum value is drawn in the small circle drawing space corresponding to the small circle. Preferably, this is a step of obtaining the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to a plurality of small circles participating in the voting of points and / or the normalized shortest distance _n d _s0 for the measurement plane. Or
Measurement points appearing on the image have intensity information,
The image measurement program includes a sixth step of setting a motion parallax τ between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point as a third parameter. ,
The second step includes the position p _{inf of the} measurement point after the infinite time has elapsed in the movement continuation state set in the first step, and the normalized shortest distance _n set in the third step. The radius R is obtained using d _s , the measurement position p _{0 at} one of the two measurement times of the measurement point, and the motion parallax τ set in the fifth step. Step,
The fourth step obtains the small circle corresponding to the measurement point, obtains the response strength of the measurement point corresponding to the motion parallax τ, and determines the obtained small circle as a small circle corresponding to the small circle. Voting the response intensity corresponding to the motion parallax τ of each measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in a circle drawing space;
For a plurality of measurement points in the measurement space, the third step and the fourth step are repeated a plurality of times while changing each value of each parameter in the first, second and sixth steps,
Subsequent to the fifth step being executed, instead of obtaining the intersection, the value obtained by voting during the repetition of the first, second, sixth, third and fourth steps becomes a maximum value. A local maximum point is obtained for each small circle drawing space, and the true moving direction is obtained by selecting a small circular drawing space corresponding to the true moving direction based on information on the local maximum value at the local maximum point. Azimuth n _{so of} a measurement plane including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the maximum points obtained for the small circle drawing space corresponding to the moving direction of the circle, and / or It is also a preferred aspect that it is a step of _{obtaining the} normalized shortest distance _n d _s0 .

Furthermore, the fourth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Each measurement position at two different measurement times of an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space, is p. ₀ , p ₁ , relative to the observation point, in the same direction as the movement direction v between the two measurement times, and the same as the movement speed between the two measurement times When the position of the measurement point after elapse of an infinite time in a state where the movement at the speed is continued as it is and when the movement is continuously planned is p _inf ,
Using a single ratio (p _inf p ₀ p ₁ ) determined by the three positions p _inf , p ₀ , and p ₁ of the measurement point, or an operation equivalent to the single ratio,
An image measurement program for obtaining a physical quantity indicating a distance between the observation point and the measurement point at one of the two measurement times is stored.

For the calculation performed by the image measurement program stored in the fourth image measurement program storage medium, the single ratio (p _inf p ₀ p ₁ ) or the equivalent of the single ratio, Instead of the two measurement positions p ₀ and p ₁ at the measurement time, the measurement position p ₀ of the measurement point at one of the two measurement times and the two measurement times of the measurement point The calculation using the motion parallax τ that is the difference in position between the two measurement positions p ₀ and p ₁ is included.

The image measurement program stored in the fourth image measurement program storage medium is a physical quantity indicating the distance between the observation point and the measurement point at one of the two measurement times. When the distance is d ₀ and the movement distance of the measurement point between the two measurement times with respect to the observation point is Δx,
_n d ₀ = d ₀ / Δx
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _inf p ₀ p ₁ )
Or a program obtained using a relational expression equivalent to the relational expression.

The fifth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A measurement plane including the measurement point in a movement continuation state in which the movement in the same direction as the general movement direction and the movement speed at the same speed as the movement speed between the two measurement times is continued a physical quantity indicative of the superimposed time overlapping the observation point, a first step of setting as a parameter coordinates voting space defined by the azimuth n _s of the measurement plane,
In the first step, the measurement position p ₀ of the measurement point at one of the two measurement times, the position p _inf after the infinite time has elapsed in the movement continuation state, and the first step. A second step of obtaining a motion parallax τ which is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the set coordinates in the voting space;
A third step of determining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

The sixth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first step of setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
In the mobile continuity state, and the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, defined by the azimuth n _s of the measurement plane, voting space corresponding to the first parameter A second step of setting the coordinates in as a second parameter;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the voting set in the second step A third step of obtaining a motion parallax τ that is a difference between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from coordinates in space;
A fourth step of obtaining response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The seventh image measurement program storage medium of the image measurement program storage medium of the present invention is:
A predetermined observation point in the measurement space for observing the predetermined measurement space, and an arbitrary measurement point in the measurement space that appears in an image when the measurement space is viewed from the observation point. to set a physical quantity that indicates the shortest distance between the measurement plane at one measurement time of the two different measurement times, the coordinates of the voting space defined by the azimuth n _s of the measuring plane as a parameter A first step;
In the same direction as the relative movement direction of the measurement point relative to the observation point between the measurement point p ₀ at one of the two measurement times and the measurement point between the two measurement times. And the position p _inf after the infinite time has elapsed in the planned movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is, and is set in the first step. A second step of obtaining a motion parallax τ that is a difference between the two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from the coordinates in the voting space;
A third step of determining a response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

Further, the eighth image measurement program storage medium of the image measurement program storage medium of the present invention is displayed in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. By setting, as the first parameter, the relative movement direction v of the arbitrary measurement point with respect to the observation point between two different measurement times, the measurement point can be moved in the same direction as the movement direction. And a first step of setting the position p _{inf of the} measurement point after elapse of an infinite time in the scheduled movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is. When,
Wherein from the observation point, comprising said measuring points, and the physical quantity that indicates the shortest distance to the measurement plane at the measurement time of one of the two measurement time is defined by the azimuth n _s of the measurement plane, the A second step of setting the coordinates in the voting space according to the first parameter as a second parameter;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the voting set in the second step A third step of obtaining a motion parallax τ that is a difference between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point from coordinates in space;
A fourth step of obtaining response intensity corresponding to the motion parallax τ of the measurement point based on two images when the inside of the measurement space is viewed from the observation point at the two measurement times;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The ninth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Two measurement positions p ₀ , at two different measurement times of arbitrary measurement points in the measurement space, which appear in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. a first step of setting, as a parameter, motion parallax τ, which is a difference in position between p ₁ ;
In the same direction as the relative movement direction of the measurement point relative to the observation point between the measurement point p ₀ at one of the two measurement times and the measurement point between the two measurement times. And the position p _inf after elapse of an infinite time of the measurement point in the movement continuation state where the movement at the same speed as the movement speed between the two measurement times is continued as it is, and the first Is determined by a physical quantity that indicates a superimposition time at which the measurement plane including the measurement point overlaps the observation point and the orientation n _{s of the} measurement plane in the movement continuation state. A second step for determining coordinates in a voting space;
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the first step is obtained at the measurement point. A third step;
A fourth step of voting the response strength determined in the third step to the coordinates determined in the second step in the voting space;
For the plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while sequentially changing the parameter values in the first step. An image measurement program to be stored is stored.

The tenth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first step of setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
A second step of setting, as a second parameter, motion parallax τ, which is a difference in position between the two measurement positions p ₀ and p ₁ at the two measurement times, of the measurement point;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the motion parallax τ set in the second step and a, the in the mobile continuity state, and the physical quantity measuring plane including the measurement point indicative of the superimposed time overlapping the observation point, defined by the azimuth n _s of the measurement plane, according to the first parameter A third step for determining coordinates in the voting space;
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the second step is obtained at the measurement point. A fourth step;
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The eleventh image measurement program storage medium of the image measurement program storage medium of the present invention is:
Two measurement positions p ₀ , at two different measurement times of arbitrary measurement points in the measurement space, which appear in an image when the measurement space is viewed from a predetermined observation point in the predetermined measurement space. a first step of setting, as a parameter, motion parallax τ, which is a difference in position between p ₁ ;
The same direction as the relative movement direction of the measurement point with respect to the observation point between the two measurement times and the measurement position p ₀ at one of the two measurement times. And a position p _inf after the infinite time of the measurement point in a state where the movement continues at the same speed as the movement speed between the two measurement times and the scheduled movement continuation state, A physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point, based on the motion parallax τ set in step 1 When, a second step of obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the first step is obtained at the measurement point. A third step;
Voting the response strength determined in the third step to the coordinates determined in the second step in the voting space; and
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

Furthermore, the twelfth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space, which appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space By setting the general moving direction v as the first parameter, the movement of the measurement point in the same direction as the movement direction and at the same speed as the movement speed between the two measurement times continues. A first step of setting the position p _{inf of the} measurement point after an infinite time has elapsed in the scheduled movement continuation state;
A second step of setting, as a second parameter, motion parallax τ, which is a difference between two measurement positions p ₀ and p ₁ at the two measurement times of the measurement point;
The measurement position p _{0 of the} measurement point at one of the two measurement times, the position p _inf set in the first step, and the motion parallax τ set in the second step And a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point, and the direction n _{s of the} measurement plane A third step for obtaining coordinates in the voting space according to the first parameter, defined by:
Based on the two images when the measurement space is viewed from the observation point at the two measurement times, the response intensity corresponding to the motion parallax τ set in the second step is obtained at the measurement point. A fourth step;
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The thirteenth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Based on two images when the measurement space is viewed at two different measurement times from a predetermined observation point in the predetermined measurement space, any two measurement points in the measurement space A first step of obtaining a response intensity corresponding to motion parallax, which is a difference between positions of two measurement positions at one measurement time;
The response intensity obtained in the first step is set in the same direction as the movement direction of the measurement point relative to the observation point between the two measurement times and between the two measurement times. A physical quantity that indicates a superposition time at which a measurement plane including the measurement point overlaps the observation point in a planned movement continuation state in which the movement at the same speed as the movement speed is continued, and an orientation of the measurement plane A second step of voting on coordinates corresponding to the measurement points and the motion parallax within a defined voting space;
An image measurement program for executing the first step and the second step a plurality of times for a plurality of measurement points in the measurement space is stored.

  Here, the image measurement program stored in the thirteenth image measurement program storage medium obtains a maximum point in the voting space where the value by the voting becomes a maximum value, thereby voting the maximum point. The program may include a third step of obtaining a azimuth of a measurement plane including a plurality of participating measurement points and / or a physical quantity indicating a superposition time at which the measurement plane overlaps the observation point.

In addition, the fourteenth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A first step of setting a general moving direction as a parameter;
Based on two images when the measurement space is viewed from the observation point at the two measurement times, the movement is a difference between the two measurement positions at the two measurement times of the measurement point. A second step for obtaining a response intensity corresponding to the parallax;
The response intensity obtained in the second step is set in the same direction as the movement direction of the measurement point relative to the observation point between the two measurement times, and between the two measurement times. The physical quantity indicating the superposition time when the measurement plane including the measurement point overlaps the observation point in the planned movement continuation state in which the movement at the same speed as the movement speed is continued as it is and the direction of the measurement plane A third step of voting on coordinates corresponding to the measurement point and the motion parallax in a voting space defined by the parameters set in the first step,
An image in which the second and third steps of the first to third steps are executed a plurality of times while changing parameter values in the first step for a plurality of measurement points in the measurement space. A measurement program is stored.

  Here, the image measurement program stored in the fourteenth image measurement program storage medium obtains a maximum point at which the value obtained by voting becomes a maximum value for each voting space, and is based on information on the maximum value at the maximum point. The true movement direction is determined by selecting a voting space corresponding to the true movement direction relative to the observation point of the measurement point, and the maximum point obtained for the voting space corresponding to the true movement direction A program having a fourth step for obtaining the orientation of the measurement plane including a plurality of measurement points participating in the voting and / or the physical quantity indicating the superposition time at which the measurement plane overlaps the observation point may be used.

The fifteenth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Based on two images when the measurement space is viewed at two different measurement times from a predetermined observation point in the predetermined measurement space, the two measurements at arbitrary measurement points in the measurement space A first step of obtaining a response intensity corresponding to a motion parallax that is a difference between positions of two measurement positions at a time;
The response intensity obtained in the first step is a physical quantity indicating the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point. Voting to coordinates corresponding to the measurement point and the motion parallax in a voting space defined by the orientation of the measurement plane,
An image measurement program for executing the first step and the second step a plurality of times for a plurality of measurement points in the measurement space is stored.

  Here, the image measurement program stored in the fifteenth image measurement program storage medium participates in the vote of the maximum point by obtaining the maximum point in the voting space where the value by the vote becomes a maximum value. A third step of obtaining a physical quantity indicating the azimuth of the measurement plane including the plurality of measurement points and / or the shortest distance from the observation point to the measurement plane at one of the two measurement times It may be a program having

Further, the sixteenth image measurement program storage medium of the image measurement program storage medium of the present invention is:
Relative to the observation point between two different measurement times of any measurement point in the measurement space that appears in the image when the inside of the measurement space is viewed from the predetermined observation point in the predetermined measurement space A first step of setting a general moving direction as a parameter;
Motion parallax, which is the difference between the two measurement positions at the two measurement times of the measurement point, based on the two images when the measurement space is viewed from the observation point at the two measurement times. A second step for obtaining a response intensity corresponding to
The response intensity obtained in the second step is a physical quantity that indicates the shortest distance from the observation point to the measurement plane at one of the two measurement times, including the measurement point. A third step of voting for the measurement point and the coordinates corresponding to the motion parallax in the voting space according to the parameter set in the first step defined by the orientation of the measurement plane Have
An image in which the second and third steps of the first to third steps are executed a plurality of times while changing parameter values in the first step for a plurality of measurement points in the measurement space. A measurement program is stored.

  Here, the image measurement program storage medium stored in the sixteenth image measurement program storage medium obtains a local maximum point at which the value obtained by voting becomes a local maximum value for each voting space, and stores the local maximum information at the local maximum point. The true movement direction is obtained by selecting a voting space corresponding to the true movement direction relative to the observation point of the measurement point based on the voting space corresponding to the true movement direction. Physical quantity that indicates the orientation of the measurement plane including a plurality of measurement points that participated in the vote of the maximum point and / or the shortest distance from the observation point to the measurement plane at one of the two measurement times It may be a program having a fourth step for obtaining.

In the present invention, the seventeenth image measurement program storage medium of the image measurement program storage medium is:
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , respectively, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the visual axis direction v connecting the two observation points is represented by p _axis . wherein one of the observation points of the point, when on the viewing plane extending parallel to the measurement plane including the measurement point, the position of intersection between the straight line and the p _c,
The four positions _{p axis,} using p _R, p L, the cross ratio _{{p axis p R p L p} c} or plurality ratio equivalent operations determined by _{p c,} the orientation of the measurement plane, and / or An image measurement program for obtaining a physical quantity indicating a distance in the visual axis direction between the measurement plane and one of the two observation points is stored.

Here, for the calculation performed by the image measurement program stored in the seventeenth image measurement program storage medium, the _cross ratio {p _axis p _R p _L p _c } or an operation equivalent to the cross ratio, Instead of the two measurement positions p _R and p _L when observed from the two observation points, the measurement position p _R when the measurement point is observed from one of the two observation points. And a calculation using the binocular parallax σ which is a difference in position between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points. .

The image measurement program stored in the seventeenth image measurement program storage medium is a physical quantity that indicates a distance in the visual axis direction between the measurement plane and the two observation points. When the distance in the visual axis direction between one of the observation points is d _c and the distance between the two observation points is Δx _LR ,
_{n d c = d c / Δx} LR
In adopting the normalized distance _n d _c represented, the normalized distance _n d _c,
n d _c = {p _axis p _R p _L p _c }
Or a program obtained using a relational expression equivalent to the relational expression.

The seventeenth image measurement program stored in the seventeenth image measurement program storage medium is:
A first step of setting, as a parameter, a physical quantity that indicates a distance in the visual axis direction between the measurement plane and one of the two observation points;
A physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points set in the first step, and the two measurement points. Each measurement position p _R , p _L when observed from the observation point, or these measurement points instead of these two measurement positions p _R , p _L were observed from one of the two observation points a binocular parallax σ between the measurement position p _R and the measurement point, the two two measurement positions p _R when observed from the observation _point, and if p _L of the time, the position p _axis of the point at infinity from a second step of using the double ratio _{{p axis p R p L p} c} or plurality ratio equivalent calculation to determine the position _{p c} of the intersection on the observation plane,
And a third step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c of the intersection on the observation plane,
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while sequentially changing the parameter values in the first step, and then
A plurality of polar lines intersecting at the intersections by obtaining intersections between the polar lines when the plurality of polar lines obtained while repeating the first to third steps a plurality of times are drawn in the polar drawing space. A physical quantity that indicates the orientation of the measurement plane including a plurality of measurement points corresponding to and / or the distance in the visual axis direction between the measurement plane and one of the two observation points is obtained. It may be a program that executes the fourth step,
In this case, the measurement point appearing on the image has intensity information, and the third step obtains the polar line and draws the obtained polar line in the polar drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the local maximum points by obtaining a local maximum point at which the value by voting becomes a maximum value while repeating the first to third steps a plurality of times Preferably, it is a step of obtaining a physical quantity indicating the orientation of the measurement plane and / or the distance in the visual axis direction between the measurement plane and one of the two observation points, or
Measurement points appearing on the image have intensity information,
A fifth step in which the image measurement program sets binocular parallax σ between two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a second parameter. A program having
The second step is a physical quantity that indicates the distance in the visual axis direction between the measurement plane and one of the two observation points set in the first step; The measurement position p _R when observed from one of the two observation points of the measurement point, the binocular parallax σ set in the fifth step, and the position p _{axis of the} infinity point with bets, a step of obtaining an intersection p _c on the observation plane,
The third step obtains a polar line corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained polar line in the polar drawing space Voting a response intensity corresponding to the binocular parallax σ of a measurement point corresponding to the polar line for each point on the trajectory of the polar line when
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter obtains the maximum point where the value obtained by voting becomes the maximum value while repeating the first, fifth, second and third steps a plurality of times instead of obtaining the intersection. Thus, the orientation of the measurement plane including a plurality of measurement points corresponding to the plurality of polar lines participating in the vote of the local maximum point, and / or one observation point of the measurement plane and the two observation points It may be a step of obtaining a physical quantity indicating the distance in the visual axis direction between the two.

The image measurement program stored in the seventeenth image measurement program storage medium is a first step of setting the infinity point position p _axis by setting the visual axis direction v as a first parameter. When,
A second step of setting, as a second parameter, a physical quantity that indicates a distance in the visual axis direction between the measurement plane and one of the two observation points;
When observing from the position p _axis set in the first step, the physical quantity indexed in the visual axis direction set in the second step, and the two observation points of the measurement point each measurement position p _{R of,} p _L or the two measuring positions p _R, replaces the p _L, of the measurement point, the measurement position when observed from one of the observation point of said two observation point p _R And the binocular parallax σ between the two measurement positions p _R and p _L when the measurement points are observed from the two observation points, or the cross ratio {p _axis p _R p _L p _c } or using said plurality ratio equivalent operation, a third step of determining the position p _c of the intersection on the observation plane of the measurement point,
The measurement point, and a fourth step of obtaining a polar corresponding to the measurement point by polar transformation the position p _c of the intersection on the observation plane,
For the plurality of measurement points in the measurement space, the third step and the fourth step of the first to fourth steps are changed to values of the first parameter and the second parameter, respectively. Are repeated several times while changing in the first step and the second step, respectively,
A plurality of polar lines obtained while repeating the first to fourth steps a plurality of times are respectively converted into corresponding polar line drawing spaces among a plurality of polar line drawing spaces according to the value of the first parameter. An intersection between polar lines at the time of drawing is obtained for each polar drawing space, and the true line drawing space corresponding to the true visual axis direction is selected based on information on the number of polar lines intersecting at the intersection. Azimuth n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of polar lines intersecting at the intersection obtained for the polar drawing space corresponding to the true visual axis direction, and / or Alternatively, the program may execute a fifth step of obtaining a physical quantity that indicates a distance in the visual axis direction between the measurement plane, the observation point, and one of the two observation points. ,
In this case, the measurement point appearing on the image has intensity information, and the fourth step obtains the polar line, and the obtained polar line corresponds to the polar line. Voting a value corresponding to the intensity of the measurement point corresponding to the polar line for each point on the trajectory of the polar line when drawn in the line drawing space, and the fifth step is a step of voting the intersection. Instead of obtaining the maximum point where the value obtained by voting during the repetition of the first to fourth steps a plurality of times is determined for each polar drawing space, and based on the information on the maximum value at the maximum point. The true visual axis direction is obtained by selecting the polar drawing space corresponding to the true visual axis direction, and the maximum point obtained for the polar drawing space corresponding to the true visual axis direction is voted Duplicate corresponding to multiple participating polar lines A physical quantity indicating the orientation of the measurement plane including the measurement point and / or the distance in the visual axis direction between the measurement plane and one of the two observation points. Is preferred, or
Measurement points appearing on the image have intensity information,
Sixth step in which the image measurement program sets binocular parallax σ between two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a third parameter. A program having
The third step is between the position _paxis set in the first step and the measurement plane and one of the two observation points set in the second step. A physical quantity indicating the distance in the visual axis direction, a measurement position p _R when observed from one of the two observation points of the measurement point, and the measurement step p _R set in the fifth step by using the binocular parallax sigma, a step of obtaining an intersection p _c on the observation plane,
In the fourth step, a polar line corresponding to the measurement point is obtained, a response intensity corresponding to the binocular parallax σ of the measurement point is obtained, and the obtained polar line is drawn in the polar drawing space. Voting a response intensity corresponding to the binocular parallax σ of a measurement point corresponding to the polar line for each point on the trajectory of the polar line when
For each of a plurality of measurement points in the measurement space, the third step and the fourth step are changed, and the values of the parameters are changed in the first step, the second step, and the sixth step. While repeating several times,
The fifth step executed thereafter is not the determination of the intersection, but the maximum value obtained by voting while repeating the first, second, sixth, third and fourth steps a plurality of times. A point is obtained for each polar drawing space, and the true visual axis direction is obtained by selecting a polar drawing space corresponding to the true visual axis direction based on the local maximum information at the local maximum point, and The orientation of the measurement plane including a plurality of measurement points corresponding to a plurality of polar lines participating in the voting of the maximum points obtained for the polar drawing space corresponding to the true visual axis direction, and / or the measurement plane and the aforementioned It may be a step of obtaining a physical quantity indicating the distance in the visual axis direction between one of the two observation points.

The 18th image measurement program storage medium of the image measurement program storage medium of the present invention is:
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the visual axis direction connecting the two observation points is v, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the visual axis direction v is p. _axis , the position of the intersection with the straight line on the observation plane including one observation point of the two observation points and extending in parallel with the measurement plane including the measurement point, p _c , and the orientation of the measurement plane where n _s
The four positions _{p axis, p R, p L} , cross ratios determined by _{p c {p axis p R p} L p c} or plurality ratio equivalent operations and the visual axis and the orientation _{n s} of the measurement plane by using the inner product of the direction v of the (n s _· v), the azimuth n _s measurement _plane, and / or the shortest between one observation point of the said measurement plane the two observation points An image measurement program for obtaining a physical quantity indicating the distance is stored.

Here, the 18th image measuring program stored image measuring program is executed, the compound ratio _{{p axis p R p L p} c} or plurality ratio equivalent operations, the measurement points, Instead of the two measurement positions p _R and p _L when observed from the two observation points, the measurement position p _R when the measurement point is observed from one of the two observation points and The calculation includes using the binocular parallax σ which is the difference in position between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points.

The image measurement program stored in the eighteenth image measurement program storage medium uses the shortest distance between the measurement plane and one of the two observation points as a physical quantity with the shortest distance as an index. D _s , the distance in the visual axis direction between the measurement plane and one of the two observation points is d _c , and the distance between the two observation points is Δx _LR of,
_n d _s = d _s / Δx _LR
The standardized shortest distance _n d _s expressed by the following formula is adopted, and the standardized shortest distance _n d _{s
n d c = d c / Δx} LR
Using the normalized distance _n d _c expressed by the above and the inner product (n _s · v),
_n d _s = _n d _c (n _s · v)
It may be a program obtained using the relational expression.

The image measurement program stored in the eighteenth image measurement program storage medium is:
A first step of setting a physical quantity indicating the shortest distance as a first parameter;
A second step of setting the inner product ( _ns / v) as a second parameter;
When observing from each of the two observation points of the measurement point, the physical quantity indicating the shortest distance set in the first step, the inner product ( _ns / v) set in the second step each measurement position p _{R of,} p _L or the two measuring positions p _R, replaces the p _L, of the measurement point, the measurement position when observed from one of the observation point of said two observation point p _R And the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points and the position p _{axis of the} infinity point of the measurement point. with _{p axis p R p L p c} } or plurality ratio equivalent operation, a third step of determining the position _{p c} of the intersection on the observation plane,
A fourth step of obtaining the polar corresponding to the position p _c by polar transformation the position p _c of the intersection on the observation plane,
The point on the polar line, and the direction v and angle of the visual axis r = cos ⁻¹ (n _s · v)
And a fifth step for obtaining a point comprising
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are set as the respective values of the first parameter and the second parameter. By repeating a plurality of times while changing in the first step and the second step, the value of the first parameter relating to one measurement point in the fifth step is the same, and the second step A curve connecting a plurality of points obtained by a plurality of executions with different values of the parameter is obtained for each of the plurality of measurement points for each value of the first parameter;
A plurality of curves corresponding to a plurality of curves intersecting at the intersection are obtained by obtaining intersections between the curves when the plurality of curves obtained while repeating the first to fifth steps a plurality of times are drawn in the curve drawing space. A sixth step of obtaining a physical quantity indicating the orientation n _{s of the} measurement plane including the measurement point and / or the shortest distance between the measurement plane and one of the two observation points is executed. It can be a program,
In that case, the measurement point appearing on the image has intensity information, and the fifth step obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to the point, wherein the sixth step repeats the first to fifth steps a plurality of times instead of obtaining the intersection by obtaining the maximum point value by vote becomes the maximum value, azimuth n _s measurement plane including a plurality of measurement points corresponding to a plurality of curves participated in the vote of ultra large _point, and / or, and the measurement plane Preferably, it is a step of obtaining a physical quantity that indicates the shortest distance between one of the two observation points, or
Measurement points appearing on the image have intensity information,
A seventh step in which the image measurement program sets binocular parallax σ between two measurement positions p _R and p _L when the measurement point is observed from the two observation points as a third parameter. A program having
In the third step, the physical quantity indexed in the shortest distance set in the first step, the inner product ( _ns / v) set in the second step, and the measurement point, Using the measurement position p _R when observing from one of the two observation points, the binocular parallax σ set in the seventh step, and the position p _{axis of the} infinity point a step of determining the position p _c of the intersection on the observation plane,
The fifth step obtains the point on the polar line corresponding to the measurement point, obtains the response intensity of the measurement point corresponding to the binocular parallax σ, and measures the point corresponding to the point on the polar line Voting a response intensity of a point corresponding to the binocular disparity σ to a point in the curve drawing space corresponding to the point on the polar line,
Steps 3 to 5 are repeated a plurality of times while changing each value of each parameter in the first, second and seventh steps for a plurality of measurement points in the measurement space, and then executed. Instead of obtaining the intersection, the sixth step is a value obtained by voting while repeating the first, second, seventh and third to fifth steps a plurality of times. By determining the local maximum point, the orientation n _{s of the} measurement plane including a plurality of measurement points corresponding to the plurality of curves that participated in the vote of the local maximum point, and / or the measurement plane and the two observation points It is also a preferred aspect that it is a step of obtaining a physical quantity that indicates the shortest distance between one observation point.

The image measurement program stored in the eighteenth image measurement program storage medium is:
A first step of setting the position p _{axis of the} infinity point by setting the direction v of the visual axis as a first parameter;
A second step of setting a physical quantity indicative of the shortest distance as a second parameter;
A third step of setting the inner product ( _ns / v) as a third parameter;
The position p _{axis of the} infinity point set in the first step, the physical quantity indicating the shortest distance set in the second step, and the inner product set in the third step ( n _s · v) and each of the measurement points p _R and p _L when observed from each of the two observation points of the measurement point, or the measurement points instead of these two measurement positions p _R and p _L , The measurement position p _R when observed from one of the two observation points and the two measurement positions p _R and p _L when observed from the two observation points. and a binocular parallax sigma, a fourth step of using the double ratio _{{p axis p R p L p} c} or plurality ratio equivalent calculation to determine the position _{p c} of the intersection on the observation plane,
A fifth step of obtaining a polar corresponding to the position p _c by polar transformation the position p _c of the intersection on the observation plane,
The point on the polar line, and the direction v and angle of the visual axis r = cos ⁻¹ (n _s · v)
And a sixth step for obtaining a point comprising
For the plurality of measurement points in the measurement space, the fourth to sixth steps of the first to sixth steps, the respective values of the first to third parameters, respectively, By repeating a plurality of times while changing in each step from 1 to 3, the value of the first parameter and the value of the second parameter are the same for one measurement point in the sixth step. Are the same, and a curve connecting a plurality of points obtained by a plurality of executions with different values of the third parameter is represented by each value of the first parameter and each value of the second parameter. For each of the combinations, the plurality of measurement points is obtained, and then
When a plurality of curves obtained while repeating the first to sixth steps a plurality of times are drawn in the corresponding curve drawing spaces among the plurality of curve drawing spaces according to the value of the first parameter. Obtaining the true visual axis direction by selecting the curve drawing space corresponding to the true visual axis direction based on the information on the number of curves intersecting at the intersection. Azimuth n _{s of the} measurement plane including a plurality of measurement points corresponding to a plurality of curves intersecting at the intersection obtained with respect to the curve drawing space corresponding to the true visual axis direction, and / or the measurement plane and the two The program may execute a seventh step for obtaining a physical quantity that indicates the shortest distance between one of the observation points.
In this case, the measurement point appearing on the image has intensity information, and the sixth step obtains the point and sets a value corresponding to the intensity of the measurement point corresponding to the point. Voting for a point in the curve drawing space corresponding to the point in which a curve including the point is drawn, wherein the seventh step is to replace the first to sixth in place of obtaining the intersection A plurality of points corresponding to a plurality of curves that participated in the voting of local maximum points corresponding to a local maximum point corresponding to a local maximum point at which the value obtained by voting during the step is repeated a plurality of times is drawn orientation n _s measurement plane including a measurement _point, and / or is preferably a step of obtaining a physical quantity that indicates the shortest distance between one of the observation points of the with the measuring plane two observation points, Or
Measurement points appearing on the image have intensity information,
An eighth step in which the image measurement program sets, as a fourth parameter, binocular parallax σ between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points; A program having
The fourth step includes the position p _{axis of the} infinity point set in the first step, the physical quantity indicating the shortest distance set in the second step, and the third step. The inner product ( _ns · v) set in step, the measurement position p _R when the measurement point is observed from one of the two observation points, and the eighth step. wherein by using the binocular parallax σ was a step of determining a position p _c of the intersection on the observation plane,
The sixth step obtains the point on the polar line corresponding to the measurement point, obtains the response intensity of the measurement point corresponding to the binocular parallax σ, and measures the point corresponding to the point on the polar line Voting a response intensity of a point corresponding to the binocular disparity σ to a point in the curve drawing space corresponding to the point on the polar line,
For a plurality of measurement points in the measurement space, the steps from the fourth to the sixth are repeated a plurality of times while changing each value of each parameter in the first, second, third and eighth steps, The value obtained by voting during the repetition of the first, second, third, eighth, and fourth to sixth steps a plurality of times, instead of the seventh step executed thereafter obtaining the intersection. Is determined for each curve drawing space by selecting a curve drawing space corresponding to the true visual axis direction based on information on the maximum value at the local maximum point. And azimuth n _{s of} a measurement plane including a plurality of measurement points corresponding to a plurality of curves that participated in the voting of the maximum points obtained for the curve drawing space corresponding to the true visual axis direction, and / or Measurement plane and 2 It is also a preferred embodiment is a step of obtaining a physical quantity that indicates the shortest distance between one point of observation of the observation point.

The 19th image measurement program storage medium of the image measurement program storage medium of the present invention is:
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the visual axis direction connecting the two observation points is v, and the position of the infinity point on the straight line including the measurement points and extending in the same direction as the two visual axes is p. _{When it is axis} ,
Using a single ratio determined by the three positions p _axis , p _R , and p _L (p _axis p _R p _L ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the measurement plane An image measurement program for obtaining a physical quantity that indicates the shortest distance between one of the two observation points is stored.

Here, for the calculation executed by the image measurement program stored in the nineteenth image measurement program, the single ratio (p _axis p _R p _L ) or the equivalent of the single ratio, the two measurement points may be Instead of the two measurement positions p _R and p _L when observed from the observation point, the measurement position p _R when the measurement point is observed from one of the two observation points and the measurement The calculation includes the binocular parallax σ which is the difference in position between the two measurement positions p _R and p _L when the point is observed from the two observation points.

The image measurement program stored in the nineteenth image measurement program storage medium adopts each position projected on a spherical surface as each position p _axis , p _R , p _L and indicates the shortest distance. As a physical quantity, when the shortest distance between the measurement plane and one of the two observation points is d _s and the distance between the two observation points is Δx _LR ,
_n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the normalized shortest distance _n d _s as a parameter;
Using the normalized shortest distance _n d _s set in the first step and a simple ratio (p _axis p _R p _L ) or an operation equivalent to the simple ratio,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A second step for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A third step of obtaining a small circle with a radius R centered on the measurement position of the measurement point when observed from one of the two observation points;
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing the parameters in the first step, and then
A plurality of small circles intersecting at the intersections by obtaining intersections between the small circles when the plurality of small circles obtained while repeating the first to third steps a plurality of times are drawn in the small circle drawing space. the azimuth n _so the measurement plane including a plurality of measurement points _{corresponding,} and / or may be a program for executing the fourth step of determining a normalized shortest distance _n d _s0 for the measurement plane,
In this case, the measurement point appearing on the image has intensity information, and the third step obtains the small circle and draws the obtained small circle in the small circle drawing space. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle, and the fourth step instead of obtaining the intersection, Including a plurality of measurement points corresponding to a plurality of small circles that participated in the voting of the local maximum by obtaining a local maximum at which the value obtained by voting becomes a local maximum while repeating the first to third steps a plurality of times Preferably, this is the step of determining the orientation n _{so of the} measurement plane and / or the normalized shortest distance _n d _s0 for the measurement plane, or
Measurement points appearing on the image have intensity information,
A fifth step in which the image measurement program sets the binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a second parameter; A program comprising
In the second step, the normalized shortest distance _n d _s set in the first step, the position p _{axis of the} infinity point, and one of the two observation points of the measurement point Using the measurement position p _R when observed from the observation point and the binocular parallax σ set in the fifth step, obtaining the radius R;
The third step obtains the small circle corresponding to the measurement point, obtains the response intensity corresponding to the binocular parallax σ of the measurement point, and draws the obtained small circle in the small circle drawing space Voting the response intensity corresponding to the binocular parallax of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing each value of each parameter in the first step and the fifth step,
The fourth step executed thereafter is not the determination of the intersection, but the maximum point at which the value by voting becomes a maximum value while repeating the first, fifth, second and third steps a plurality of times. By obtaining the azimuth n _{so of the} measurement plane including a plurality of measurement points corresponding to the plurality of small circles participating in the voting of the local maximum point, and / or the normalized shortest distance _n d _s0 for the measurement plane is obtained. It is also a preferred embodiment that it is a step.

The image measurement program stored in the nineteenth image measurement program storage medium adopts each position projected on a spherical surface as each of the positions p _axis , p _R , and p _L and sets the shortest distance. as a physical quantity that indicates, the shortest distance d _s between one observation point within the measurement plane and the two observation _points, and the distance between the each other two observation point was set to [Delta] x _LR,
n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the position p _{axis of the} infinity point by setting the visual axis direction v as a first parameter;
A second step of setting the normalized shortest distance _n d _s as a second parameter;
The position p _{axis of the} infinity point set in the first step, the normalized shortest distance _n d _s set in the second step, and a simple ratio (p _axi p _R p _L ) or the Using simple ratio and equivalent operations,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A third step for obtaining a radius R defined by the following relational expression or a relational expression equivalent to the relational expression:
A fourth step of obtaining a small circle having a radius R around the measurement position of the measurement point when observed from one of the two observation points;
For the plurality of measurement points in the measurement space, the third step and the fourth step are performed, and the values of the first parameter and the second parameter are respectively set to the first step and the second step. Repeat multiple times while changing in steps, then
A plurality of small circles obtained while repeating the first to fourth steps a plurality of times are respectively transferred to corresponding small circle drawing spaces among the plurality of small circle drawing spaces according to the value of the first parameter. The intersection between the small circles at the time of drawing is obtained for each small circle drawing space, and the true circle is selected by selecting the small circle drawing space corresponding to the true visual axis direction based on the information on the number of small circles intersecting at the intersection. An orientation n _{so of} a measurement plane that includes a plurality of measurement points corresponding to a plurality of small circles intersecting at the intersection determined for the small circle drawing space corresponding to the true visual axis direction, and / or , A program for executing the fifth step for obtaining the normalized shortest distance _n d _s0 for the measurement plane may be used,
In this case, the measurement point appearing on the image has intensity information, and the fourth step obtains the small circle, and the obtained small circle corresponds to the small circle. Voting a value corresponding to the intensity of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in a circle drawing space, and the fifth step is for locating the intersection. Instead of obtaining, a local maximum point at which the value by voting during the first to fourth steps is repeated a plurality of times is determined for each small circle drawing space, and based on the local maximum information at the local maximum point. The true visual axis direction is obtained by selecting the small circle drawing space corresponding to the true visual axis direction, and the maximum point obtained for the small circle drawing space corresponding to the true visual axis direction is voted The compound corresponding to the participating small circles Orientation n _so the measurement plane including a measurement _point, and / or is preferably a step of obtaining a normalized shortest distance _n d _s0 for the measurement plane, or,
Measurement points appearing on the image have intensity information,
A sixth step in which the image measurement program sets binocular parallax σ between the two measurement positions p _R and p _L when observed from the two observation points of the measurement point as a third parameter; A program comprising
In the second step, the position p _{axis of the} infinity point set in the first step, the normalized shortest distance _n d _s set in the second step, and the measurement point The step of obtaining the radius R using the measurement position p _R when observed from one of the two observation points and the binocular parallax σ set in the fifth step. ,
The fourth step obtains the small circle corresponding to the measurement point, obtains a response intensity corresponding to the binocular parallax σ of the measurement point, and corresponds the obtained small circle to the small circle. Voting the response intensity corresponding to the binocular parallax σ of the measurement point corresponding to the small circle for each point on the trajectory of the small circle when drawn in the small circle drawing space;
For a plurality of measurement points in the measurement space, the third step and the fourth step are repeated a plurality of times while changing each value of each parameter in the first, second and sixth steps, and thereafter Instead of obtaining the intersection, the fifth step to be executed is a local maximum where the value obtained by voting during the first, second, sixth, third and fourth steps is repeated a plurality of times. A point is obtained for each small circle drawing space, and the true visual axis direction is obtained by selecting a small circular drawing space corresponding to the true visual axis direction based on the local maximum information at the local maximum point, and Direction n _{so of} a measurement plane including a plurality of measurement points corresponding to a plurality of small circles that participated in voting for the maximum point obtained for the small circle drawing space corresponding to the true visual axis direction, and / or the measurement plane About standardization shortest It is also a preferred embodiment is a step of obtaining a release _n d _s0.

Further, the twentieth image measurement program storage medium of the image measurement program storage medium of the present invention described above,
Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. When the measurement position is p _R , p _L , the measurement point, and the position of the point at infinity on a straight line extending in the same direction as the visual axis direction v connecting the two observation points is p _axis ,
The three positions _{p axis,} using p R, _{p L,} single-ratio determined by the _{(p axis p R p L)} or single-ratio and equivalent operation, and the measuring point, one of the two observation points An image measurement program for obtaining a physical quantity that indicates the distance to one observation point is stored.

For the calculation performed by the image measurement program recorded in the twentieth image measurement program storage medium, the unit ratio (p _axis p _R p _L ) or the equivalent of the unit ratio, the two measurement points are Instead of the two measurement positions p _R and p _L when observed from each observation point, the measurement position p _R when the measurement point is observed from one of the two observation points; Computation using the binocular parallax σ which is the difference in position between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is included.

The image measurement program stored in the twentieth image measurement program storage medium uses the distance between the measurement point and one of the two observation points as d ₀ as a physical quantity indicating the distance. , when the distance between the each other two observation point was set to [Delta] x _LR,
n d ₀ = d ₀ / Δx _LR
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _axis p _R p _L )
Or a program obtained using a relational expression equivalent to the relational expression.

The twenty-first image measurement program storage medium of the image measurement program storage medium of the present invention is
A measurement plane including an arbitrary measurement point in the measurement space, which appears in an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space, and one of the two observation points the set of between observation point, a physical quantity indicative of the distance of the visual axis direction connecting the two observation points with each other, the coordinates of the voting space defined by the azimuth n _s of the measuring plane as a parameter 1 step,
A measurement position p _R when the measurement point is observed from one of the two observation points, and an infinite point on a straight line including the measurement point and extending in the same direction as the visual axis direction. From the position p _axis and the coordinates in the voting space set in the first step, between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points. A second step of obtaining a binocular parallax σ which is a position difference;
A third step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

The twenty-second image measurement program storage medium of the image measurement program storage medium of the present invention is
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the inside of the measurement space can be viewed from the two observation points. A first step of setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and extending in the same direction as the visual axis direction, which appears in the image of
A physical quantity indicative of the distance of the visual axis direction between one of the observation point of said the measurement plane including a measurement point the two observation points, defined by the azimuth n _s of the measurement plane, the A second step of setting the coordinates in the voting space according to the first parameter as a second parameter;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. Further, from the coordinates in the voting space, a third binocular parallax σ that is a difference between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is obtained. And the steps
A fourth step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The twenty-third image measurement program storage medium of the image measurement program storage medium of the present invention is
One of the two observation points in the measurement space that observes the measurement space, and the measurement space that appears in the image when the measurement space is viewed from the two observation points a physical quantity which indicates the shortest distance between the measurement plane including any measuring point of the inner, a first step of setting as a parameter coordinates voting space defined by the azimuth n _s of the measurement plane,
A measurement position p _R when the measurement point is observed from one of the two observation points, and the visual axis direction that includes the measurement point and connects the two observation points. Two measurement positions when the measurement point is observed from the two observation points, based on the position p _{axis of the} infinity point on the straight line and the coordinates in the voting space set in the first step. a second step for obtaining a binocular parallax σ which is a difference between positions of p _R and p _L ;
A third step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fourth step of voting the response intensity obtained in the third step to the coordinates set in the first step in the voting space,
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

Further, the 24th image measurement program storage medium of the image measurement program storage medium of the present invention is:
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the measurement space is viewed from the two observation points. A first step of setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and extending in the same direction as the visual axis direction, which appears in the image of
And one of the observation point of said two observation points, the physical quantity that indicates the shortest distance to the measurement plane including a measurement point is defined by the azimuth n _s of the measurement plane, said first parameter A second step of setting the corresponding coordinates in the voting space as a second parameter;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. Further, from the coordinates in the voting space, a third binocular parallax σ that is a difference between the two measurement positions p _R and p _L when the measurement point is observed from the two observation points is obtained. And the steps
A fourth step of obtaining a response intensity corresponding to the binocular parallax σ of the measurement point based on two images when the inside of the measurement space is viewed from the two observation points;
A fifth step of voting the response strength obtained in the fourth step to the coordinates set in the second step in a voting space corresponding to the first parameter;
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The 25th image measurement program storage medium of the image measurement program storage medium of the present invention is:
Between two measurement positions p _R and p _{L at} an arbitrary measurement point in the measurement space, which appear in two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space A first step of setting the binocular parallax σ, which is the difference in position, as a parameter;
A measurement position p _R when the measurement point is observed from one of the two observation points, and the visual axis direction that includes the measurement point and connects the two observation points. From the position p _{axis of the} infinity point on the straight line and the binocular parallax σ set in the first step, the measurement plane including the measurement point and one observation point of the two observation points a physical quantity which indicates the, the distance of the visual axis direction between, a second step of obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
A third step of obtaining a response intensity corresponding to the binocular parallax σ set in the first step based on two images when the measurement space is viewed from the two observation points. When,
A fourth step of voting the response strength determined in the third step to the coordinates determined in the second step in the voting space;
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

The 26th image measurement program storage medium of the image measurement program storage medium of the present invention is:
By observing the inside of the predetermined measurement space and setting the visual axis direction v connecting the two predetermined observation points in the measurement space as the first parameter, the measurement space is viewed from the two observation points. A first step of setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and extending in the same direction as the visual axis direction, which appears in the image of
A second step of setting, as a second parameter, binocular parallax σ which is a difference between two measurement positions p _R and p _L when the measurement point is observed from the two observation points;
The measurement point p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. From the binocular parallax σ, the physical quantity indicating the distance in the visual axis direction between the measurement plane including the measurement point and one of the two observation points, and the orientation of the measurement plane is defined by a n _s, a third step of obtaining the coordinates of the voting space corresponding to the first parameter,
Fourth step of obtaining response intensity corresponding to the binocular parallax σ set in the second step based on two images when the measurement space is viewed from the two observation points. When,
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The twenty-seventh image measurement program storage medium of the image measurement program storage medium of the present invention is
Between two measurement positions p _R and p _{L at} an arbitrary measurement point in the measurement space, which appear in two images when the measurement space is viewed from two predetermined observation points in the predetermined measurement space A first step of setting the binocular parallax σ, which is the difference in position, as a parameter;
A measurement position p _R when the measurement point is observed from one of the two observation points and the visual axis direction that includes the measurement point and connects the two observation points. Measurement including one observation point of the two observation points and the measurement point based on the position p _{axis of the} infinity point on the straight line and the binocular parallax σ set in the first step a physical quantity which indicates the shortest distance between the plane and a second step of obtaining the coordinates of the voting space defined by the azimuth n _s of the measurement plane,
A third step of obtaining a response intensity corresponding to the binocular parallax σ set in the first step based on two images when the measurement space is viewed from the two observation points. When,
Voting the response strength determined in the third step to the coordinates determined in the second step in the voting space; and
For a plurality of measurement points in the measurement space, the second to fourth steps among the first to fourth steps are executed a plurality of times while changing parameter values in the first step. An image measurement program is stored.

Further, the twenty-eighth image measurement program storage medium of the image measurement program storage medium of the present invention is:
By observing the inside of the predetermined measurement space and setting the visual axis direction direction v connecting the two predetermined observation points in the measurement space as the first parameter, the two observation points in the measurement space are set. A first step of setting a position p _axis of an infinite point on a straight line including an arbitrary measurement point in the measurement space and appearing in the image when viewed, and extending in the same direction as the visual axis direction;
A second step of setting, as a second parameter, binocular parallax σ which is a difference between two measurement positions p _R and p _L when the measurement point is observed from the two observation points;
The measurement position p _R when the measurement point is observed from one of the two observation points, the position p _axis set in the first step, and the second step are set. Based on the binocular parallax σ, the physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the direction n _{s of the} measurement plane A third step for obtaining coordinates in the voting space according to the first parameter, defined by:
Fourth step of obtaining response intensity corresponding to the binocular parallax σ set in the second step based on two images when the measurement space is viewed from the two observation points. When,
Voting the response intensity obtained in the fourth step to the coordinates obtained in the third step in the voting space corresponding to the first parameter,
For a plurality of measurement points in the measurement space, the third to fifth steps of the first to fifth steps are changed, and each value of each parameter is changed in the first and second steps. However, an image measurement program to be executed a plurality of times is stored.

The 29th image measurement program storage medium of the image measurement program storage medium of the present invention is:
When observing an arbitrary measurement point in the measurement space from the two observation points based on two images when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space A first step of obtaining a response intensity corresponding to binocular parallax, which is a difference between positions of the two measurement positions;
The visual axis direction connecting the two observation points between the measurement plane including the measurement points and one observation point of the two observation points with the response intensity obtained in the first step. A second step of voting to coordinates corresponding to the measurement point and the binocular parallax in a voting space defined by a physical quantity that indicates the distance of the measurement plane and the orientation of the measurement plane;
An image measurement program for executing the first step and the second step a plurality of times for a plurality of measurement points in the measurement space is stored.

  Here, the image measurement program stored in the 29th image measurement program storage medium obtains a maximum point in the voting space where the value by voting becomes a maximum value, thereby voting the maximum point. The orientation of the measurement plane including a plurality of participating measurement points and / or the physical quantity indicating the distance in the visual axis direction between the measurement plane and one of the two observation points is obtained. It may be a program having a third step.

The thirtieth image measurement program storage medium of the image measurement program storage medium of the present invention is:
A first step of setting a visual axis direction connecting two predetermined observation points in the measurement space as a parameter for observing the predetermined measurement space;
Based on two images when the inside of the measurement space is viewed from the two observation points, a position between two measurement positions of the arbitrary measurement point in the measurement space when observed from the two observation points A second step of obtaining a response intensity corresponding to binocular parallax that is a difference between
The response intensity obtained in the second step is a physical quantity indicating the distance in the visual axis direction between the measurement plane including the measurement point and one of the two observation points. A third step of voting on coordinates corresponding to the measurement point and the binocular parallax in a voting space defined by the orientation of the measurement plane and according to the parameter set in the first step And
An image in which the second and third steps of the first to third steps are executed a plurality of times while changing parameter values in the first step for a plurality of measurement points in the measurement space. A measurement program is stored.

  Here, the image measurement program stored in the thirtieth image measurement program storage medium obtains a maximum point at which the value of the vote becomes a maximum value for each voting space, and is based on information on the maximum value at the maximum point. The true visual axis direction is determined by selecting a voting space corresponding to the true visual axis direction relative to the observation point of the measurement point, and the voting space corresponding to the true visual axis direction is determined. The orientation of the measurement plane including a plurality of measurement points that participated in the vote of the given maximum point and / or the visual axis direction between the measurement plane and one of the two observation points A program having a fourth step for obtaining a physical quantity indicating the distance may be used.

The thirty-first image measurement program storage medium of the image measurement program storage medium of the present invention is
When an arbitrary measurement point in the measurement space is observed from the two observation points based on two images when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space. A first step of obtaining a response intensity corresponding to binocular parallax, which is a difference between positions of two measurement positions;
The response intensity obtained in the first step is calculated by using a physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the measurement plane Voting to coordinates corresponding to the measurement point and the binocular parallax σ in a voting space defined by an orientation,
An image measurement program for executing the first step and the second step a plurality of times for a plurality of measurement points in the measurement space is stored.

  Here, the image measurement program stored in the thirty-first image measurement program storage medium participates in the vote of the maximum point by obtaining the maximum point in the voting space where the value by the vote becomes a maximum value. A third step of obtaining a physical quantity indicating the azimuth of the measurement plane including the plurality of measurement points and / or the shortest distance between one of the two observation points and the measurement plane. It may be a program.

Further, the thirty-second image measurement program storage medium of the image measurement program storage medium of the present invention is:
A first step of observing a predetermined measurement space and setting a visual axis direction connecting two predetermined observation points in the measurement space as a parameter;
Based on two images when the inside of the measurement space is viewed from the two observation points, a position between two measurement positions of the arbitrary measurement point in the measurement space when observed from the two observation points A second step of obtaining a response intensity corresponding to binocular parallax that is a difference between
The response intensity obtained in the second step is obtained by using a physical quantity indicating the shortest distance between one of the two observation points and the measurement plane including the measurement point, and the measurement plane. And a third step of voting on coordinates corresponding to the measurement point and the binocular parallax in the voting space according to the parameter set in the first step, defined by the azimuth,
An image in which the second and third steps of the first to third steps are executed a plurality of times while changing parameter values in the first step for a plurality of measurement points in the measurement space. A measurement program is stored.

  Here, the image measurement program stored in the thirty-second image measurement program storage medium obtains a maximum point at which the value obtained by voting becomes a maximum value for each voting space, and is based on information on the maximum value at the maximum point. The true visual axis direction is obtained by selecting the voting space corresponding to the true visual axis direction relative to the observation point of the measurement point, and the voting space corresponding to the true visual axis direction is obtained. A physical quantity indicating the azimuth of a measurement plane including a plurality of measurement points that participated in voting for the maximum point and / or the shortest distance between one of the two observation points and the measurement plane is obtained. It may be a program having a fourth step.

1. Motion vision algorithm (1) Effect of the method for determining the normalized time _n t _{c As} described in 1.3.2, the current time position p ₀ , the next time position p ₁ , and the position after infinite time has elapsed Knowing p _inf , it is possible to determine the three-dimensional orientation n _s of the plane and the normalized time _n t _c until crossing it.

This time is very useful for robots, automatic driving of automobiles, and automatic landing of airplanes. For example, it is performed as follows. An example will be described in which the robot walks in the hallway. When moving diagonally (in the direction of collision) into the corridor, one point in the columnar array in FIG. 10B ignites, and the normalized time _n t _c until “crossing the wall (coinciding with the wall)” as its height coordinate ", And as a coordinate in the cross-sectional circle," wall normal vector _ns "is obtained. When the normalized time _n t _c is multiplied by “a time difference from the present time to the next time (that is, a time difference between image frames) Δt”, the time is converted to a time t _c until the plane is hit. Based on the time t _c and the normal vector n _s , the robot can take an avoidance action as follows. That is, if the lower "limit avoidance time determined by the speed, etc., inertia and driving torque of the robot" measured t _c does not collide with the wall direction (i.e., the normal vector n _s perpendicular measured by the Control to change the direction. When the orientation changes, the next time “time to hit the wall t _c ” returns to a value exceeding the avoidance limit value, and when the avoidance progresses and starts moving parallel to the wall, “t _c is infinite” If you move in this direction, you can grasp that it will not collide. By measuring the time t _c until this manner bump into it and the three-dimensional orientation n _s of the wall, it can be moved without colliding even tortuous corridor.

A major feature of this method is that it is possible to predict the “time t _c until hitting the wall” without knowing the moving speed V just by knowing the positions p ₀ , p ₁ and p _inf at three times. . In the method of measuring and avoiding the distance to the wall with ultrasonic waves or the like, measurement of the moving speed V is indispensable in order to convert the distance into “time until hitting the wall”.

Further, if the method described in 1.6 is used, it is not necessary to know even p _inf (that is, the moving direction v), and only by knowing the positions p ₀ and p ₁ at two times, n _s and _n t _c Can be measured. By this method, for example, even when the moving direction during photographing is not known in the Internet, video, movie image, etc., the “three-dimensional orientation n _{s of the} plane and the normalized time _n t _c until crossing” can be measured. Also, when the plane is moving, the moving direction is generally not known, but even in that case, “the direction and time” can be measured together with the moving direction v.

  There has been no report on how to do this.

(2) Effect of Method for Determining Normalized Shortest Distance _n d _{s As} described in 2.1 and 2.2.3, the current time position p ₀ , the next time position p ₁ , and after an infinite time has elapsed. knows the position _{p inf,} it can determine the 3-dimensional orientation _{n s} and the normalized shortest distance _n d _s plane.

This standardized shortest distance is very useful for “separating multiple objects and environments that appear to overlap each other in an image (called depth separation)”. That is, since the “relative depth” and “inclination” of the plane constituting the object or the environment can be measured as the normalized shortest distance _n d _s and the direction n _s , they can be separated and identified even if they overlap. it can.

This depth separation can be performed only by knowing the positions p ₀ , p ₁ , and p _inf at three times without knowing the moving speed V, “the time difference Δt from the present time to the next time, or the moving distance Δx”. Is a big feature. When the shortest distance d _s to the plane is required, the normalized shortest distance _n d _s can be obtained by multiplying “the moving distance Δx from the present time to the next time”.

Furthermore, if the method described in 2.5 is used, it is not necessary to know even p _inf (that is, the moving direction v), and the normalized shortest distance _n d _{s is obtained} only by knowing the positions p ₀ and p ₁ at two times. to be able to measure the orientation _{n s.} By this method, for example, even when the moving direction during photographing is not known in Internet images, videos, movies, etc., “the distance and direction” can be measured, and the depth separation can be performed based on the distance. Further, when the plane is moving, the moving direction is generally not known, but even in that case, “the distance and the direction” can be measured together with the moving direction v.

There has been no report on how to do this.
2. Binocular vision algorithm (1) Effect of the method for determining the normalized shortest distance _n d _{s As} described in 4.3.1 and 4.3.2, the position p _{L in} the left camera image, position _{p R} and, "the position on the visual axis connecting the left and right cameras _{p axis"} know, can determine the 3-dimensional orientation _{n s} and the normalized shortest distance _n d _s plane.

This normalized shortest distance is very useful for “depth separation of multiple objects and environments that appear to overlap each other in an image”. That is, since the “relative depth” and “inclination” of the plane constituting the object or the environment can be measured as the normalized shortest distance _n d _s and the direction n _s , they can be separated and identified even if they overlap. it can.

A major feature is that this depth separation can be performed without knowing the distance Δx _LR between the left and right cameras, only by knowing the positions p _L and p _R on the left and right cameras and the position p axis on the visual _axis . In addition, when the shortest distance d _s to the plane is necessary, it can be obtained by multiplying the normalized shortest distance _n d _s by “distance Δx _LR between left and right cameras”.

Furthermore, if the method described in 4.3.3 is used, it is not necessary to know even p _axis (that is, the visual axis direction a _xis ), and the standardized shortest distance can be obtained only by knowing the positions p _L and p _R in the left and right cameras. _n d _s and the orientation _{n s} can be measured. According to this method, for example, in the case of a stereo image such as the Internet, even when the visual axis direction during photographing is not known, “the distance and the azimuth” can be measured, and the depth separation can be performed based thereon.

Incidentally, the depth separation above, also be carried out using the the "standardization of the visual axis direction until across the plane distance _{n d c"} and "3-dimensional orientation n _s plane" measured in 4.2.3 it can.

  There has been no report on how to do this.

  Hereinafter, embodiments of the present invention will be described.

  FIG. 2 is a diagram showing an external appearance of a computer system that is employed as an embodiment of the image measurement apparatus of the present invention.

  The computer system 300 includes a main body 301 incorporating a CPU, a RAM memory, a magnetic disk, a communication board, an image input board, etc., a CRT display 302 for displaying a screen in accordance with an instruction from the main body 301, and a user instruction in the computer. And a keyboard 303 for inputting character information, and a mouse 304 for inputting an instruction corresponding to an icon or the like displayed at that position by designating an arbitrary position on the display screen of the CRT display 302. . The main body 301 has a floppy disk loading port 301a and a MO loading port 301b in which a floppy disk and an MO (magneto-optical disk) can be taken out in terms of appearance. Inside the main body 301, a loaded floppy disk is mounted. A floppy disk driver for driving the MO and an MO driver are also incorporated.

  FIG. 3 is a hardware configuration diagram of the computer system shown in FIG.

  This hardware configuration diagram includes a central processing unit (CPU) 311, a RAM 312, a magnetic disk controller 313, a floppy disk driver 314, an MO driver 315, a mouse controller 316, a keyboard controller 317, a display controller 318, a communication board 319, And an image input board 320 are shown interconnected by a bus 310.

  The magnetic disk controller 313 accesses the magnetic disk 321 built in the main body 301 (see FIG. 2).

  Further, as described with reference to FIG. 2, the floppy disk driver 314 and the MO driver 315 are loaded with the floppy disks 322 and MO323, respectively, and access the loaded floppy disks 322 and MO323.

  A mouse controller 316 and a keyboard controller 317 transmit operations of the mouse 304 and the keyboard 303 to the inside of the computer, respectively.

  The display controller 318 is a controller that displays an image on the CRT display 302 in accordance with a program that runs on the CPU 311.

  Further, the communication board 319 is connected to a communication network such as a LAN or the Internet via the communication line 400, and the communication board 319 plays a role of receiving image data flowing through the communication network, for example. .

  Further, the image input board is connected to an external camera 11 (such as an electronic still camera or a video camera), and plays a role of capturing image data obtained by photographing with the camera 11 into the computer. . Although only one camera is shown in FIG. 3, two cameras are connected to the image input board 320, and the same camera is viewed from different directions corresponding to the parallax of human eyes, for example. It is also possible to input two images obtained by photographing the subject at the same time.

  The computer system 300 is installed with a program stored in the floppy disk 322 or the MO 323 or installed via the communication line 400 and operates as an image measuring apparatus according to the present invention described later. Therefore, an embodiment of the image measurement apparatus of the present invention is realized as a combination of the hardware of the computer system shown in FIGS. 2 and 3 and a program installed and executed in the computer system. The program for operating this computer system as the image measuring apparatus of the present invention corresponds to the image measuring program according to the present invention, and when the image measuring program is stored in the floppy disk 322 or the MO 323, the image measuring program is stored. The floppy disk 322 and the MO 323 in which the program is stored correspond to the image measurement program storage medium referred to in the present invention. When the image measurement program is installed in the computer system, the installed image measurement program is stored in the magnetic disk 321. Therefore, the magnetic disk 321 that stores the image measurement program also corresponds to an example of the image measurement program storage medium of the present invention.

  Here, the function of the computer system 300 shown in FIGS. 2 and 3 as an image measuring device is shown in various functional block diagrams to be described later. It is shown in the flowchart. Also, the various flowcharts described later can be regarded as “methods”, and therefore the various flowcharts described later correspond to various embodiments of the image measurement method of the present invention.

  Here, the description of the embodiment of the present invention is temporarily interrupted, the principle of image measurement according to the present invention will be described, and then various embodiments of the present invention will be described.

1. Method of measuring plane 3D orientation and time until crossing Of 3D geometric information of plane, a method of measuring plane 3D orientation n _s and time t _c until crossing the plane is provided.

It is assumed that the plane is moving in a certain direction v. Plane of the normal vector n _s (3-dimensional orientation of n _s plane), the state moves from the current time t ₀ to the next time t _1, which plane further time t _c crosses camera center O, As shown in FIG. The vertex (white circle) of the triangle on the plane of each time is the intersection (black circle) of the line connecting the camera center and each vertex and the eyeball (or spherical camera) on the retina (or spherical camera) of the eyeball. It is reflected. Hereinafter, for the sake of simplicity, the diameter of the eyeball (or spherical camera) is assumed to be 1. Therefore, the vector connecting the camera center and the black circle is a “unit vector” of size 1.

1.1. Shown in FIG. 5 by extracting state at time t _c of the principles Figure 4 for measuring the three-dimensional orientation n _s plane. At this time, since the plane passes through the camera center O, the point group on the plane (white circle) is projected as a “point group on the great circle g _ns (black circle)” on the spherical surface. This great circle is the intersection line between the plane and spherical, therefore the vector p _c perpendicular to the normal vector n _s plane. From this relationship, the normal vector n _s of the plane can be measured in the following manner as a "polar transformation" of the vector p _c. That is, a great circle centered on it for each vector p _c - great circle the largest circle on the sphere - the draw, they great circle group intersect at one point, the normal vector of the plane as the intersection ( That is, the three-dimensional orientation) _ns is measured. Thus, when it is found p _c for a plurality of points on a plane, it is possible to obtain a three-dimensional orientation of the plane by performing a polar transformation. (Here, previously described a polar transformation terms, p _c and great circle points on the spherical surface are respectively referred to as "electrode" and "polar" polar (great circle that pole p _c ) Is referred to as “polar transformation” (or dual transformation).)
1.2. Principle of Measuring Normalized Time _n t _c until Crossing the Plane The principle of measuring the normalized time _n t _c will be described. Here, the normalization time is a time obtained by normalizing the time t _c until crossing the plane by “a time difference Δt between the current time t ₀ and the next time t ₁ ”, and is represented by Expression (1).

_n t _c = t _c / Δt (1)
Δt = t ₁ −t ₀ (2)
FIG. 6 shows how the triangle (black circle) projected on the spherical surface in FIG. 4 moves. Triangle ₁ p ₀ at the present time _{t 0, 2 p 0, 3} p 0 is moved to the next time _{t 1} in _{1 p 1, 2 p 1,} 3 p 1, then at time _{t c} across the plane "plane degenerated three-point _{1 p c, 2 p c,} 3 p c on the great circle _{g ns} "that is perpendicular to the normal vector _{n s} move in, after last infinite time to converge to the moving direction v . Each vertex moves on “movement direction v and great circles ₁ g, ₂ g, ₃ g connecting them”. In addition, since the moving direction v is a position after an infinite time has elapsed, it is also referred to as p _inf hereinafter.

In preparation for measuring the normalized time _n t _c, the position p ₀ at the present _time, the position p ₁ at the next _time, and to know the position p _inf after infinite time has elapsed, the three-dimensional distance of a point at the present time FIG. 7 shows that d ₀ can be measured (ie, the distance from the camera center O to the point P ₀ ). FIG. 7 shows a cross section of the spherical surface cut by “one of the great circles g representing the movement locus” in FIG. The use of sine theorem triangle _OP 0 _{P 1,} between the distance _{d 0} to the point _{P 0} "moving distance [Delta] x from the current to the next _{time" Δx / sin (p 0 p} 1) = d 0 / sin (P _inf p ₁ ) (3)
There is a relationship. Here, p ₀ p ₁ is a central angle from p ₀ to p ₁ , and p _inf p ₁ is a central angle from p _inf to p ₁ . By transforming equation (3), the distance d ₀ to the point P ₀ is calculated by the following equation.

d ₀ = Δx sin (p _inf p ₁ ) / sin (p ₀ p ₁ ) (4)
Here, the “single ratio (abc)” for the three points a, b, and c of the great circle is (abc) = ac / bc = sin (ac) / sin (bc) (5)
(Page 6 of “Projective Geometry (Gourievic, Tokyo Book)”). When this single ratio is used, Expression (4) is expressed by Expression (6). Therefore no longer depend on type of central projection by simple ratio expression, "spherical camera or eye" above three points p ₀ to move the large circle _of, p well _1, p _c, "the linear of the planar camera image" it is possible to measure the distance _{d 0} from any three points _{p 0, p 1, p c} to move over. That is, regardless of the camera system for capturing an image, can measure the distance d ₀ of the point three-dimensional.

d ₀ = Δx (p _inf p ₀ p ₁ ) (6)
Based on the above preparation, the principle of measuring the normalized time _n t _c will be described. The "plane crosses the camera center O at time t _c" that was additionally written in FIG. 7 is a diagram 8. Assuming that the moving speed is V, “the moving distance Δx from the present time to the next time” is “the time difference Δt from the present time to the next time” and Δx = V Δt (7)
Therefore, substituting equation (7) into equations (4) and (6) gives the following equation.

d ₀ = V Δt sin (p _inf p ₁ ) / sin (p ₀ p ₁ ) (8a)
= V Δt (p _inf p ₀ p ₁ ) (8b)
Also, assuming that the time from the present to crossing the plane is t _c , when the sine theorem is applied to the triangle O ₀ PP _c and deformed in the same manner as described above, the distance d ₀ is obtained by the following equation.

_{d 0 = V t c sin (} p inf p c) / sin (p 0 p c) (9a)
_{= V t c (p inf p} 0 p c) (9b)
If the ratio of the formula (8) and the formula (9) is taken and arranged, the normalized time _n t _c until the plane is crossed is obtained by the following formula.

_{n t c} = t _c / _{Δt
= (Sin (p inf p 1} ) / sin (p 0 p 1)) / (sin (p inf p c)
/ Sin (p _{0 pc} )) (10a)
_{= (P inf p 0 p 1} ) / (p inf p 0 p c) (10b)
Here, the double ratio {abcd} for the four points a, b, c, d on the great circle is defined by the equation (11a) as “a ratio of two simple ratios, (abc) and (abd)”. The cross-ratio has the relationship of formula (11c) (pages 257 and 119 of “Projective geometry (Gourievic, Tokyo Book)”).

{Abcd} = (abc) / (abd)
= (Ac / bc) / (ad / bd) (11a)
= (Sin (ac) / sin (bc))
/ (Sin (ad) / sin (bd)) (11b)
{Abcd} = {badc} = {cdab} = {dcba} (11c)
Using this definition of the cross ratio (formula (11a)), formula (10) becomes
_{n t c = {p inf p} 0 p 1 p c} (12a)
It is expressed. As described above, when the four points p ₀ , p ₁ , p _c , and p _inf on the movement trajectory are known, the normalized time _n t _c is obtained as a cross ratio of Expression (12a).

Here, the projective geometric meaning of equation (12a) is considered. According to page 86 of “Projective geometry (by Yasunaga and Hirano, Asakura Shoten)”, the coordinate λ of d based on the base point system a, b, c is called the cross ratio, and is expressed as {abcd}. (The same description is given on page 119 of “Projective Geometry (Gourievic, Tokyo Book)”). In this definition, if the base point system a, b, c is changed to the base point system p _inf , p ₀ , p ₁ , and the cross ratio value λ is changed to _n t _c , “the base point system p _inf , p ₀ , p ₁ the coordinates _n t _c of p _c called Fukuhi _will represented "in _{{p inf p 0 p 1 p} c}. Therefore, equation (12a) is
“The normalized time _n t _c is the coordinates of _pc measured by the base point system with the origin, the infinity point, and the unit point as p ₀ , p _inf , and p ₁ ” (12b)
Means.

The cross-ratio in equation (12a) is a fundamental invariant of projective geometry and does not change for any projection and cut--that is, to “an arbitrary camera-style image” such as a spherical camera or planar camera. On the other hand, the cross ratio does not change. Thus, "spherical camera or eye" above four points _p 0 to move the large circle _{of, p 1, p c, p} inf well, four-point moves on the "straight line of the planar camera image" _{p 0} , P ₁ , p _c , and p _inf , the “normalized time _n t _c until crossing the plane” can be measured as a cross ratio. That is, it shows that the standardized time _n t _c can be measured regardless of the method of the camera that captures the image.

1.3. Four positions used in the above principle method of obtaining the three-dimensional geometric data of the plane by converting the polar transformation cross ratio _{p 0, p 1, p c} , consider whether it is possible to know the p _inf. First, the position p ₁ of the position p ₀ and the next time the present time can be known from the camera image. Next, the position at infinite time can be known because it is equal to the moving direction v of the plane (or camera). Can not be known directly of the four positions described above is a position p _c at time t _c which plane crosses the camera center.

By "cross-ratio conversion" obtained by modifying Equation (10a) or Formula (12a) can estimate its position p _c, then by "polar transformation" in the process of the p _c 1.1, plane It is shown below that the three-dimensional geometric information (normalized time _n t _c0 until crossing with the three-dimensional orientation n _s0 ) can be measured.

1.3.1 Multi-ratio transform This multi-ratio transform is important for determining three-dimensional geometric information by knowing the normalized time _n t _c and the positions p ₀ , p ₁ , and p _inf at three times. The position p _c ″ is calculated. Four variable _n t _c in the formula (12a), since _{p 0, p 1, p inf} is known, it is possible to determine the remaining variables _{p c} - This calculation may as double ratio calculation method of projective geometry Are known.

This calculation is specifically expressed by a mathematical formula. 9 is taken out and shown in FIG. and a p _inf base point indicates the position of _{p 0, p 1, p c} in the central angle a, b, x therefrom (Note: This base may be any position). Various central angles are summarized below.

p _inf p ₀ = a (13)
p _inf p ₁ = b
p _inf p _c = x
p ₀ p ₁ = ba
p _{0 pc} = x−a
Using these central angles, the above-described cross ratio conversion is expressed by a mathematical expression. When the right side of the equation (10a), that is, the cross ratio is expressed using the central angle of the equation (13),
_n t _c = (sin (b) / sin (ba))
/ (Sin (x) / sin (x−a)) (14a)
It becomes. By transforming this, the central angle x between _{p c} and _{p inf} is ^{x = tan -1 ((sin (} a) sin (b))
_{/ (Cos (a) sin (} b) - n t c sin (b-a)))
(14b)
Given in. Therefore, given the normalized time _n t _c and “three time positions p ₀ , p ₁ , p _inf ”, the position p _c at the time crossing the plane is calculated by the equation (14b), The formula for ratio conversion is shown.

Here, in the study of the general moving image processing and optical flow instead of used in the above "position p ₁ at the next _time", "change in p ₁ -p 0 from the current time _(i.e., in motion parallax τ In most cases, the center angle p ₀ p ₁ ) ”is handled. Each arrow in the optical flow pattern (FIG. 1) is this change, and the start point of the arrow is the current time position p ₀ and the end point is the next time position p ₁ . In FIG. 9, the amount of change is represented by an angle τ. The formula for the cross ratio conversion in this case is shown below. Various central angles are p _inf p ₀ = a (15)
p ₀ p ₁ = τ
p _inf p _c = x
p _inf p ₁ = a + τ
p _{0 pc} = x−a
And the right side of equation (10a) is expressed using the central angle of equation (15)
_n t _c = (sin (a + τ) / sin (τ))
/ (Sin (x) / sin (x−a)) (16a)
It becomes. By transforming this, _{p c} and _p central angle x is x ⁼ tan with _{inf -1 ((sin (a)} sin (a + τ))
/ (Cos (a) sin ( a + τ) - n t c sin (τ))) (16b)
And another formula for cross ratio conversion was obtained.

1.3.2 using 3-dimensional orientation as the cross normalized time Determination above cross ratio conversion to the plane, a method of determining a normalized time _n t _c to traverse the three-dimensional orientation n _s plane Will be explained. This is done in four steps:

(1) The normalized time parameter _n t _c is arbitrarily set.

(2) For each point in the image, the positions p ₀ and p ₁ at the current time and the next time are known from the camera image, the position p _inf after elapse of infinite time is known from the moving direction v, and they are expressed by the formula ( 14b) or by performing a multi-ratio conversion into equation (16b), calculates the position _{p c.}

(3) 1.1 based on "the principle of measuring the three-dimensional orientation _{n s} plane", determining a candidate of the normal vector _{n s} plane. That is, polar transformation to p _c obtained in step (2), draw a great circle on a sphere. Here, the meaning of drawing a great circle will be described. If the normalized time parameter _n t _c given in step (1) is the true normalized time _n t _c0 , as described in FIG. 5, the plane normal vector n _s0 is used as the intersection of these great _circles. I want. However, in step (1), since the parameter _n t _c is arbitrarily given, generally it does not intersect at one point. Therefore, great circle drawn here would determine the candidate of the normal vector n _s plane. The intensity of the great circle is set to “brightness at position p ₀ in the image”, and at a place where a plurality of great circles intersect, the intensity of the great circles is added.

(4) The above steps are performed while changing the normalized time parameter _n t _c, and a parameter value _n t _{c0 at} which a plurality of great circles drawn in step (3) intersect at one point is obtained. As the parameter value, “normalized time _n t _c0 until crossing the plane” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

Here, a description of the geometrical meaning of the position p _c calculated by the cross-ratio conversion step (2). Its p _c is the position at any time _n t _c Delta] t obtained by "prediction". This prediction is clear from the derivation process of the equation (10a) that is the base of the cross ratio conversion, but intuitively, the equation (12b) can be rephrased as “arbitrary time _n t _c Δt (ie, normalized time _n t _c Situated in predicting of), the origin, the point of infinity, _p 0 units points _{respectively, p} inf, in origin system to _{p 1,} coordinates may be obtained a position _{p c} of the _n t _c "
You can understand that. Time p _c predicted this manner are arranged on the great circle is the "time _n t _c0 Delta] t across the plane center of the camera (i.e., t _c0)" shown in FIG. 5, in response to the time “Normalized time _n t _c0 until crossing the plane” is determined. Great circle obtained by converting these p _c poles meet at one point, the three-dimensional orientation n _s0 plane are determined as the coordinates of the intersection (Fig. 5).

  1.3.3 Geometric meaning of the above steps

The geometric meaning of each step described above will be described with reference to FIG. As shown in FIG. 10A, each point on the spherical camera image moves from the current time position p ₀ to the next time position p ₁ , and then the “plane crosses the camera center. "through the position p _c in the last and after infinite time has elapsed" time to reach the plane (or camera) position equivalent p _inf in the moving direction v of _"(see FIG. 6).

Determination of the position _{p c} (meaning of Step (2)): "Present and position _p 0 of the next time, _{p 1"} and was given in step (1) "normalized time parameter _n t _c" Equation (14b) converts cross ratio by, determining the position p _c for "time plane crosses camera center". This state is shown in FIG. When the cross ratio conversion of the equation (16b) is used, “the difference vector τ from the current time position p ₀ to the next time position p ₁ ” is used instead of the position p ₁ of the next time.

Drawing candidates plane orientation {n _s} (meaning of Step (3)): to convert the position p _c determined above polar great circle on the sphere, i.e. candidates plane orientation {n _s} Is drawn as shown in FIG. If the normalized time parameter _n t _c given in step (1) is the true normalized time _n t _c0 , the normal vector n _{s0 of the} plane as the intersection of these great circles corresponding to a plurality of points in the image. Is obtained.

3D geometric information is determined as the coordinate value of the cylindrical array (meaning of step (4)): the spherical surface in FIG. 10A is projected onto a plane, and the image on the spherical surface is converted into a “circle”. --There are known three-dimensional projection method, equidistant projection method, orthographic projection method, etc. ("Latest lens design course 23 Problems associated with lens design (1) (Photo by Nakagawa, Photography) "Industry, 1982": Section 4.2.2.1 of the "Research Results Report on Working Robot Development Contracted for Working Nuclear Power Generation Facilities in 1984 (Extreme Working Robot Technology Research Association)"; Section 4.2.2.1 of “Work Robot Development Contract Research Results Report (Extreme Work Robot Technology Research Association)”. The circles are stacked with the normalized time parameter _n t _c as the vertical axis to form a “cylindrical array” as shown in FIG. In this way, the geometric meaning of step (1) becomes clear. That is, the normalized time parameter _n t _c arbitrarily given in step (1) specifies the height coordinate of this cylinder, and in steps (2) and (3), the cross-sectional circle at that height -Drawing the spherical image of Fig. 10 (A) into a "circle". Since the parameter _n t _c is arbitrarily given in step (1), the great circles do not intersect at one point as shown in the upper surface of FIG. 10B, but the height is _{equal to the} true normalization time _n t _c0 . In equal cross-sectional circles, the great circles intersect at one point. Accordingly, the normalized time of the plane as "height coordinate" _n t _c0 cylindrical can be obtained and the three-dimensional orientation n _s as "cross coordinates in the section circle" is obtained (FIG. 10 (B)).

1.4 Confirmation by simulation It is shown by computer simulation that the “algorithm for measuring three-dimensional geometric information of a plane” described in 1.3.2 and 1.3.3 is correct (FIG. 11). The simulation was performed according to the flow of Embodiment A-1.

First, input data will be described. There is a plane perpendicular to the front of the spherical camera (or eyeball), and the distance to the camera center is 3 m. The plane moves in a direction perpendicular to it (ie, in a direction parallel to the normal vector n _{s0 of the} plane) toward the camera at a speed of 1 m / sec. There are eight points on the plane, and “the current and next time positions p ₀ , p ₁ ” of these points are observed as input image data. The time difference Δt from the present time to the next time is 0.05 seconds, and the position p _{inf at the} infinite time is equal to the moving direction v of the camera and at the center of the visual field. From the above, the time t _c until the camera crosses the plane is 3/1 = 3 seconds, and thus the normalized time _n t _c0 is 3 / 0.05 = 60. The plane normal vector _ns0 is at the center of the visual field.

From the current position _p 0 of the next time, _{p 1} and the position _{p inf} after infinite time has elapsed, by the algorithm described above (cross-ratio conversion and polar transformation), the three-dimensional geometric data of the plane _{(n t c0} and _{n s0)} FIG. 11 shows a simulation result obtained for the above. “Cylinder cross-sectional circles” at the normalized time parameters _n t _c described in 1.3.3 are shown side by side. Each cross-sectional circle is projected on the plane passing through the sphere by the “equal distance projection method described in 1.3.3 (see equation (103c) in the embodiment)” in FIG. 10A. The lower right is a cross-sectional circle at _n t _c = 0 corresponding to the current time, and is arranged in order of increasing parameters _n t _c toward _n t _c = infinite corresponding to the upper left infinite time. Each cross-sectional circle will be described. Each circular section, the position p _0, p ₁ and the parameter _n t _c position p _c calculated by the "cross-ratio conversion" from is depicted by dots, 8 corresponding to the eight points on a plane _{p c} is depicted in - these _{p c,} as described in 1.3.2, is a position overlooking the respective points at an arbitrary time _n t _c Delta] t obtained by "prediction". Then, they p _c eight great circles that "polar transformation" is depicted.

In the first cross-sectional circle (bottom right, _n t _c = 0), these great circles are scattered, but converge as the parameter _n t _c increases, and the cross-sectional circle with _n t _c of 60 (upper right end) In the second), these great circles intersect at one point. When _n t _c becomes larger than this, it diverges again. Thus, the great circles intersect at one point only at the height of _n t _c = 60. This height _n t _c is equal to the value 60 of the above-mentioned “normalized time _n t _c0 until crossing the plane”. The azimuth intersecting one point is at the center of the field of view and is equal to the above-described “plane normal vector n _s0 ”. From the above simulation, it was confirmed that the “algorithm for measuring three-dimensional geometric information of a plane” described in 1.3.2 and 1.3.3 was correct.

1.5 normalized distance _n d _c measures how normalized distance _n d _c up across the plane in the moving direction, "the distance _{d c} (this distance to crossing the plane direction of movement Vt _c of FIG. 8 Is standardized by “movement distance Δx from the present time to the next time”, and is expressed by the following equation.

_n d _c = d _c / Δx
= Vt _c / Δx (17a)
Substituting equation (7) and transforming using equation (1)
_{n d c = Vt c / (} VΔt) = t c / Δt
= _N t _c (17b)
Thus, it was shown that the normalized distance _n d _c is equal to the aforementioned “normalized time _n t _c ”. Therefore, measuring the normalized distance _n d _c used as it is a method (1.3.2) to obtain the normalized time _n t _c.

1.6 In the moving direction v or good way without having to know, as the movement direction v is known, describes a method for measuring the normalized time _n t _c to the three-dimensional orientation n _s plane thereacross It was. That is, determine the "position of the infinite time p _inf" from the direction v, then performs a cross-ratio conversion using the position, the end performing polar transformation three-dimensional orientation n _s and the normalized time _n t _c Asked.

Here, the moving direction v is to provide a method which can measure the "three-dimensional orientation n _s and the normalized time _{n t} c _plane" in unknown. By this method, for example, even when the moving direction during shooting is not known in the Internet, video, movie image, etc., “the direction and time” can be measured. Further, when the plane is moving, the moving direction is generally not known, but even in that case, “the direction and time” can be measured together with the moving direction v. The outline of the method will be described. It is assumed that the moving direction v can be in all directions, and “3.2. When calculating the moving direction thereof poles lines intersect at a certain, but it is true of the movement direction v _0, also the three-dimensional orientation n _s and the normalized time _n t _c of the plane is determined as the coordinates of the intersection is there. It is done in the following steps.

  (1) The moving direction parameter v is arbitrarily set.

(2) The direction of the parameter v is set to “infinite time position p _inf ”.

(3) Steps (1) to (4) in 1.3.2 are executed, and polar lines for all the normalized time parameters _n t _c are drawn inside the cylindrical array (FIG. 10B).

(4) The above steps are performed by changing the moving direction parameter v, and a parameter value v _{0 at} which the polar lines drawn in the above steps intersect at one point is obtained. This parameter value is true in the moving direction v _0. As the coordinates of the intersection, the plane direction n _s0 and the normalized time _n t _c0 until crossing the plane are obtained. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

2. A method of measuring the normalized shortest distance _n d _s to 3-dimensional orientation n _s and the plane of the normalized shortest distance measurement method plane to plane. Here, the normalized shortest distance is a distance obtained by normalizing the shortest distance d _s to the plane by “distance Δx that the camera (or plane) has moved from the present time to the next time”, and is expressed by Expression (19). The

_n d _s = d _s / Δx (19)
Between this normalized shortest distance _n d _s and “standardized time _n t _c ” described in “1”, _ns is a three-dimensional orientation of a plane, v is a moving direction, and () is a scalar product.
_{n d s = n t c (} n s · v) (20)
There is a relationship. The reason will be described with reference to FIG. 12—a cross section in a plane in which the plane vector n _s and the moving direction v lie. Since “the shortest distance d _s from the camera center O to the plane” is a normal direction component of “the distance Vt _c to cross the plane in the moving direction”,
d _s = Vt _c (n _s · v) (21)
-V is the speed of movement. When both sides of this equation are normalized by the movement distance Δx, the following equation is obtained and equation (20) is obtained.

_n d _s = d _s / Δx
_{= (Vt c / Δx) (} n s · v)
_{= (Vt c / (VΔt)} ) (n s · v)
_{= N t c (n s ·} v) (22)
In the above modification, the following relationship between “the time difference Δt from the present time to the next time” and “the distance Δx moving between them” was used.

Δx = VΔt (23)
2.1 the plane normalized shortest distance _n d _s and algorithms described in method 1.3.2 of measuring the three-dimensional orientation _{n s} of (cross-ratio conversion and polar transformation) _"in the n _{d s} and _n t _c combining relationships (equation (20)) ", can be measured three-dimensional orientation _{n s} and the normalized shortest distance _n d _s plane.

  This will be described with reference to FIG. It is done in the following six steps.

(1) The normalized shortest distance parameter _n d _s is arbitrarily set.

(2) Considering a small circle centered on the moving direction v, the radius r is arbitrarily set (FIG. 13). In step (4), the “plane three-dimensional orientation candidate n _s ” on the small circle is obtained. For this purpose, it is necessary to set the normalized time parameter _n t _c to a value determined by the following equation.

_n t _c = _n d _s / cos (r) (24)
The reason will be explained. First, since n _s is above “a small circle with radius r centered on v”, there is a relationship between n _s , v, and r:
cos (r) = (n _s · v) (25a)
Also, since n _s needs to satisfy equation (20), substituting equation (25a) into equation (20)
_{n d s = n t c cos} (r) (25b)
This is transformed to obtain equation (24).

(3) For each point in the image, the positions p ₀ and p ₁ at the current time and the next time are known from the camera image, and the position p _inf after the infinite time has elapsed is known from the moving direction v and normalized with them. The position parameter p _c is calculated by substituting the time parameter _n t _c into the equation (14b) or the equation (16b) and performing a cross ratio conversion.

(4) Step (3) to polar transformation to _{p c} determined, draw a great circle _{g pc} on a spherical surface. Two intersections _r n _{s +} and _r n _{s− of the} great circle and the small circle in step (2) are “planar three-dimensional orientation candidates” (FIG. 13). In addition, it mentions later that this intersection is represented by Formula (29).

(5) Step (2) to perform (4) by changing the radius r, the two intersection points _r n _s + which is obtained in step _(4), a curve consisting of r _{n s-(13)} - This curve As described in 2.2, it becomes “a small circle centered on p ₀ ”. Here, the meaning of drawing this curve will be described. If the normalized shortest distance parameter _n d _s given in step (1) is the true normalized shortest distance _n d _s0 , a plane normal vector n _s0 is obtained as the intersection of these curves. However, in step (1), since the parameter _n d _s is arbitrarily given, generally it does not intersect at one point. Thus, the curve drawn here would determine the candidate of the normal vector n _s plane. It should be noted that the intensity of the curve is “brightness at position p ₀ in the image”, and the intensity obtained by adding the intensities of the curves at a place where a plurality of curves intersect.

(6) The above steps are performed by changing the normalized shortest distance parameter _n d _s, and a parameter value _n d _{s0 at} which the plurality of curves drawn in step (5) intersect at one point is obtained. As the parameter value, “standardized shortest distance _n d _s0 from the camera center O to the plane” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

2.2 Another method for measuring the standardized shortest distance _n d _s and the three-dimensional orientation n _s of the plane First, it is shown that the curve drawn in 2.1 (5) is a small circle, and then the small A method for measuring the standardized shortest distance _n d _s0 and the three-dimensional azimuth n _s0 of the plane using circle drawing will be described.

2.2.1 Proof of becoming a small circle FIG. 14 shows various parameters written in FIG. 13 for proof. Using these parameters, steps (3) to (5) of 2.1 are expressed by the following mathematical formulas, indicating that the curve of 2.1 (5) is a small circle.

First, the “position p _c ” described in step (3) is expressed by a mathematical expression. That, _{p c} and _{p inf} and central angle x (Fig. 9) of is represented by the following equation by substituting equation (24) into equation (16b). In addition, although Formula (16b) is used here, even if it uses Formula (14b) equivalent to it, it can prove similarly.

tax (x) = (sin (a) sin (a + τ))
/ (Cos (a) sin (a + τ)
_{- (n d s / cos (} r)) sin (τ)) (26)
Next, expressed intersection between the small circle "of Step (2) and" great circle that polar converts the _{p c} ", i.e. three-dimensional orientation candidate _r n _s + _plane, the r _{n s-a} formula. Triangle _r When you apply the cosine theorem to _{n s + p c v cos (} π / 2) = cos (r) cos (-x)
+ Sin (r) sin (-x ) cos (π- (r α s + -α a))
= Cos (r) cos (x)
+ Sin (r) sin (x ) cos (r α s + -α a) (27)
become. [Pi / 2 in the above equation due to the fact that the orientation candidate point _r n _{s +} is on "polar of _{p c".} In equation (27) by substituting the equation (26), clearing the _{x cos (r α s + -α} a) = - cot (r) cot (x)
= -Cot (r) (cos (a) sin (a + τ)
_{- (n d s / cos (} r)) sin (τ))
/ (Sin (a) sin (a + τ))
=-(Cos (r) cos (a) sin (a + τ)
_{- n d s sin (τ)} )
/ (Sin (r) sin (a) sin (a + τ))
To become, when further conversion cos (r) cos (a) + sin (r) sin (a) cos (r α s + -α a)
_{= N d s sin (τ)} / sin (a + τ) (28)
become. The longitude coordinate points _r α _{s +} and _r α _s− of the candidate points _r n _{s +} and _r n _{s− in} the plane orientation are obtained by transforming the equation (28).
_{r α s + = α a +} (r α s + -α a)
= Α _a + cos ⁻¹ ((( _n d _s sin (τ) / sin (a + τ))
-Cos (r) cos (a)) / (sin (r) sin (a)))
_{r α s- = α a - (} r α s + -α a)
^{_{= Α a -cos -1 (((}} n d s sin (τ) / sin (a + τ))
-Cos (r) cos (a)) / (sin (r) sin (a)))
(29)
Calculated by Accordingly, the longitude coordinates r of the plane orientation candidate points _r n _{s +} and _r n _s− are set as “the radius r of the small circle in step (2)”, and the longitude coordinate points _r α _{s +} and _r α _s− are expressed by the equation (29). ).

Based on the above preparation, it is shown that “a curve formed by two intersections _r n _{s +} and _r n _s− ” is a small circle centered on p ₀ . A small circle of radius R centered at p ₀ applies the cosine theorem to the triangle _r n _{s +} p ₀ v, and cos (r) cos (a)
+ Sin (r) sin (a) cos ( _{r [} alpha] _{s +-[} alpha] _a ) = cos (R) (30)
(Page 72 of “Geometrical Dictionary 2 (Ishida Ishiyasu, Tsuji Shoten)”). Comparing equation (30) and equation (28), equation (28) _{—that is} , a curve composed of two intersections _r n _{s +} and _r n _s −—is “p ₀ where radius R is expressed by the following equation It can be seen that it is a small circle centered on.

^{_{R = cos -1 (n d s}} sin (τ) / sin (a + τ)) (31)
Thus, + two intersections _r n _s in FIG. _14, the curve consisting of r _{n s-(i.e.} formula (28)) has been shown to be a _{"p 0} radius small circle R around the". This can be regarded as converting an arbitrary point p ₀ into this small circle, and this conversion will be referred to as “small circle conversion”. Here, the position _{p 0} in the current time equation (31), previously represented by the position _{p inf} position _{p 1} and infinite time, the next time. When τ and a + τ are expressed by the central angle of Expression (15), Expression (31) is expressed by the following expression.

^{_{R = cos -1 (n d s}} sin (p 0 p 1) / sin (p inf p 1)) (32)
While representing the radius R using the equation (31), the motion parallax tau, if it is found position p ₁ of the next time is given by the following equation by substituting equation (13) into equation (32).

^{_{R = cos -1 (n d s}} sin (b-a) / sin (b)) (33)
2.2.2. Geometric meaning of small circle transformation

  (1) Geometric meaning of radius R

Wherein when the (32) is deformed by using the equation _{^{(4) R = cos -1 (}} n d s / (d 0 / Δx)) (34a)
Then, if the equation (19) is further substituted and transformed, R = cos ⁻¹ ((d _s / Δx) / (d ₀ / Δx))
^{_{= Cos -1 (d s / d}} 0) (34b)
become.

The geometric meaning of the radius R will be described with reference to FIG. R determined by the equation (34b) indicates a necessary condition for the existence of a plane that passes through “a point P ₀ at a distance d ₀ from the camera center O” and that has a shortest distance d _s . That is, due to the presence of the plane shows that the orientation n _s must have inclined by "angle R determined by equation (34b)" in the direction p ₀ of the point.

There are many such “planes passing through the point P _{0 and} having the shortest distance d _s ” with p ₀ as the rotation axis. When you draw all foot P _h of their planes vertical, "right cone (crest length d _s, apex angle 2R) whose vertices camera center O" of FIG. 16 become the bottom of the. The line of intersection between the edge of this right cone and the “unit sphere centered on O” becomes a small circle. The small circle transformation described in 2.2.1 is to transform the “point direction p ₀ ” into this small circle.

The expression (34b) can also be expressed as the expression (35) using the “standardized shortest distance _n d _s of the plane” and the “standardized shortest distance _n d ₀ of the point” defined by the expression (36). ,
R = cos ⁻¹ ( _n d _s / _n d ₀ ) (35)
_n d ₀ = d ₀ / Δx (36)
Also, the following equation can be expressed by substituting equation (6) into equation (34a).

^{_{R = cos -1 (n d s}} / (p inf p 0 p 1)) (37)
(2) Geometric meaning of small circle transformation

From the above consideration, the small circle transformation on the spherical surface of 2.2.1 is equivalent to the next transformation in the three-dimensional space. That is, the “point P _{0 in} space (direction p ₀ , distance d ₀ )” is the circumference of the bottom of the right cone in FIG. 16—that is, the shortest distance from the camera center O through “P ₀ is d _s . This is equivalent to transforming into the “all-plane” normal vector group { _ns }. This is the geometric meaning of small circle transformation.

2.2.3. Another method of measuring the standardized shortest distance _n d _s and the three-dimensional orientation n _s of the plane 2.2.2. A method of measuring the standardized shortest distance _n d _s and the three-dimensional orientation n _s of the plane using the “small circle transformation” described in the above will be described. This is done in four steps:

(1) The normalized shortest distance parameter _n d _s is arbitrarily set.

(2) For each point in the image, the current and next time positions p ₀ and p ₁ are known from the camera image, and the position p _inf after the infinite time has elapsed is known from the moving direction v, and the radius of the small circle transformation R is calculated by equation (32).

(3) to a small circle converts each point _{p 0.} That is, a small circle having a radius R centered on p ₀ is drawn on the spherical surface. Here, the meaning of drawing this small circle will be described. If the normalized shortest distance parameter _n d _s given in step (1) is the true normalized shortest distance _n d _s0 , a plane normal vector n _s0 is obtained as an intersection of these small circles. However, in step (1), since the parameter _n d _s is arbitrarily given, generally it does not intersect at one point. Therefore, small circle drawn here would determine the candidate of the normal vector n _s plane. The intensity of the small circle is set to “brightness at position p ₀ in the image”, and at a place where a plurality of small circles intersect, the intensity of the small circles is added.

(4) The above steps are performed by changing the normalized shortest distance parameter _n d _s, and a parameter value _n d _{s0 at} which a plurality of small circles drawn in step (3) intersect at one point is obtained. As the parameter value, “standardized shortest distance _n d _s0 from the camera center O to the plane” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

Here, the above small circle transformation method will be described geometrically with reference to FIG. For each point in the image, the current and next times p ₀ , p ₁ are known from the camera image, the position p _inf after the infinite time has elapsed is known from the moving direction v, and the normalized shortest distance parameter _n d _s is Given in step (1). By substituting these into equation (32), the radius R is determined, and small circle transformation is performed as shown in FIG. In other words, draw a small circle of a radius R on a spherical surface around the p _0. Next, the spherical surface of FIG. 17A is projected onto a plane in the same manner as in Step (4) of 1.3.3, and the image on the spherical surface is converted into a “circle”. The circles are stacked with the normalized shortest distance parameter _n d _s as the vertical axis to form a “cylindrical array” in FIG.

The normalized shortest distance parameter _n d _s given in step (1) specifies the height coordinate of this cylinder. In steps (2) and (3), the cross-sectional circle at that height— A spherical image of 17 (A) is converted into a “circle”. Since the parameter _n d _s is arbitrarily given in step (1), the small circles do not intersect at one point as shown in the upper surface of FIG. 17B, but the height is a true normalized shortest distance _n d _s0. In the cross-sectional circle equal to, the small circles intersect at one point. Therefore, the standardization of the plane as "height coordinate" of the cylinder shortest distance _n t _c0 can be obtained and the three-dimensional orientation n _s as "intersection coordinates in cross-section circle" is obtained.

Here, the range and nature of the parameter _n d _s in which the small circle transformation is established are examined. This conversion is established because | cos (R) | ≦ 1 in equation (35), that is,
| _N d _s | ≦ | _n d ₀ | (38)
When it is in the range. In this range, the radius R varies according to the parameter _n d _{s as} follows. _{When n} d _s = 0, the radius R of the small circle becomes π / 2 (that is, the great circle), and as | _n d _s | becomes larger, the radius R becomes smaller, and | _n d _s | ≦ | _n d ₀ | R becomes zero. The "radius R is zero _n d _s" is 15, corresponding to the "plane orthogonal as vectors 0P ₀ points _{P 0".}

Based on the above discussion, FIG. 17C will explain to which geometrical shape in the cylindrical array an arbitrary point p ₀ in the image is transformed into a small circle. Since the center of the small circle to be transformed is at p ₀ regardless of the height _n d _s and the radius R changes as described above, the point p ₀ is the surface of a “spheroid-like solid” Is converted to a small circle. Its vertex is _n d ₀ and its rotation axis is p ₀ . If there are a plurality of points in the image, the “spheroid surface” obtained by transforming them into small circles intersects at one point, and the coordinates of the intersection include the three-dimensional azimuth n _s0 of the plane and the normalized shortest distance _n d _s0. can get. In FIG. 17C, the “spheroid” is drawn when p ₀ is at the center of the cylinder, but when it is off the center, the ellipsoid protrudes from the cylinder, but that part need not be drawn.

2.3 Confirmation by simulation It is shown by computer simulation that the “algorithm for measuring three-dimensional geometric information of a plane” explained in 2.2.3 is correct (FIG. 18). The simulation was performed according to the flow of Embodiment A-3.

First, input data will be described. There is a plane perpendicular to the front of the spherical camera (or eyeball), and the distance to the camera center O is 3 m. The plane moves in a direction perpendicular to it (ie, in a direction parallel to the normal vector n _{s0 of the} plane) toward the camera at a speed of 1 m / sec. There are eight points on the plane, and “the current and next time positions p ₀ , p ₁ ” of these points are observed as input image data. Note that the time difference Δt from the present time to the next time is 0.05 seconds, and therefore the movement distance Δx from the present time to the next time is 0.05 × 1 = 0.05 m. The position p _{inf at} infinite time is equal to the moving direction v of the camera and is at the center of the visual field. From the above, the normalized shortest distance _n d _s0 of the plane is 30.05 = 60 according to the equation (19), and the normal vector n _{s0 of the} plane is at the center of the visual field.

From position _{p inf} current position _p 0, _{p 1} and infinite time of the next time, the small circle conversion algorithm 2.2.3, was determined three-dimensional geometric data of the plane _{(n d s0} and _{n s0)} The simulation result is shown in FIG. “Cylinder cross-sectional circles with each normalized shortest distance parameter _n d _s ” described in FIG. Each cross-sectional circle is projected on the plane passing through the sphere by the “equal distance projection method described in 1.3.3 (see equation (107c) of the embodiment)” in FIG. 17A. The lower right is a cross-sectional circle at _n d _s = 0, and is arranged in the order of increasing parameter _n d _s toward the upper left. Each cross-sectional circle will be described. Position p ₀ of the current time are depicted with a dot. The “small circle with the radius R calculated by the equation (32)” is drawn corresponding to the eight points on the plane with the p ₀ as the center.

In the first cross-sectional circle (bottom right, _n d _s = 0), these small circles are scattered, but as the parameter _n d _s increases, it converges, and the cross-sectional circle with _n d _s of 60 (upper right corner) In the second), these small circles intersect at one point. If _n d _s becomes larger than this, it diverges again. Thus, the small circles intersect at one point by the height of _n d _s = 60. The height _n d _s is _equal to the above-described value 60 of “normalized time _n d _s0 until crossing the plane”. Also, the direction intersecting at one point is at the center of the field of view, and is equal to the above-described “plane normal vector n _s0 ”. From the above simulation, it was confirmed that the “small conversion algorithm” described in 2.2.3 is correct. In addition, it has been confirmed by simulation that the three-dimensional geometric information of the plane can be obtained correctly by the equivalent method 2.1.

2.4 Method of not needing to know the moving direction v 2.1 and 2.2.3. In the moving direction v as it is known, to determine the three-dimensional orientation n _s and the normalized shortest distance _n d _s plane. Here, a method is provided in which the direction and distance can be measured even if the moving direction v is unknown. The method is similar to 1.6. By this method, for example, even when the moving direction during photographing is not known in the Internet, video, movie image, etc., “the direction and distance” can be measured. Further, when the plane is moving, the moving direction is generally unknown, but even in that case, “the direction and distance” can be measured together with the moving direction v. The outline of the method will be described with respect to 2.2.3 (the same applies to 2.1). Assuming that the moving direction v can be in all directions, a small circle is drawn by performing “small circle transformation” of 2.2.3 for each moving direction. When they small circles obtain the moving direction intersecting at a point, what it is true of the movement direction v _0, also required a three-dimensional orientation n _s and the normalized shortest distance _n d _s of the plane as the coordinates of the intersection It is. It is done in the following steps.

  (1) The moving direction parameter v is arbitrarily set.

(2) The direction of the parameter v is set to “infinite time position p _inf ”.

(3) 2.2.3. Steps (1) to (4) are executed, and small circles for all normalized shortest distance parameters _n d _s are drawn inside the cylindrical array (FIG. 17B).

(4) The above steps are performed by changing the moving direction parameter v, and a parameter value v ₀ at which the small circle drawn in the above steps intersects at one point is obtained. This parameter value is true in the moving direction v _0. As the coordinates of the intersection, the plane direction n _s0 and the normalized shortest distance _n d _s0 are obtained. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

3. Generalization 3.1 Plane movement and camera movement In the above, it has been described that the plane moves, but even when the camera moves, the three-dimensional geometric information of the plane can be measured with the same algorithm. This is because the plane movement and the camera movement are relative movements, and therefore the same algorithm is obtained if the movement direction v is reversed. The equivalence will be described with reference to FIG.

FIG. 19B shows a case where the plane has moved in the direction v, and points on the plane also move in space with p ₀ and p ₁ . Positions on the spherical surface p ₀ and p ₁ (FIG. 8) at each time are determined by being observed as angles α ₀ and α ₁ viewed from the camera center O. The spherical position p _inf after the infinite time has elapsed is equal to the plane moving direction v. To predict these three spherical surface position _{p 0, p 1, p inf} " sphere on a position p _c at the time that the plane crosses the center of the _camera" by cross-ratio conversion from, then the plane with polar converts the p _c As described in 1.3.2 and 2.2.3, it is possible to measure the three-dimensional geometric information (three-dimensional orientation n _s , normalized time _n t _c until crossing, normalized shortest distance _n d _s ). Note that t _c , Δt, Δx, and V are respectively the time to cross the plane, the time difference from the present to the next time, the movement distance to the next time, and the magnitude of the movement speed.

On the other hand, FIG. 19A shows a case where the camera has moved. When the camera center moves to O ₀ , O ₁ , O _c while viewing the point P on the plane, the point P is observed as angles α ₀ , α ₁ , α _c viewed from the camera center at each time, and is on the spherical surface. Positions p ₀ , p ₁ , _pc are determined. The position on the spherical surface p _inf after the infinite time has elapsed is equal to the moving direction v of the plane. These positions _{p 0, p 1, p c} , when modified 19 a (A) to draw a _{p inf} on one sphere, that figure is the same as "If the plane has moved (FIG. 19 (B))" become. The only difference from FIG. 19B is that the moving direction is opposite. From the above examination, even when the camera moves, as long as the moving direction v is half-turned, the aforementioned “algorithm when the plane moves (1.3.2, 2.1, 2.2.3)” It can be seen that the three-dimensional geometric information of the plane can be obtained using. Similarly, the measurement of the point distance when the camera moves can also be performed using the “algorithm when the plane moves (formula (6))”.

3.2 Voting for Cylinder Arrangement In FIG. 10B, it is assumed that “curve obtained by projecting a great circle on a spherical surface onto a plane” is “drawn” on each cross-sectional circle of the cylinder. Instead of drawing, each cross-sectional circle may be constituted by a memory array or a register array, and “voting” may be performed on the memory or register corresponding to the curve.

  In FIG. 17B, it is assumed that a “curve obtained by projecting a small circle on a spherical surface onto a plane” is “drawn” on each cross-sectional circle of the cylinder. Instead of this drawing, “voting to a memory or register” may be used as described above.

3.3 Polar transformation on a plane First, a general definition of polar transformation (also called dual transformation) will be described with reference to FIG. Consider an arbitrary vector a and a “plane π that makes it a normal vector and passes through the center O”. The conversion from the vector a to the plane π is a “polar conversion (or dual conversion)” in a broad sense. In order to express the vector and the plane two-dimensionally, an intersection or a line of intersection with the spherical surface or the plane is used.

In the case of a spherical surface, if the intersection point with the vector a is a _sphere and the intersection line with the plane π (that is, a great circle) is g, a _sphere and g are a spherical pole and a polar line, respectively. The conversion from the polar a _sphere to the polar line g is the “polar conversion on a spherical surface”. On the other hand, in the case of a plane, if the intersection with the vector a is a _plane and the intersection line (ie, a straight line) with the plane π is l, a _plane and l are a pole and a polar line on the plane, respectively. The conversion from the polar a _plane to the polar line l is “polar conversion on a plane”.

1.3.2 (3), drawing the polar converting the _{p c} electrode as a "great circle on the sphere." Based on the above description, the polar line may be drawn as a “straight line”.

In 2.1 (5) and 2.2.3 (3), a small circle obtained by converting p ₀ into a small circle is drawn on the “spherical surface”. This small circle may be projected from the camera center onto an “arbitrary plane” and drawn as an ellipse.

3.4. Method of measuring the normalized distance _n d ₀ to the point Knowing the current time, the next time, and the infinite time on the spherical surface p ₀ , p ₁ , p _inf , measure the normalized distance _n d ₀ to the point What can be done is shown in FIG. Here, the normalized distance _n d ₀ of the point is the distance to the point (that is, the distance from the camera center O to the point P ₀ in FIG. 7) d ₀ , and the camera (or point) from the present time to the next time. The distance is normalized by the moving distance Δx. The measurement method is shown below.

It was described in 1.2 that the distance d ₀ to the point can be measured by the equation (6) using the three positions p ₀ , p ₁ and p _inf . When both sides of Equation (6) are normalized by the camera movement distance
_n d ₀ = d ₀ / Δx
= (P _inf p ₀ p ₁ ) (39a)
Therefore, it was shown that the normalized distance _n d ₀ can be measured as a _single ratio (p _inf p ₀ p ₁ ). Substituting equation (13) (or equation (15)) into equation (39a) and expressing it as a central angle yields the following equation.

_n d ₀ = sin (b) / sin (ba) (39b)
= Sin (a + τ) / sin (τ) (39c)
3.5 Planar Camera Although it has been described above that an image from a spherical camera (or eyeball) is used, an image from a plane camera can also be used (FIG. 21). In that case, an image captured by a plane camera (white circle triangle) may be converted into an image on a spherical surface (black circle triangle) and the above-described “spherical algorithm” may be executed.

4). Method of measuring three-dimensional geometric information from stereo images The parameter names of the motion vision algorithm described above--that is, the algorithm for measuring three-dimensional geometric information of a plane by moving the camera (or plane)-binocular vision In other words, 3D geometric information can be measured from a stereo image. That is, the current time position p ₀ , the next time position p ₁ , and the infinite time position p _inf of the kinematic algorithm are respectively represented as a position p _{R in} the right camera image, a position p _L in the left camera image, and When replaced with “position p _axis on the visual axis connecting the left and right cameras”, the algorithm is used to normalize (i) the three-dimensional orientation n _{s of} the plane from the stereo image, and (ii) cross the plane in the visual axis direction. The distance _n d _c , (iii) the normalized shortest distance _n d _s to the plane, and (iv) the normalized distance _n d ₀ to the point can be determined. A method for measuring the azimuth and distance will be described below. Here, the normalized distance _n d _{c to} the plane, the normalized shortest distance _n d _s to the plane, and the normalized distance _n d ₀ to the point are respectively “from the right camera until the plane crosses the plane in the visual axis direction”. The distance d _c ”,“ the shortest distance d _s from the right camera to the plane ”, and“ the distance d ₀ from the right camera to the point ”are normalized by the distance Δx _LR between the left and right cameras (FIG. 22). reference). These distances are expressed by the following equation. Note that the right camera and the left camera may be exchanged as described above.

_{n d c = d c / Δx} LR (50)
_n d _s = d _s / Δx _LR (51)
_n d ₀ = d ₀ / Δx _LR (52)
4.1 Correspondence with Motion Vision The binocular vision parameters (FIG. 22) correspond to the parameters (FIG. 19) used in the above motion vision algorithm as follows.

FIG. 22A shows a state when “point P on the plane” is viewed with both eyes. Center _O L of the right and left cameras, _O angle of the point P as viewed from the _{R α L, α R} is observed, whereby the position _p L on the spherical surface, _{p R} is determined. A position p _axis on the visual axis connecting the left and right cameras—corresponding to a position p _inf after an infinite time (for example, FIG. 19B) _—is equal to the visual axis direction a _xis . The position of the center of the camera when it is assumed that the right camera moves in the visual axis direction until it crosses the plane is O _c . O angle of the point P as viewed from the _c alpha _c is observed, whereby the position p _c on the spherical surface is determined.

When this figure is compared with the above-mentioned FIG. 19A, it is exactly the same except for the name. That is, the current time position p ₀ , the next time position p ₁ , the infinite time position p _inf , the unit movement distance Δx, and the distance Vt _c to cross the plane in the movement direction in FIG. Replace with the position p _{R in} the right camera image, the position p _L in the left camera image, the position p axis on the visual _axis , the distance Δx _LR between the left and right cameras, and the distance d _c until crossing the plane in the visual axis direction. And FIG. 22 (A). FIG. 22B is a modification of FIG. 22A in which the four positions p _R , p _L , p _c , and p _axis are overlaid on the “sphere centered on O _c ”. in _{and, "p 0, p 1,} p c, p inf in visual motion" is the same as FIG. 19 is replaced in (B) their four positions.

  Thus, the geometrical relationship between motion vision and stereo vision is exactly the same when the parameter names are exchanged as described above. Therefore, it can be seen that the three-dimensional geometric information of a plane can be measured from a stereo image by the same algorithm by changing the parameter names of the above-mentioned motion vision algorithms (1.3.2, 2.1, 2.2.3). This will be described in detail below.

4.2 Similar Procedure and three-dimensional orientation _{n s} and how visual motion algorithm that measures the "normalized distance _n d _c until crossing the visual axis direction in the plane" plane (1.2 and 1.3), it can be measured three-dimensional orientation n _s and the normalized distance _{n d} c _"plane.

4.2.1 In cross ratio of the normalized distance _{n d c {p axis p R} p L p c} represent views 22 by (B), the use of sine theorem triangle _P R _P L _{O c,} from _{O c} point _P and the distance _{d 0} to _R between the distance [Delta] x _LR between the left and right cameras _{Δx LR / sin (α L -α} R) = d 0 / sin (π-α L) (53a)
When this is expressed by the positions p _L , p _R , and p _axis on the spherical surface, the following equation is obtained.

Δx _LR / sin (p _R p _L ) = − d ₀ / sin (p _axis p _L ) (53b)
When this is deformed, the distance d ₀ is obtained by the following equation.

d ₀ = −Δx _LR sin (p _axis p _L ) / sin (p _R p _L ) (54a)
= −Δx _LR (p _axis p _R p _L ) (54b)
Further, when the sine theorem is applied to the triangle P _R P _L O _c and deformed in the same manner as described above, the distance d ₀ is obtained by the following equation.

_{d 0 = -d c sin (p} axis p c) / sin (p R p c) (55a)
_{= -D c (p axis p R} p c) (55b)
If the ratio of the formula (54) and the formula (55) is taken and arranged, the normalized distance _n d _c is obtained as a double ratio {p _axis p _R p _L p _c } by the following formula.

_{n d c = d c / Δx} LR
= (Sin (p _axis p _L ) / sin (p _R p _L ))
_{/ (Sin (p axis p c} ) / sin (p R p c)) (56a)
= {P _axis p _R p _L p _c } (56b)
Note that this equation is expressed by the equation (12a) in which “motion vision parameters _n t _c , p _inf , p ₀ , p ₁ , p _c ” are changed to “binocular vision parameters _n d _c , p _axis , p _R , p _R , p _L, has become to those exchange in _{p c} ".

4.2.2 cross-ratio conversion The cross ratio conversion, four variables _{n d c, p R, p} L, knowing _{p axis,} the remaining variables _{p c} in the conversion be calculated using equation (56) is there. This corresponds to the cross ratio conversion (1.3.1) in the case of motion vision.

This multi-ratio conversion is specifically shown by a mathematical expression. A cross section of the sphere of FIG. 22B is taken out and shown in FIG. and the p _axis base point indicates the position of the _{p R, p L, p c} then the central angle c, d, with x (Note: This base may be any position). These central angles are summarized below.

p _axis p _R = c (57)
p _axis p _L = d
p _axis p _c = x
p _R p _L = d-c
p _R p _c = _x-c
Using these center angles, the cross ratio conversion is expressed by a mathematical expression. When the right side of the formula (56a), that is, the cross ratio is expressed using the central angle of the formula (57),
_{n d c = (sin (d} ) / sin (d-c))
/ (Sin (x) / sin (x−c)) (58a)
It becomes. By transforming this, the central angle x between _{p c} and _{p axis} is ^{x = tan -1 ((sin (} c) sin (d))
_{/ (Cos (c) sin (} d) - n d c sin (d-c))) (58b)
Given in. Accordingly, the normalized distance _n d _c and "on three spherical position _{p R, p L, p axis} " Given a "position _{p c} of when the right camera across the plane (FIG. 22 (B))" has the formula (58b). This is the mathematical formula for the cross ratio conversion. Note that equation (58b), the parameter of the visual motion in the formula (14b) (normalized time _n t _c and the central angle _{p inf p 0, p inf p} 1, p 0 p 1, p inf p c) binocular viewing parameters (normalized distance _n d _c and the central angle _{p axis p R, p axis p} L, p R p L, p axis p c) to those that have been replaced - i.e. parameters of visual motion (normalized time _n t _c spherical on position _{p inf, p 0, p 1} , p c) the parameters of binocular vision (normalized distance _n d _c and the central angle _{p axis, p R, p L} , which was replaced with _{p c)} - -It has become.

Here, in general stereo image processing research, instead of the “left camera position p _L ” used above, “a change from the right camera p _L −p _R (that is, binocular parallax σ and a central angle) p _R p is represented by _L) it is often handle ". A mathematical expression of the conversion of the cross ratio in this case is shown below. Various central angles are p _axis p _R = c (59)
p _R p _L = σ
p _axis p _c = x
p _axis p _L = c + σ
p _R p _c = _x-c
And the right side of Expression (56a) is expressed using the central angle of Expression (59).
_{n d c = (sin (c} + σ) / sin (σ))
/ (Sin (x) / sin (x−c)) (60a)
It becomes. By transforming this, _{p c} and _p central angle x is x ⁼ tan with _{inf -1 ((sin (c)} sin (c + σ))
/ (Cos (c) sin (c + σ) _−n d _c sin (σ))) (60b)
And another mathematical expression of the cross ratio conversion was obtained. Note that equation (60b), the parameter of the visual motion in the formula (16b) (normalized time _n t _c and the central angle _{p inf p 0, p inf p} 1, p 0 p 1, p inf p c) binocular viewing parameters (normalized distance _n d _c and the central angle _{p axis p R, p axis p} L, p R p L, p axis p c) to those that have been replaced - i.e. parameters of visual motion (normalized time _n t _c spherical on position _{p inf, p 0, p 1} , p c) the parameters of binocular vision (normalized distance _n d _c and spherical on position _{p axis, p R, p L} , which was replaced with _{p c)} --It has become.

4.2.3 using 3-dimensional orientation _{n s} and cross-ratio conversion method for determining the above normalized distance _n d _c plane, a method of determining the three-dimensional orientation _{n s} and the normalized distance _n d _c of the plane explain. This is the same as in the case of motion vision (1.3.2). This is done in four steps:

(1) The normalized distance parameter _n d _c is arbitrarily set.

(2) For each point in the image, the positions p _L and p _R on the left and right cameras are known from the camera image, and the position p _axis on the visual _axis is _known from the visual axis direction a _xis , and these are _expressed by the equation (58b) Alternatively, the ratio _pc is calculated by substituting into the equation (60b) to perform the cross ratio conversion.

(3) 1.1 based on "the principle of measuring the three-dimensional orientation _{n s} plane", determining a candidate of the normal vector _{n s} plane. That is, specifically, to polar transformation to p _c obtained in step (2), draw a great circle on a sphere. Here, the meaning of drawing a great circle will be described. If the normalized time parameter _n d _c given in step (1) is the true normalized distance _n d _c0 , as described in FIG. 5, the plane normal vector n _s0 is used as the intersection of these great _circles. I want. However, in step (1), the parameter _n d _c is arbitrarily given, and therefore generally does not intersect at one point. Therefore, great circle drawn here would determine the candidate of the normal vector n _s plane. The intensity of the great circle is the "brightness of the position p _R within the image", also in places where a plurality of great circle intersects to the strength obtained by adding the intensity thereof great circle.

(4) The above steps are performed by changing the standardized distance parameter _n d _c, and a parameter value _n d _{c0 at} which a plurality of great circles drawn in step (3) intersect at one point is obtained. As the parameter value, “normalized distance _n d _c0 until the plane is crossed in the visual axis direction” is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

In this method, in 1.3.2, the parameters for motion vision (standardized time _n t _c and positions on the spherical surface p _inf , p ₀ , p ₁ ) are changed to binocular vision parameters (normalized distance _n dc ₎ . The positions on the spherical surface are changed to p _axis , p _R , p _L ).

  4.2.4 Geometric meaning of the above steps

  The geometric meaning of the above steps will be described with reference to FIG. In the “geometrical meaning in the case of motion vision” described in 1.3.3, the parameters for motion vision are changed to the parameters for binocular vision.

Drawing candidates plane orientation {n _s} (Step (3)): Step (2) to polar transformation the position p _c as determined by the cross-ratio conversion, a great circle on the sphere, i.e. the plane orientation candidate The group {n _s } is drawn as shown in FIG.

Determine the three-dimensional geometric information as the coordinate value of the cylindrical array (step (4)): project the spherical surface of FIG. 24 (A) onto the plane in the same manner as in step (4) of 1.3.3, Convert the image on the sphere into the inside of a “circle”. The circles are stacked with the normalized shortest distance parameter _n d _c as the vertical axis to form a “cylindrical array” as shown in FIG. In this way, the geometric meaning of step (1) becomes clear. That is, the normalized distance parameter _n d _c arbitrarily given in step (1) specifies the height coordinate of this cylinder, and in steps (2) and (3), the cross-sectional circle at that height is specified. —Drawing the spherical image in FIG. 24 (A) into a “circle” — Since the parameter _n d _c is arbitrarily given in step (1), the great circles do not intersect at one point as shown in the upper surface of FIG. 24B, but the height is _{equal to the} true normalized distance _n d _c0 . In equal cross-sectional circles, the great circles intersect at one point. Accordingly, the normalized distance _n d _c0 of the plane as "height coordinate" cylinder is obtained, also 3-dimensional orientation n _s as "cross coordinates in the section circle" is obtained (FIG. 24 (B)).

4.2.5 visual axis direction _{a xis} more good way without having to know, as a visual axis direction _{a xis} is known, a method of measuring the three-dimensional orientation _{n s} and the normalized distance _n d _c of the plane Stated. Namely, such provision of the p _axis as spherical surface positioned equal to the direction a _xis, then performs a cross-ratio conversion using the position was determined at the end thereof by performing a polar transformation orientation n _s and the distance _n d _c.

Here is the visual axis direction _{a xis} --1.6 a similar manner to provide a method which can measure even "three-dimensional orientation _{n s} and the normalized distance _n d _c plane" unknown. By this method, for example, in the case of a stereo image such as the Internet, even when the visual axis direction during photographing is not known, it is _possible to measure with “the direction and distance” along the visual axis direction axis. The outline of the method will be described. _Assuming that the visual axis direction a _xis can be in all directions, _42.3 “multi-ratio conversion and polar conversion” is performed on each visual axis direction to obtain a polar line. Draw. When they pole line seek visual axis direction intersecting at one point, it is required a true a visual axis direction a _Xis0, also the three-dimensional orientation n _s and the normalized distance _n d _c of the plane as the coordinates of the intersection Is. It is done in the following steps.

(1) The movement direction parameter a _xis is arbitrarily set.

(2) The direction of the parameter a _xis is set to “position p _axis on the spherical surface”.

(3) 4.2.3. Step (1) running through (4), draw a polar for all normalized distance parameter _n d _c inside the cylinder arrangement (FIG. 24 (B)).

(4) The above steps carried out by changing the visual axis direction parameter a _xis, obtaining the parameter values a _Xis0 the polar depicted in step intersect at one point. This parameter value is the true visual axis direction _axis0 . As the coordinates of the intersection, a plane direction n _s0 and a normalized distance _n d _c0 until crossing the plane are obtained. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

4.3 visual motion described in --2 to provide a method for measuring the "normalized shortest distance _{n d} s of formula _(51)" 3-dimensional orientation n _s and a normalized shortest distance measurement method plane to the plane This is done in the same way as the algorithm. Between the standardized shortest distance _n d _s and “standardized distance _n d _c ” described in 4.2.3, n _s is the three-dimensional orientation of the plane, a _xis is the visual axis direction, and () As a scalar product
_n d _s = _n d _c (n _s · a _xis ) (61)
There is a relationship. The reason will be described with reference to FIG. FIG. 25 shows a cross section in the plane in which the normal vector n _{s of the} plane and the visual axis direction a _xis lie. Since "right camera center O shortest distance from _R to the plane d _s" is the normal direction of the component of the "distance d _c of the visual axis direction until crossing the _{plane",
d s = d c (n s} · a xis) (62)
There is a relationship. To normalize both sides of this equation by the camera distance [Delta] x _LR, equation becomes the following equation (61) is obtained.

_n d _s = d _s / Δx _LR = (d _c / Δx _LR ) (n _s · a _xis )
_{= N d c (n s ·} a xis) (63)
4.3.1 Method of measuring standardized shortest distance _n d _s and three-dimensional orientation n _s of the plane “ _n d _s and _n d _c ” is applied to the algorithm described in 4.2.3 (multi-ratio conversion and polar conversion). combining the relationship between the (formula (61)) ", will be described with reference to FIG. 26 to be able to measure the three-dimensional orientation n _s and the normalized shortest distance _n d _s plane. --This is done in the same way as the motion vision algorithm described in 2.1. It is performed in 6 steps.

(1) The normalized shortest distance parameter _n d _s is arbitrarily set.

(2) Considering a small circle centered on the visual axis direction a _xis , its radius r is arbitrarily set (FIG. 26). In step (4), the “plane three-dimensional orientation candidate n _s ” on the small circle is obtained. For that purpose, it is necessary to set the normalized distance parameter _n d _c to a value determined by the following equation.

_n d _c = _n d _s / cos (r) (64)
The reason will be explained. First, since n _s is above “a small circle with a radius r centered on a _xis ”, there is a relationship between n _s , a _xis , and r:
_{cos (r) = (n s} · a xis) (65a)
Also, since n _s needs to satisfy equation (61), substituting equation (65a) into equation (61)
_n d _s = _n d _c cos (r) (65b)
This is transformed to obtain equation (64).

(3) For each point in the image, the spherical positions p _L and p _R of the left and right cameras are known from the camera image, and the position p _axis on the visual _axis is _known from the visual axis direction a _xis , and their normalized distances The position n _c is calculated by substituting the parameter _n d _c into the equation (58b) or the equation (60b) to perform the cross ratio conversion.

(4) Step (3) to polar transformation to _{p c} determined, draw a great circle _{g pc} on a spherical surface. The great circle, the small circle in step (2), and the two intersections _r n _{s +} and _r n _s− are “planar three-dimensional orientation candidates” (FIG. 26). Note that the latitude coordinate of this intersection is r (see FIG. 26), and the longitude coordinate points _r α _{s +} and _r α _s− are _expressed as “motion vision parameters α _a , a, and τ” in Equation (29). Substituting the binocular vision parameters α _c , c, σ ″,
_r α _{s +} = α _c + cos ⁻¹ ((( _n d _s sin (σ) / sin (c + σ))
-Cos (r) cos (c)) / (sin (r) sin (c)))
^{_{r α s- = α c -cos -1}} (((n d s sin (σ) / sin (c + σ))
-Cos (r) cos (c)) / (sin (r) sin (c)))
(65c)
Calculated by Here, α _c and c are the longitude and latitude coordinates of p _R with respect to “parameters α _a and a of motion vision in FIG. 14 (that is, longitude and latitude coordinates of p ₀ )”.

(5) Step (2) is performed by changing-the radius r (4), two intersections _r n _s + which is obtained in step _(4), a curve consisting of r _{n s-(Figure} 26) - This curve , the "small circle around the _{p R"} as described in 4.3.2. Here, the meaning of drawing this curve will be described. If the normalized shortest distance parameter _n d _s given in step (1) is the true normalized shortest distance _n d _s0 , a plane normal vector n _s0 is obtained as the intersection of these curves. However, in step (1), since the parameter _n d _s is arbitrarily given, generally it does not intersect at one point. Thus, the curve drawn here would determine the candidate of the normal vector n _s plane. The intensity of the curve is the "brightness of the position p _R within the image", and in places where multiple curves intersect to strength obtained by adding the intensity thereof curves.

(6) The above steps are performed by changing the normalized shortest distance parameter _n d _s, and a parameter value _n d _{s0 at} which the plurality of curves drawn in step (5) intersect at one point is obtained. As the parameter value "right camera center O standardization from _R to the plane shortest distance d _s0" is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection. Instead of detecting the above intersection, a point where the intensity reaches a peak may be detected.

4.3.2 normalized shortest distance of the plane _n d _s and another method Figure 26 to measure the three-dimensional orientation _{n s} "parameter _p R about binocular _{vision, p L, p axis, a} xis" a "movement in other the parameter _{p 0, p 1, p inf} , v " relates view the same as FIG. 13. Therefore, the curve in step (5) of 4.3.1 can be used as it is in the proof of 2.2.1, and is a “small circle with radius R around p _R ”. The radius R is equal to “parameters for motion vision τ, a, p ₀ p ₁ , p _inf p ₁ ” in equations (31) and (32) and “parameters σ, c, p _R p _{L for} binocular vision”. , P _axis p _L ″ is expressed by the following formula.

^{_{R = cos -1 (n d s}} sin (σ) / sin (c + σ)) (66a)
^{_{= Cos -1 (n d s sin}} (p R p L) / sin (p axis p L)) (66b)
From the above, it was shown that the curve composed of the two intersections _r n _{s +} and _r n _s− in FIG. 26 is “a small circle with radius R centering on p _R ”. This can be regarded as converting an arbitrary point p _R into this small circle, and this conversion will be referred to as “small circle conversion”. The nature of this small circle transformation is the same as that of motion vision and is shown in 2.2.2. Although the radius R is expressed using the binocular parallax σ in the equation (66a), when the position P _L with the left camera is known, the equation (57) is substituted into the equation (66b) and Given.

^{_{R = cos -1 (n d s}} sin (d-c) / sin (d)) (66c)
Based on the above preparation, another method for measuring the standardized shortest distance _n d _s and three-dimensional orientation n _s of the plane using “small circle transformation” will be described. It is performed in the following four steps-the same procedure as the motion vision algorithm described in 2.2.3.

(1) The normalized shortest distance parameter _n d _s is arbitrarily set.

(2) For each point in the image, know the positions p _R and p _L on the left and right cameras from the image, and know the “position p _axis on the visual axis connecting the left and right cameras” from the visual axis direction a _xis. The radius R of the circle conversion is calculated by the equation (66b).

(3) small circle convert each point _{p R.} - That, draw a small circle of radius R centered on p _R on a sphere. Here, the meaning of drawing this small circle will be described. If the normalized shortest distance parameter _n d _s given in step (1) is the true normalized shortest distance _n d _s0 , a plane normal vector n _s0 is obtained as an intersection of these small circles. However, in step (1), since the parameter _n d _s is arbitrarily given, generally it does not intersect at one point. Therefore, small circle drawn here would determine the candidate of the normal vector n _s plane. The intensity of the small circle in the "brightness of positions p _R within the image", and in a plurality of small circles intersect location to the strength obtained by adding the intensity of their small circle.

(4) The above steps are performed by changing the normalized shortest distance parameter _n d _s, and a parameter value _n d _{s0 at} which a plurality of small circles drawn in step (3) intersect at one point is obtained. As the parameter value "right camera center O standardization from _R to the plane shortest distance _{n d s0"} is obtained. In addition, the plane direction _ns0 is obtained as the coordinates of the intersection. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

Here, the above small circle transformation method will be described geometrically with reference to FIG. For each point in the image, know the positions p _R and p _L on the left and right cameras from the image, know the “position p _axis on the visual axis connecting the left and right cameras” from the visual axis direction a _xis , and further normalize the shortest distance The parameter _n d _s is given in step (1). By substituting these into equation (66b), radius R is determined, and small circle transformation is performed (FIG. 27 (A)). In other words, draw a small circle of a radius R on a spherical surface around the p _R. Next, the spherical surface of FIG. 27A is projected onto a plane in the same manner as in step (4) of 1.3.3, and the image on the spherical surface is converted into a “circle”. The circles are stacked with the normalized shortest distance parameter _n d _s as the vertical axis to form the “cylindrical array” in FIG.

The normalized shortest distance parameter _n d _s given in step (1) specifies the height coordinate of this cylinder. In steps (2) and (3), the cross-sectional circle at that height— This is a 27-A spherical image converted into a “circle”. Since the parameter _n d _s is arbitrarily given in step (1), the small circles do not intersect at one point as shown in the upper surface of FIG. 27B, but the height is the true normalized shortest distance _n d _s0. In the cross-sectional circle equal to, the small circles intersect at one point. Accordingly, the standardized shortest distance _n d _s0 of the plane is obtained as the “height coordinate” of this cylinder, and the three-dimensional azimuth _{ns 0} is obtained as the “intersection coordinate within the cross-sectional circle”.

4.3.3 visual axis direction _a good way without having to know the _xis 4.3.1 and in 4.3.2, visual axis direction _a as _xis is known, three-dimensional orientation _{n s} plane and standards The shortest distance _n d _s was obtained. Here, visual axis direction a _xis is to provide a method that can measure their orientation and distance in the unknown. This is similar to 2.5. In this way, for example in a stereo image, such as the Internet, a "them bearing and distance" even if the visual axis direction in the imaging is unknown can be measured with a visual axis direction a _xis. An outline of the method will be described with respect to 4.3.2 (the same can be applied to 4.3.1). _Assuming that the visual axis direction a _xis can be in all directions, a small circle is drawn by performing “small circle transformation” of 4.3.2 for each visual axis direction. When they small circles seek visual axis direction intersecting at one point, it is true of the visual axis direction a _Xis0, also obtains the three-dimensional orientation n _s and the normalized shortest distance _n d _s of the plane as the coordinates of the intersection It is what It is done in the following steps.

(1) The movement direction parameter a _xis is arbitrarily set.

(2) The direction of the parameter a _xis is a position p _axis on the spherical surface.

(3) Steps (1) to (4) of 4.3.2 are executed, and small circles for all normalized shortest distance parameters _n d _s are drawn inside the cylindrical array (FIG. 27B).

(4) The above steps are performed while changing the visual axis direction parameter a _xis, and a parameter value a _xis0 at which the small circle drawn in the above step intersects at one point is _obtained . This parameter value is the true visual axis direction _axis0 . As the coordinates of the intersection, the plane direction n _s0 and the normalized shortest distance _n d _s0 are obtained. Instead of the above intersection detection, a point where the intensity reaches a peak may be detected.

4.4 Generalization 4.4.1 Voting for Cylinder Array In FIG. 24B, it is assumed that “curve obtained by projecting a great circle on a spherical surface onto a plane” is “drawn” on each cross-sectional circle of the cylinder. Instead of drawing, each cross-sectional circle may be constituted by a memory array or a register array, and “voting” may be performed on the memory or register corresponding to the curve.

  In FIG. 27B, it is assumed that a “curve obtained by projecting a small circle on a spherical surface onto a plane” is “drawn” on each cross-sectional circle of the cylinder. Instead of this drawing, “voting to a memory or register” may be used as described above.

4.4.2 polar transformation on a plane 4.2.3 (3), drawing the polar converting the _{p c} electrode as a "great circle on the sphere." The polar line may be drawn as a “straight line on the plane” similarly to 3.3.

Further, in 4.3.1 (5) and 4.3.2 (3), depicting the small circle and a small circle converts _{p R} on the "sphere". As in the case of 3.3, this small circle may be projected from the camera center onto an “arbitrary plane” and drawn as an ellipse.

4.4.3 Method for Measuring Normalized Distance _n d ₀ to Point In Expression (39a), “Moving Vision Parameters Δx, p ₀ , p ₁ , p _inf ” is set to “Binocular Viewing Parameters Δx _LR , p _R , p _L , p _axis ”
_n d ₀ = d ₀ / Δx _LR
= (P _axis p _R p _L ) (67a)
become. With this equation, the normalized distance _n d ₀ to the point can be measured. Substituting equation (57) (or equation (59)) into equation (67a) and expressing it as a central angle yields the following equation.

_n d ₀ = sin (d) / sin (dc) (67b)
= Sin (c + σ) / sin (σ) (67c)
4.4.4 Planar Camera In the above description, it is assumed that an image from a spherical camera (or eyeball) is used, but an image from a plane camera can also be used. In that case, the algorithm described above may be executed by converting the image captured by the plane camera onto a spherical surface, as in 3.5.

  Above, the fundamental description of the present invention is finished, and various embodiments of the present invention will be described below. Note that the various block diagrams described below can be regarded as functional blocks of the computer system 300 shown in FIGS. 2 and 3, and are regarded as various embodiments when the image measuring apparatus of the present invention is configured by hardware. You can also. The various flowcharts described below can also be regarded as various embodiments of the image measurement program according to the present invention installed and executed in the computer system 300 shown in FIGS. 2 and 3. It can also be understood as various embodiments of the method.

Embodiment A Motion Vision Embodiment A-1. (Measurement of normalized time _n t _c0 until crossing with the three-dimensional orientation n _{s0 of the} plane)
The method of posting, that is, the method of 1.3.2 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG.
(Start)
(1) The position p _inf at infinite time is set as follows (A-1-1). Obtained by the camera 11, the position of all points in the current and next time _{{i p 0}, {i} p 1} the "image register 12 at the present time _{t 0",} "image register 13 at the next time _{t 1} To the “extraction unit 14 in the moving direction v” to extract the moving direction v—a method and an embodiment for performing this extraction are described in a patent (Japanese Patent Publication No. 06-14355). Next, the “p _inf setting unit 15” sets “infinite time position p _inf ” as being equal to the moving direction v.

(2) The standardized time parameter _n t _c is scanned from the lower limit value _n t _{c, min} to the upper limit value _n t _{c, max} by the “ _n t _c parameter scanning unit 16” (A-1-2, A-1 -3, A-1-16).
(Scans _n t _c )
(3) Each address i in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-1-4, A-1-5, A-1-15).
(Scan i)
(4) The current and next time positions _i p ₀ and _i p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-1-6).

(5) The four parameters _n t _c , _i p ₀ , _i p ₁ , and p _inf set as described above are input to the “multi-ratio conversion unit 17” and the position _{i pc} is output. Calculation of the position _i p _c is in the unit takes place in the next two steps.

_(A) calculation of the center angle _i x and i _{p c} and _{p inf} (A-1-7)
_n t _c, from _{i p 0, i p 1,} p inf, the _i x calculated by the following equation according to equation (14b) (see FIG. 9).

^{_{i x = tan -1 ((sin}} (i a) sin (i b))
_{/ (Cos (i a) sin} (i b) - n t c · sin ((i b) - (i a)))
(100a)
_(B) calculation of i _{p c} (A-1-8)
The position _i p _c on the spherical surface is calculated by the following equation using the above _{i x.}

_{i pc} = cos ( _ix ) Γ _x + sin ( _ix ) Γ _y (100b)
Here, Γ _x and Γ _y are calculated by the following equations using [×] and || as vector outer product operations and absolute value operations, respectively.

Γ _x = v (100c)
Γ _z = [v × _i p ₀ ] / | [v × _i p ₀ ] |
Γ _y = [Γ _z × Γ _x ]
(6) the position _i p _c of the poles converted to great circle on the sphere by the "polar transformation unit 18 '(see 1.3.3 and FIG. 10 (A)). The polar conversion is performed in the following two steps in the unit (see FIG. 30A).

(A) Convert _{i pc} to polar coordinates (A-1-9)
_{i pc} in Cartesian coordinates and "polar coordinates on a sphere"
_{i p c = (i p cX} , i p cY, i p cZ) (101a)
_{= (I α c, i β} c) (101b)
It is expressed _as, calculates polar components i _{p c} a (longitude _i alpha _c and latitude _i beta _c) by the following equation.

^{_{i α c = tan -1 (i}} p cY / i p cX) (101c)
^{_{i β c = tan -1 (√}} (i p cX 2 + i p cY 2) / i p cZ)) (101d)
Here, O is the “origin on the spherical surface” unit vector, and X and Y are the X-axis and Y-axis unit vectors.
(Scan k)
_(B) polar transformation of i _{p c} (A-1-12)
_i p _c poles converted great circle on a sphere - that is, any point _k p _GC constituting the great circle coordinates (the address to k) (longitude _k alpha _GC and latitude _k beta _GC) - - is determined by the coordinates _{i α c, i β c} of _i p _c, the formula (102b) (geometric dictionary 2 (Iwata ItaruYasushi, 72 pp Maki Shoten) ").

_{k p GC = (k α GC} , k β GC) (102a)
_{cos (k α GC - i α} c) = - cot (k β GC) cot (i β c) (102b)
A specific calculation of the great circle is performed as follows by scanning k from the lower limit value k _min to the upper limit value k _max . The latitude _k β _GC is calculated as kΔβ _GC using Δβ _GC as the latitude resolution, and the longitude _k α _GC is calculated by the following equation using the latitude.

_{k α GC = i α c +} cos -1 (cot (k β GC) cot (i β c)) (102c)
(7) The point _k p _GC constituting the great circle is defined as polar coordinates (inclination angle _k α _{GC, Proj} , radius _k β _{GC, in} the cross-sectional circle of the height _n t _c of the “cylindrical array voting unit 19” _{. Proj} ) is converted into a point _k p _{GC, Proj} represented by voting (A-1-13)-the voting is performed by adding “brightness at position _i p ₀ ” (1.3. 3, see FIG. 10 (B) and FIG. 30 (B)). The transformation is generally f () as a projection function,
_{k β GC, Proj = f (} k β GC) (103a)
_k α _{GC, Proj} = _k α _GC (103b)
-The details of f () are in the literature ("Latest Lens Design Lecture 23 Problems Associated with Lens Design (1) (by Nakagawa, Photo Industry, 1982)": Refer to Section 4.2.2.1 of “Work Robot Development Contract Research Results Report (Extreme Work Robot Technology Research Association)”. In the case of the equidistant projection method, f () = 1, and _k β _{GC, Proj} is given by the following equation.

_k β _{GC, Proj} = _k β _GC (103c)
Summarizing the above, the “point _k p _GC on the spherical surface” constituting the great circle is converted into the “point _k p _{GC, Proj} on the plane” in the cross-sectional circle, and the brightness of the position _i p ₀ is added to the point. Vote “Add”. Each cross-sectional circle can be realized by a register arrangement or a memory arrangement. The “algorithm for polar conversion to a great circle on a spherical surface” and its “algorithm for projecting the great circle into a circle and voting” are available in the literature (“Development research report on the work robot development commissioned in 1985” (Section 4.2.2.1) of the “Summary (Robot Engineering Research Association)”.
(Scan k (A-1-14))
(8) By the processing so far, one great circle in which the point at the position _i p ₀ has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n t _c . However, the great circle is converted by the equation (103).
(Scan i (A-1-15))
(9) In the processing up to this point, the height in the cross section circle of _n t _c, "all the points in the image _{{i p 0}"} is cross-ratio conversion and polar converted great circles group is drawn. Note that the above-described hardware for voting the polar transformation on the spherical surface to the “register arrangement corresponding to the inside of the circle” is Japanese Patent Publication No. 01-57831, Japanese Patent Publication No. 01-59619, Japanese Patent Publication No. 06. -70795 and Japanese Patent Publication No. 06-70796.
(Scanning _n t _c (A-1-16))
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n t c},} the "all points _{{i p} 0} in the _image" are cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed in all cross-sectional circles of the columnar arrangement.

(11) The “peak extraction unit 20” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized time _n t _c0 until crossing the plane” is obtained, and as the “coordinate within the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (A-1). -17). FIG. 11 shows the result of computer simulation performed using this flow.

(End)
Embodiment A-2. (Measurement of “normalized time _n t _c0 until crossing with three-dimensional orientation n _{s0 of} plane” without knowing moving direction v)
The posting measurement, that is, the method of 1.6 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG. 32, which is a modification of Embodiment A-1. The following steps (2) to (10) are the same as the corresponding steps in the embodiment A-1.

(Start)
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (from the lower limit value v _min to the upper limit value v _max ).
(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (A-2-3).

(2) The normalization time parameter _n t _c is scanned from the lower limit value _n t _{c, min} to the upper limit value _n t _{c, max} by the “ _n t _c parameter scanning unit 16” (A-2-4, A− 2-5, A-2-16).
(Scans _n t _c )
(3) Each address i in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-2-6, A-2-7, A-2-15).
(Scan i)
(4) The current and next time positions _i p ₀ and _i p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-2-8).

(5) The four parameters _n t _c , _i p ₀ , _i p ₁ , and p _{inf set above} are input to the “multi-ratio conversion unit 17”, and the position _{i pc} is output (A-2- 9).

(6) the position _i p _c of the poles converted to great circle on the sphere by the "polar transformation unit 18 '(A-2-10).
(Scan k)
(7) The point _k p _GC constituting the great circle is converted into “a point in the cross-sectional circle of height _n t _c ” of the “cylindrical array voting unit 19” and voted (A-2-13).

Scan k (A-2-14)
(8) By the processing so far, one great circle in which the point at the position _i p ₀ has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n t _c . However, the great circle is converted by the equation (103).
(Scan i (A-2-15))
(9) In the processing up to this point, the height in the cross section circle of _n t _c, "all the points in the image _{{i p 0}"} is cross-ratio conversion and polar converted great circles group is drawn.
(Scanning _n t _c (A-2-16))
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n t c},} the "all points _{{i p} 0} in the _image" are cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed in all cross-sectional circles of the columnar arrangement.
(Scanning v (A-2-17))
(11) In the processing so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all the moving direction parameters v”.

(12) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 20”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point at which the intensity reaches a peak in the array is extracted, the normalization time _n t _c0 until the plane is crossed as the “height coordinate” of the point is obtained, and the coordinate of the plane is obtained as the “coordinate in the cross-sectional circle”. A three-dimensional orientation _ns0 is obtained (A-2-18).

(End)
Embodiment A-3. (Measurement of three-dimensional orientation n _{s0 of} plane and normalized shortest distance _n d _s0 )
The posted measurement, that is, the method of 2.2.3 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG.
(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, the “p _inf setting unit 15” sets “infinite time position p _inf ” as being equal to the moving direction v (A-3-1).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (A-3-2, A- 3-3, A-3-15).
(Scans _n d _s )
(3) Each address i in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-3-4, A-3-5, A-3-14).
(Scan i)
(4) Output the current and next time positions _i p ₀ and _i p ₁ from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-3-6).

(5) The four parameters _n d _s , _i p ₀ , _i p ₁ , and p _{inf set above} are input to the “radius R calculation unit 23”, and the radius _i R and the position _i p ₀ are output. (A-3-7). Within that unit, the radius _i R is calculated according to the following equation using equation (33).

^{_{i R = cos -1 (n d}} s sin (i b- i a) / sin (i b)) (104)
(6) The radius _i R and the position _i p ₀ are input to the “small circle transform unit 24”, and the small circle transform—that is, the position _i p ₀ on the “spherical surface” with the radius _i R centered on it. To “small circle” in FIG. 17 (see 2.2.1 and FIG. 17A). The small circle conversion is performed in the following two steps in the unit (see FIG. 35A).

(A) Convert _i p ₀ to polar coordinates (A-3-8)
_i p ₀ in Cartesian coordinates and “Polar coordinates on sphere”
_i p ₀ = ( _i p _0X , _i p _0Y , _i p _0Z ) (105a)
_{= (I α 0, i β} 0) (105b)
And the polar coordinate components of _i p ₀ (longitude _i α ₀ and latitude _i β ₀ ) are calculated by the following equation.

_i α ₀ = tan ⁻¹ ( _i p _0Y / _i p _0X ) (105c)
^{_{i β 0 = tan -1 (√}} (i p 0X 2 + i p 0Y 2) / i p 0Z)) (105d)
Scan k (A-3-9, A-3-10, A-3-13)
(B) Small circle conversion (A-3-11)
A great circle on the sphere obtained by transforming _i p ₀ into a small circle--that is, the coordinates (longitude _k α _SC and latitude _k β _SC ) of an arbitrary point _k p _SC (whose address is k) constituting the small circle -Is determined by the coordinates _i α ₀ and _i β ₀ of _i p ₀ and the radius _i R, and is expressed by the equation (106b). This expression represents Expression (30) using the parameters shown in FIG.

_k p _SC = ( _k α _SC , _k β _SC ) (106a)
_{cos (k β SC) cos (} i β 0)
_{+ Sin (k β SC) sin} (i β 0) cos (k α SC - i α 0)
_{= Cos (i R) (106b} )
The specific calculation of the small circle is performed as follows by scanning k. Latitude _k β _SC is calculated as kΔβ _SC with Δβ _SC as the latitude resolution, and longitude _k α _SC is calculated by the following equation using the latitude.

_{k α SC = i α 0 +} cos -1 ((cos (i R) -cos (k β SC) cos (i β 0))
_{/ (Sin (k β SC)} sin (i β 0))) (106c)
(7) above the _k p _SC points constituting the small circle, in cross-section circle of height _n d _s of "columnar array voting unit 25", a polar coordinate (tilt _k alpha _{SC, Proj,} radial _k beta _{SC, Proj} ) is converted to a point _k p _{SC, Proj} expressed by voting (A-3-12)-the voting is performed by adding “brightness at position _i p ₀ ” (2.2. 3, see FIG. 17 (B) and FIG. 35 (B)). The conversion is expressed by the following equation using f () as a projection function.

_{k β SC, Proj = f (} k β SC) (107a)
_k α _{SC, Proj} = _k α _SC (107b)
In the case of the equidistant projection method, f () = 1, and _k β _{SC, Proj} is given by the following equation. (See Embodiment A-1 (7)).

_k β _{SC, Proj} = _k β _SC (107c)
Summarizing the above, the “point _k p _SC on the spherical surface” constituting the small circle is converted into the “point _k p _{SC, Proj} on the plane” in the cross-sectional circle, and the brightness of the position _i p ₀ is added to the point. Vote “Add”. Each cross-sectional circle can be realized by a register arrangement or a memory arrangement.
(Scan k (A-3-13))
(8) Through the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (107).
(Scan i (A-3-14))
(9) Through the processing so far, a small circle group in which “all the points in the image { _{ip 0} }” are transformed is drawn in the cross-sectional circle having the height of _n d _s .
(Scan _n d _s (A-3-15)
(10) In the process up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(11) “Peak extraction unit 26” extracts “a point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “a place where small circles intersect at one point” It is. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (A-3-16). FIG. 18 shows the result of the computer simulation performed using this flow.
(End)
Embodiment A-4. (Measured without knowing the direction of movement v of the "three-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane")
The posting measurement, that is, the method of 2.4 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG. 37 modified from the embodiment A-3. The following steps (2) to (10) are the same as the corresponding steps in the embodiment A-3.

(Start)
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (A-4-1, A-4-2, A-4-17).
(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (A-4-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (A-4-4, A -4-5, A-4-16).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-4-6, A-4-7, A-4-15).
(Scan i)
(4) Output the current and next time positions _i p ₀ and _i p ₁ from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-4-8).

(5) The four parameters _n d _s , _i p ₀ , _i p ₁ , and p _{inf set above} are input to the “radius R calculation unit 23”, and the radius _i R and the position _i p ₀ are output. (A-4-9).

(6) The radius _i R and the position _i p ₀ are input to the “small circle conversion unit 24” and converted into a “small circle on the sphere” having a radius _i R centered on the position _i p ₀ (A− 4-10).
(K is scanned A-4-11, A-4-12, A-4-14)
(7) The point _k p _SC constituting the small circle is converted into a “point in the cross-sectional circle of height _n d _s ” of the “cylindrical array voting unit” and voted (A-4-13).
(K is scanned A-4-14)
(8) Through the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (107).
(Scan i (A-4-15))
(9) Through the processing so far, a small circle group in which “all the points in the image { _{ip 0} }” are transformed is drawn in the cross-sectional circle having the height of _n d _s .
(Scan _{n d s (A-4-16)} )
(10) In the process up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed in all cross-sectional circles of the columnar arrangement.
(Scanning v (A-4-17))
(11) In the processing so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all the moving direction parameters v”.

(12) The “specific column array” having the maximum intensity is extracted by the “peak extraction unit 26” from the above-described column array group. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (A-4-18).
(End)
Embodiment A-5. (Measurement of “standardized distance _n d ₀ of points”)
The posting measurement, that is, the method of 3.4 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG.
(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, the “p _inf setting unit 15” sets “position p _inf after elapse of infinite time” as being equal to the moving direction v (A-5-1).

(2) The current and next time positions p ₀ and p ₁ are output from “image register 12 at current time t ₀ ” and “image register 13 at next time t ₁ ” (A-5-2).

(3) The three parameters p ₀ , p ₁ , and p _inf set above are input to the “point distance calculation unit 27”, and the normalized distance _n d ₀ to the point is output (A-5-5). 3). The calculation of _n d ₀ is performed by calculating equation (39b) within the unit.
(End)
Embodiment A-6. ("Measured normalization time _n t _c0 until crossing with three-dimensional orientation n _{s0 of} plane” by motion parallax τ)
In the above-described measurement, that is, in the method of 1.3.2, the case of using “multi-ratio conversion by motion parallax τ (formula 16b))” will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. The following steps (7) to (12) are the same as the corresponding steps in the embodiment A-1.

(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, the “p _inf setting unit 15” sets “infinite time position p _inf ” as being equal to the moving direction v (A-6-1).
(2) The normalization time parameter _n t _c is scanned from the lower limit value _n t _{c, min} to the upper limit value _n t _{c, max} by the “ _n t _c parameter scanning unit 16” (A-6-2, A- 6-3, A-6-16).
(Scans _n t _c )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-6-4, A-6-5, A-6-15).
(Scan i)
(4) The current and next time positions _i p ₀ and _i p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-6-6).

(5) The positions _i p ₀ and _i p ₁ at the current time and the next time are input to the “τ determination unit 28”, and the motion parallax _i τ (that is, _i p _{1 −i} p ₀ ) is output (A− 6-7). This motion parallax is a local motion described in “Prior art and problems to be solved by the invention”, and an algorithm for measuring the motion parallax and a method for realizing the same have been reported in detail (Japanese Patent Laid-Open No. 05-165958, special feature). Kaihei 05-165957, JP-A-06-044364, JP-A-09-081369; “Method of performing two-dimensional correlation and convolution one-dimensionally along the ρ coordinate of the Hough plane,” Kawakami and Okamoto , IEICE Technical Report, vol.IE 96-19, pp. 31-38, 1996; "A cell model for the detection of the local image motion on the magnecellular path of." motor, H., Vision Research, vol. 36, pp. 117-147, 1996).

(6) The four parameters _n t _c , _i τ, _i p ₀ , and p _inf set above are input to the “multi-ratio conversion unit 17” and the position _{i pc} is output. Calculation of the position _i p _c is in the unit takes place in the next two steps.

_(A) calculation of the center angle _i x and i _{p c} and _{p inf} (A-6-8)
_{From n} t _c , _i p ₀ , _i p ₁ , and p _inf , _i x is calculated by the following equation using equation (16b).

^{_{i x = tan -1 ((sin}} (i a) sin (i a + i τ))
_{/ (Cos (i a) sin} (i a + i τ) - n t c sin (i τ)))
(110a)
_(B) calculation of i _{p c} (A-6-9)
The position _i p _c on the spherical surface is calculated by the following equation using the above _{i x.}

_{i pc} = cos ( _ix ) Γ _x + sin ( _ix ) Γ _y (110b)
Here, Γ _x and Γ _y are calculated by the equation (100c) of the embodiment A-1.

(7) the position _i p _c of the poles converted to great circle on the sphere by the "polar transformation unit 18 '(A-6-10).
(Scanning k (A-6-11, A-6-12, A-6-14))
(8) The point _k p _GC constituting the great circle is converted into a “point in the cross-sectional circle of height _n t _c ” of the “cylindrical array voting unit 19” and voted (A-6-13).
(Scan k (A-6-14))
(9) In the process so far, one great circle in which the point at the position _i p ₀ has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n t _c . However, the great circle is converted by the equation (103).
(Scan i (A-6-15))
(10) With the processing so far, a great circle group in which “all points in the image { _{ip 0} }” are subjected to the cross ratio conversion and the polar conversion is drawn in the cross-sectional circle having the height of _n t _c .
(Scan _n nt _c (A-6-16))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n t c},} "all the points in the image _{{i p 0}"} is the cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(12) The “peak extraction unit 20” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized time _n t _c0 until crossing the plane” is obtained, and as the “coordinate within the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (A-6). -17).
(End)
Embodiment A-7. ("Standardized time _n t _c0 until crossing with the three-dimensional orientation n _{s0 of the} plane” is measured by the motion parallax τ without knowing the moving direction v)
The embodiment shown in FIG. 42 will be described in the case of using the above-described measurement, that is, “multi-ratio conversion by motion parallax τ (formula 16b))” in the method of 1.6. It is performed according to the following flow shown in FIG. 43, which is a modification of Embodiment A-6. The following steps (2) to (11) are the same as the corresponding steps in the embodiment A-6.

(Start)
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (A-7-1, A-7-2, A-7-18).
(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (A-7-3).

(2) The normalization time parameter _n t _c is scanned from the lower limit value _n t _{c, min} to the upper limit value _n t _{c, max} by the “ _n t _c parameter scanning unit 16” (A-7-4, A− 7-5, A-7-17).
(Scans _n t _c )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-7-6, A-7-7, A-7-16).
(Scan i)
(4) The current and next time positions _i p ₀ and _i p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-7-8).

(5) The positions _i p ₀ and _i p ₁ at the current time and the next time are input to the “τ determination unit 28”, and the motion parallax _i τ (that is, _i p _{1 −i} p ₀ ) is output (A− 7-9).

(6) The four parameters _n t _c , _i τ, _i p ₀ , and p _inf set above are input to the “multi-ratio conversion unit 17” and the position _{i pc} is output (A-7-10) ).

(7) the position _i p _c of the poles converted to great circle on the sphere by the "polar transformation unit 18 '(A-7-11).
(Scanning k (A-7-12, A-7-13, A-7-15))
(8) The point _k p _GC constituting the great circle is converted into “a point in the cross-sectional circle of height _n t _c ” of the “cylindrical array voting unit 19” and voted (A-7-14).
(Scan k (A-7-15))
(9) In the process so far, one great circle in which the point at the position _i p ₀ has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n t _c . However, the great circle is converted by the equation (103).
(Scan i (A-7-16))
(10) With the processing so far, a great circle group in which “all points in the image { _{ip 0} }” are subjected to the cross ratio conversion and the polar conversion is drawn in the cross-sectional circle having the height of _n t _c .
(Scan _n nt _c (A-7-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n t c},} "all the points in the image _{{i p 0}"} is the cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
(Scan v (A-7-18))
(12) In the process so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all moving direction parameters v”.

(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 20”. As the movement direction parameter corresponding to this array, the true movement direction v ₀ is obtained. Further, when a point at which the intensity reaches a peak in the array is extracted, the normalized time _n t _c0 until the plane is crossed as the “height coordinate” of the point, and the plane of the plane as the “coordinate in the cross-section circle”. A three-dimensional orientation _ns0 is obtained (A-7-19).
(End)
Embodiment A-8. ( "3-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane" measured by the motion parallax tau)
The case where “small circle transformation by motion parallax τ (formula 31)” is used in the above-described measurement, that is, the method of 2.2.3, will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. The following “(1) to (4)” and “(7) to (12)” are the same as the corresponding steps of the embodiment A-3, and (5) corresponds to the embodiment A-6. Same as step.

(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, the “p _inf setting unit 15” sets “position p _inf after elapse of infinite time” as being equal to the moving direction v (A-8-1).
(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (step A-8-2, A-8-3, A-8-15).
(Scans _n d _s )
(3) Scan each address i in the image from the lower limit i _min to the upper limit i _max (A-8-4, A-8-5, A-8-14).
(Scan i)
(4) Output the current and next time positions _i p ₀ and _i p ₁ from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-8-6).

(5) The positions _i p ₀ and _i p ₁ at the current time and the next time are input to the “τ determination unit 28”, and the motion parallax _i τ (that is, _i p _{1 −i} p ₀ ) is output (A− 8-7).

(6) The four parameters _n d _s , _i τ, _i p ₀ , and p _inf set above are input to the “radius R calculation unit 23”, and the radius _i R and the position _i p ₀ are output ( A-8-8). Within that unit, the radius _i R is calculated according to the following equation using equation (31).

^{_{i R = cos -1 (n d}} s sin (i τ) / sin (i a + i τ)) (115)
(7) The radius _i R and the position _i p ₀ are input to the “small circle conversion unit 24” and converted into a “small circle on the sphere” having a radius _i R centered on the position _i p ₀ (A− 8-9).
(Scanning k (A-8-10, A-8-11, A-8-13)).

(8) The point _k p _SC constituting the small circle is converted into a “point in the sectional circle of height _n d _s ” of the “cylindrical array voting unit 25” and voted (A-8-12).
(Scan k (A-8-13))
(9) With the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (107).
(Scan i (A-8-14))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p 0}"} has been converted is drawn.
(Scans _n d _s (A-8-15))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed in all cross-sectional circles of the columnar arrangement.

(12) The “peak extraction unit 26” extracts “the point where the voted intensity is the maximum (peak)” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point” It is. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (A-8-16). FIG. 18 shows the result of the computer simulation performed using this flow.
(End)
Embodiment A-9. (Without knowing the "3-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane" direction of movement v of and measured by motion parallax tau)
A case where “small circle transformation by motion parallax τ (formula 31)” is used in the above-described measurement, that is, the method of 2.5 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. 47, which is a modification of Embodiment A-8. The following steps (2) to (11) are the same as the corresponding steps in embodiment A-8.

(Start)
(0) “V-parameter scanning unit 21” scans “movement direction parameter v” in all possible directions (lower limit value v _min to upper limit value v _max ) (A-9-1, A -9-2, A-9-18).
(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” as being equal to this parameter v (A-9-3).

(2) The normalized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (A-9-4, A -9-5, A-9-17).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (A-9-6, A-9-7, A-9-16).
(Scan i)
(4) Output the current and next time positions _i p ₀ and _i p ₁ from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-9-8).

(5) The positions _i p ₀ and _i p ₁ at the current time and the next time are input to the “τ determination unit 28”, and the motion parallax _i τ (that is, _i p _{1 −i} p ₀ ) is output (A− 9-9).

(6) The four parameters _n d _s , _i τ, _i p ₀ , and p _inf set above are input to the “radius R calculation unit 23”, and the radius _i R and the position _i p ₀ are output ( A-9-10).

(7) The radius _i R and the position _i p ₀ are input to the “small circle conversion unit 24” and converted into a “small circle on the sphere” having a radius _i R centered on the position _i p ₀ (A− 9-11).
(Scanning k (A-9-12, A-9-13, A-9-15))
(8) The point _k p _SC constituting the small circle is converted into a “point in the cross-sectional circle of height _n d _s ” of the “cylindrical array voting unit 25” and voted (A-9-14).
(K is scanned A-9-15))
(9) With the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (107).
(Scan i (A-9-16))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p 0}"} has been converted is drawn.
(Scans _n d _s (A-9-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
(Scanning v (A-9-18))
(12) In the process so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all moving direction parameters v”.

(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 26”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the plane three-dimensional is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (A-9-19).
(End)
Embodiment A-10. (Measures “normalized distance _n d ₀ of point” by motion parallax τ)
The case of using the “measurement method by motion parallax τ (formula 39c))” in the above-described measurement, that is, the method of 3.4 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG.

(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, the “p _inf setting unit 15” sets “position p _inf after elapse of infinite time” as being equal to the moving direction v (A-10-1).
(2) The current and next time positions p ₀ and p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (A-10-2).

(3) The positions p ₀ and p ₁ at the current time and the next time are input to the “τ determination unit 28”, and the motion parallax τ (ie, p ₁ −p ₀ ) is output (A-10-3).

(4) The three parameters τ, p ₀ , and p _inf set above are input to the “point distance calculation unit 27”, and the normalized distance _n d ₀ to the point is output (A-10-4). ). The calculation of _n d ₀ is performed by calculating equation (39c) in the unit.
(End)
Embodiment B Binocular View Embodiment B-1. (Measurement of normalized distance _n d _c0 until crossing with plane three-dimensional direction n _s0 )
The posting measurement, that is, the method of 4.2.3 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG. Step (6) and subsequent steps are the same as those in the embodiment A-1.
(Start)
(1) The “visual axis direction a _xis connecting the left and right cameras” is generally known from the geometric arrangement of the stereo camera. The “p _axis setting unit 115” sets the “position p _axis on the visual axis” as being equal to the visual axis direction a _xis .

(2) The standardized distance parameter _n d _c is scanned from the lower limit value _n d _{c, min} to the upper limit value _n d _{c, max} by the “ _n d _c parameter scanning unit 116” (B-1-2, B− 1-3, B-1-16).
(Scans _n d _c )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-1-4, B-1-5, B-1-15).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-1-6).

(5) The four parameters _n d _c , _i p _R , _i p _L , and p _axis set as described above are input to the “multi-ratio conversion unit 117” and the position _{i pc} is output. Calculation of the position _i p _c is in the unit takes place in the next two steps.

_(A) calculation of the center angle _i x and i _{p c} and _{p axis} (B-1-7)
_From nd _c , _i p _R , _i p _L , and p _axis , _i x is calculated by the following formula using formula (58b) (see FIG. 23).

^{_{i x = tan -1 ((sin}} (i c) sin (i d))
_{/ (Cos (i c) sin} (i d) - n d c sin (i d) - (i c)))
(150a)
The _(b) i _p position _i p _c of the calculated sphere of _c, is calculated by the following equation using the above _i x.

_{i pc} = cos ( _ix ) Γ _x + sin ( _ix ) Γ _y (150b)
Here, Γ _x and Γ _y are calculated by the following equations using [×] and || as vector outer product operations and absolute value operations, respectively.

Γ _x = a _xis (150c)
Γ _z = [a _xis x _i p _R ] / | [a _xis x _i p _R ] |
Γ _y = [Γ _z × Γ _x ]
(6) the position _i p _c of the pole into a great circle on the sphere by the "polar transformation unit 118 '(see FIG. 24 (A)). The polar conversion is performed in the following two steps in the unit (see FIG. 30A).

(A) Converting _{i pc} to polar coordinates (B-1-9)
_{i pc} in Cartesian coordinates and "polar coordinates on a sphere"
_{i p c = (i p cX} , i p cY, i p cZ) (151a)
_{= (I α c, i β} c) (151b)
It is expressed _as, calculates polar components i _{p c} a (longitude _i alpha _c and latitude _i beta _c) by the following equation.

^{_{i α c = tan -1 (i}} p cY / i p cX) (151c)
^{_{i β c = tan -1 (√}} (i p cX 2 + i p cY 2) / i p cZ)) (151d)
(Scan k (B-1-10, B-1-11, B-1-14))
_(B) polar transformation of i _{p c} (B-1-12)
_i p great circle on the sphere that polar transformation a _c - i.e., any point _k p _GC constituting the great circle coordinates (the address to k) (longitude _k alpha _GC, latitude _k beta _GC) - - is determined by the coordinates _{i α c, i β c} of _i p _c, the formula (152 b).

_{k p GC = (k α GC} , k β GC) (152a)
_{cos (k α GC - i α} c) = - cot (k β GC) cot (i β c) (152b)
A specific calculation of the great circle is performed as follows by scanning k from the lower limit value k _min to the upper limit value k _max . The latitude _k β _GC is calculated as kΔβ _GC using Δβ _GC as the latitude resolution, and the longitude _k α _GC is calculated by the following equation using the latitude.

_{k α GC = i α c +} cos -1 (cot (k β GC) cot (i β c)) (152c)
(7) a _k p _GC points constituting the great circle, in cross-section circle of height _n d _c of the "cylindrical array voting unit 119", a polar coordinate (tilt _k alpha _{GC, Proj,} radial _k beta _{GC, Proj} ) is converted into a point _k p _{GC, Proj} represented by voting (B-1-13). The transformation is generally f () as a projection function,
_{k β GC, Proj = f (} k β GC) (153a)
_k α _{GC, Proj} = _k α _GC (153b)
It is represented by In the case of the equidistant projection method, f () = 1, and _k β _{GC, Proj} is given by the following equation.

_k β _{GC, Proj} = _k β _GC (153c)
Summarizing the above, the “point _k p _GC on the spherical surface” constituting the great circle is converted into the “point _k p _{GC, Proj} on the plane” in the cross-sectional circle, and the brightness of the position _i p _R is added to the point. Vote “Add”. Each cross-sectional circle can be realized by a register arrangement or a memory arrangement.
(Scan k (B-1-14))
(8) With the processing so far, one great circle whose point _i p _R has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n d _c . However, the great circle is converted by the equation (153).
(Scan i (B-1-15))
(9) In the processing up to this point, the height in the cross section circle of _n d _c, "all the points in the image _{{i p R}"} is cross-ratio conversion and polar converted great circles group is drawn.
(Scan _{n d c (B-1-16)} )
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n d c},} "all the points in the image _{{i p R}"} is the cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed in all cross-sectional circles of the columnar arrangement.

(11) The “peak extraction unit 120” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized distance _n d _c0 until crossing the plane” is obtained, and as the “coordinate in the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (B-1). -17).
(End)
Embodiment B-2. (Measured "3D orientation _{n s0} and the normalized distance _n d _c0 plane" without knowing the visual axis direction _{a xis)}
The posted measurement, that is, the method of 4.2.5 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG. 53, which is a modification of Embodiment B-1. The following steps (2) to (10) are the same as the corresponding steps in the embodiment B-1.

(Start)
(0) “ _Axis parameter scanning unit 121” scans the “visual axis direction parameter a _xis ” in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ). (B-2-1, B-2-2, B-2-17).
( _{Scan axis} )
(1) by _{'p axis} setting unit 115 ", the" viewing position _{p axis} on the axis "is set as equal to the parameter _{a xis} (B-2-3).

(2) The standardized distance parameter _n d _c is scanned from the lower limit value _n d _{c, min} to the upper limit value _n d _{c, max} by the “ _n d _c parameter scanning unit 116” (B-2-4, B− 2-7, B-2-16).
(Scans _n d _c )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-2-6, B-2-7, B-2-15).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-2-8).

(5) The four parameters _n d _c , _i p _R , _i p _L , and p _axis set as described above are input to the “multi-ratio conversion unit 117”, and the position _{i pc} is output (B-2- 9).

(6) the position _i p _c of the poles converted to great circle on the sphere by the "polar transformation unit 118" (B-2-10).
(Scanning k (B-2-11, B-2-12, B-2-14))
(7) a _k p _GC points constituting the great circle, to vote converted to "points in circular section of height _n d _c" of the "cylindrical array voting unit 119" (B-2-13).

(8) With the processing so far, one great circle whose point _i p _R has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n d _c . However, the great circle is converted by the equation (153).
(Scan i (B-2-15))
(9) In the processing up to this point, the height in the cross section circle of _n d _c, "all the points in the image _{{i p R}"} is cross-ratio conversion and polar converted great circles group is drawn.
(Scan _{n d c (B-2-16)} )
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n d c},} "all the points in the image _{{i p R}"} is the cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed in all cross-sectional circles of the columnar arrangement.
( _{Scan axis} (B-2-17))
(11) In the processing so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all visual axis direction parameters a _xis ”.

(12) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 120”. As the visual axis direction parameter corresponding to this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized distance _n d _c0 until the plane is crossed as the “height coordinate” of the point, and the plane coordinate as the “coordinate in the cross-sectional circle” is obtained. A three-dimensional orientation _ns0 is obtained (B-2-18).
(End)
Embodiment B-3. (Measurement of three-dimensional orientation n _{s0 of} plane and normalized shortest distance _n d _s0 )
The measurement of posting, that is, the method of 4.3.2 will be described in the embodiment of FIG. This is performed in the following flow shown in FIG. Step (7) and subsequent steps are the same as those in the embodiment A-3.
(Start)
(1) With the “p _axis setting unit 115”, the position p _axis on the visual _axis is set to be equal to the visual axis direction a _xis (see Embodiment B-1 (B-3-1)).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (B-3-2, B- 3-3, B-3-15).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-3-4, B-3-5, B-3-14).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-3-6).

(5) The four parameters _n d _s , _i p _R , _i p _L , and p _axis set as described above are input to the “radius R calculation unit 123”, and the radius _i R and the position _i p _R are output. (B-3-7). Within the unit, the radius _i R is calculated by the following formula by the formula (66c).

^{_{i R = cos -1 (n d}} s sin (i d- i c) / sin (i d)) (154)
(6) above the radius _i R and the position _i p _R enter the "small circle conversion unit 124", the small circle conversion - that is, the "sphere of radius _i R centered on it the position _i p _R To “small circle of” (refer to 4.3.2 and FIG. 27A). The small circle conversion is performed in the following two steps in the unit (see FIG. 56A).

(A) Convert _i p _R to polar coordinates (B-3-8)
_i p _R in Cartesian coordinates and “polar coordinates on a sphere”
_i p _R = ( _i p _RX , _i p _RY , _i p _RZ ) (155a)
_{= (I α R, i β} R) (155b)
And the polar coordinate components (longitude _i α _R and latitude _i β _R ) of _i p _R are calculated by the following equations.

_i α ₀ = tan ⁻¹ ( _i p _RY / _i p _RX ) (155c)
^{_{i β 0 = tan -1 (√}} (i p RX 2 + i p RY 2) / i p RZ) (155d)
(Scan k)
(B) Small circle conversion (B-3-11)
_A small circle on the sphere obtained by transforming _i p _R into a small circle--that is, the coordinates (longitude _k α _SC , latitude _k β _SC ) of an arbitrary point _k p _SC (whose address is k) constituting the small circle -Is determined by the coordinates _i α _R and _i β _R of the _i p _R and the radius _i R and is expressed by the equation (156b). This expression represents Expression (30) using the parameters shown in FIG.

_k p _SC = ( _k α _SC , _k β _SC ) (156a)
_{cos (k β SC) cos (} i β R)
_{+ Sin (k β SC) sin} (i β R) cos (k α SC - i α R)
_{= Cos (i R) (156b} )
The specific calculation of the small circle is performed as follows by scanning k. Latitude _k β _SC is calculated as kΔβ _SC with Δβ _SC as the latitude resolution, and longitude _k α _SC is calculated by the following equation using the latitude.

_{k α SC = i α R +} cos -1 ((cos (i R) -cos (k β SC) cos (i β R))
_{/ (Sin (k β SC)} sin (i β R))) (156c)
(7) above the _k p _SC points constituting the small circle, in cross-section circle of height _n d _s of "columnar array voting unit 125", a polar coordinate (tilt _k alpha _{SC, Proj,} radial _k beta _{SC, Proj} ) is converted to a point _k p _{SC, Proj} represented by voting (B-3-12). The conversion is expressed by the following equation using f () as a projection function.

_{k β SC, Proj = f (} k β SC) (157a)
_k α _{SC, Proj} = _k α _SC (157b)
In the case of the equidistant projection method, f () = 1, and _k β _{SC, Proj} is given by the following equation.

_k β _{SC, Proj} = _k β _SC (157c)
To summarize the above, the “point _k p _SC on the spherical surface” forming the small circle is converted into the “point _k p _{SC, Proj} on the plane” in the cross-sectional circle, and the brightness of the position _i p _R is added to the point. Vote “Add”. Each cross-sectional circle can be realized by a register arrangement or a memory arrangement.
(Scan k (B-3-13))
(8) With the processing so far, one small circle whose point _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (157).
(Scan i (B-3-14))
(9) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scanning _n d _s (B-3-15)
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(11) The “peak extraction unit 126” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point” It is. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-section circle” (B-3-16).
(End)
Embodiment B-4. (Measurement "3-dimensional orientation _{n s0} plane and the normalized shortest distance _n d _s0" a without knowing the visual axis direction _{a xis)}
The measurement of the posting, that is, the method of 4.3.3 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG. 58 obtained by modifying Embodiment B-3. The following steps (2) to (10) are the same as the corresponding steps in embodiment B-3.

(Start)
(0) The “visual axis parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scan unit 121”. (B-4-1, B-4-2, B-4-17).
( _{Scan axis} )
(1) The “p _axis setting unit 115” sets the “position p _axis on the visual _axis ” to be equal to the parameter a _xis (B-4-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (B-4-4, B -4-5, B-4-16).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-4-6, B-4-7, B-4-15).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-4-8).

(5) The four parameters _n d _s , _i p _R , _i p _L , and p _axis set as described above are input to the “radius R calculation unit 123”, and the radius _i R and the position _i p _R are output. (B-4-9).

(6) The radius _i R and the position _i p _R are input to the “small circle conversion unit 124” and converted into a “small circle on the sphere” having a radius _i R centered on the position _i p _R (B− 4-10).
(Scan k (B-4-11, B-4-12, B-4-14))
(7) The point _k p _SC constituting the small circle is converted into a “point in the sectional circle of height _n d _s ” of the “cylindrical array voting unit 125” and voted (B-4-13).
(Scanning k (B-4-14))
(8) With the processing so far, one small circle whose point _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (157).
(Scan i (B-4-15))
(9) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scan _{n d s (B-4-16)} )
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed in all cross-sectional circles of the columnar arrangement.
( _{Scan axis} (B-4-17))
(11) In the processing so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all visual axis direction parameters a _xis ”.

(12) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 126”. As the visual axis direction parameter for this array, the true movement direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the plane three-dimensional is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (B-4-18).
(End)
Embodiment B-5. (Measurement of “standardized distance _n d ₀ of points”)
The posted measurement, that is, the method of 4.4.3 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG.
(Start)
(1) The “p _axis setting unit 115” sets “position p _axis on the visual _axis ” as being equal to the visual axis direction a _xis (see Embodiment B-1 (B-5-1)).

(2) The positions p _R and p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-5-2).

(3) The three parameters p _R , p _L , and p _axis set above are input to the “point distance calculation unit 127”, and the normalized distance _n d ₀ to the point is output (B-5-5). 3). The calculation of _n d ₀ is performed by calculating Expression (67b) in the unit.
(End)
Embodiment B-6. (Measured by binocular parallax σ "3-dimensional orientation _{n s0} and the normalized distance _n d _c0 plane")
The case where “the double ratio conversion by binocular parallax σ (formula 60b)” is used in the above-described measurement, that is, the method of 4.2.3 will be described with reference to the embodiment of FIG. This is performed according to the following flow shown in FIG. The following steps (7) to (12) are the same as the corresponding steps in embodiment B-1.

(Start)
(1) With the “p _axis setting unit 115”, the position p _axis on the visual _axis is set equal to the visual axis direction a _xis (see Embodiment B-1).

(2) The normalized distance parameter _n d _c is scanned from the lower limit value _n d _{c, min} to the upper limit value _n d _{c, max} by the “ _n d _c parameter scanning unit 116” (B-6-2, B- 6-3, B-6-16).
(Scans _n d _c )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-6-4, B-6-5, B-6-15).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-6-6).

(5) The positions _i p _R and _i p _L of the left and right cameras are input to the “σ determination unit 128”, and the binocular parallax _i σ (that is, _i p _{L −i} p _R ) is output (B-6) -7). An algorithm for measuring the binocular parallax and a method for realizing the algorithm have been reported (Japanese Patent Laid-Open Nos. 05-165957, 05-165957, 06-04364, and 09-081369). “A method for performing two-dimensional correlation and convolution one-dimensionally along the ρ coordinate of the Hough plane,” Kawakami and Okamoto, Shingaku Giho, vol. IE 96-19, pp. 31-38, 1996).

(6) The four parameters _n d _c , _i σ, _i p _R , and p _axis set as described above are input to the “multi-ratio conversion unit 117”, and the position _{i pc} is output. Calculation of the position _i p _c is in the unit takes place in the next two steps.

_(A) Calculation of the center angle _i x and i _{p c} and _{p axis} (B-6-8)
_From nd _c , _i σ, _i p _R , and p _axis , _i x is calculated by the following equation using equation (60b)

^{_{i x = tan -1 ((sin}} (i c) sin (i c + i σ))
_{/ (Cos (i c) sin} (i c + i σ) - n d c sin (i σ)))
(160a)
_(B) calculation of i _{p c} (B-6-9)
The position _i p _c on the spherical surface is calculated by the following equation using the above _{i x.}

_{i pc} = cos ( _ix ) Γ _x + sin ( _ix ) Γ _y (160b)
Here, Γ _x and Γ _y are calculated by the equation (150c).

(7) poles into a great circle on the sphere position _i p _c of the by 'transform unit 118 "(B-6-10).
(Scanning k (B-6-11, B-6-12, B-6-14))
(8) a _k p _GC points constituting the great circle, to vote converted to "points in circular section of height _n d _c" of the "cylindrical array voting unit 119" (B-6-13).
(Scan k (B-6-14))
(9) Through the processing so far, one great circle whose point _i p _R has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n d _c . However, the great circle is converted by the equation (153).
(Scan i (B-6-15))
(10) In the processing up to this point, the height in the cross section circle of _n d _c, "all the points in the image _{{i p R}"} is cross-ratio conversion and polar converted great circles group is drawn.
(Scan _{n d c (B-6-16)} )
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d c},} the "all points _{{i p} R} in the _image" are cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(12) “Peak extraction unit 120” extracts “the point where the voted intensity becomes the maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized distance _n d _c0 until crossing the plane” is obtained, and as the “coordinate in the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (B-6). -17).
(End)
Embodiment B-7. (“Normalized distance _n d _c0 until crossing with the three-dimensional orientation n _{s0 of the} plane” is measured by the binocular parallax σ without knowing the visual axis direction a _xis )
In the case of using the above-described measurement, that is, the method of 4.2.5, “multi-ratio conversion by binocular parallax σ (formula 60b))” will be described with reference to the embodiment of FIG. This is performed according to the following flow shown in FIG. 64 obtained by modifying the embodiment B-6. The following steps (2) to (11) are the same as the corresponding steps in the embodiment B-6.

(Start)
(0) “ _Axis parameter scanning unit 121” scans the “visual axis direction parameter a _xis ” in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ). (B-7-1, B-7-2, B-7-18).
( _{Scan axis} )
(1) The “position on the visual axis p _axis ” is set to be equal to the parameter a _xis by the “p _axis setting unit 115” (B-7-3).

(2) The standardized distance parameter _n d _c is scanned from the lower limit value _n d _{c, min} to the upper limit value _n d _{c, max} by the “ _n d _c parameter scanning unit 116” (B-7-4, B- 7-5, B-7-17).
(Scans _n d _c )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-7-6, B-7-7, B-7-16).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-7-8).

(5) The positions _i p _R and _i p _L of the left and right cameras are input to the “σ determination unit 128”, and the binocular parallax _i σ (that is, _i p _L − _i p _R ) is output (B-7) -9).

(6) The four parameters _n d _c , _i τ, _i p _R , and p _axis set as described above are input to the “multi-ratio conversion unit 117”, and the position _{i pc} is output (B-7-10). ).

(7) the position _i p _c of the poles converted to great circle on the sphere by the "polar transformation unit 118" (B-7-11).
(Scan k B-7-12, B-7-13, B-7-15))
(8) a _k p _GC points constituting the great circle, to vote converted to "points in circular section of height _n d _c" of the "cylindrical array voting unit" (B-7 - 14).
(K is scanned B-7-15)
(9) Through the processing so far, one great circle whose point _i p _R has been “multi-ratio transformed and pole transformed” is drawn in a cross-sectional circle having a height of _n d _c . However, the great circle is converted by the equation (153).
(Scan i (B-7-16))
(10) In the processing up to this point, the height in the cross section circle of _n d _c, "all the points in the image _{{i p R}"} is cross-ratio conversion and polar converted great circles group is drawn.
(Scan _{n d c (B-7-17)} )
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d c},} the "all points _{{i p} R} in the _image" are cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
( _{Scan axis} (B-7-18))
(12) by the processing up to here, the vote total cross-sectional circle of the "cylinder arrangement group for all of the visual axis direction parameter a _xis" is performed.

(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 120”. As the visual axis direction parameter corresponding to this arrangement, the true visual axis direction _axis0 is obtained. In addition, when a point where the intensity reaches a peak in the array is extracted, the normalized distance _n d _c0 until the plane is crossed as the “height coordinate” of the point is obtained, and the coordinate of the plane is obtained as “coordinate in the cross-sectional circle”. A three-dimensional orientation _ns0 is obtained (B-7-19).
(End)
Embodiment B-8. (Measured by binocular parallax σ "3-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane")
A case where “small circle transformation by binocular parallax σ (formula (66a))” is used in the above-described measurement, that is, the method of 4.3.2 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. The following “(1) to (4)” and “(7) to (12)” are the same as the corresponding steps of the embodiment B-3, and (5) corresponds to the embodiment B-6. Same as step.

(Start)
(1) With the “p _axis setting unit 115”, the position p _axis on the visual _axis is set to be equal to the visual axis direction a _xis (see Embodiment B-1) (B-8-1).

(2) The normalized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (B-8-2, B -8-3, B-8-15).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-8-4, B-8-5, B-8-14).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-8-6).

(5) The positions _i p _R and _i p _L of the left and right cameras are input to the “σ determination unit 128”, and the binocular parallax _i σ (that is, _i p _{L −i} p _R ) is output (B-8) -7).

(6) The four parameters _n d _s , _i σ, _i p _R , and p _axis set as described above are input to the “radius R calculation unit 123”, and the radius _i R and the position _i p _R are output ( B-8-8). Within the unit, the radius _i R is calculated by the following equation by the equation (66a).

^{_{i R = cos -1 (n d}} s sin i σ / sin (i c + i σ)) (115)
(7) The radius _i R and the position _i p _R are input to the “small circle conversion unit 124” and converted into a “small circle on the sphere” having a radius _i R centered on the position _i p _R (B− 8-9).
(Scanning k (B-8-10, B-8-11, B-8-13))
(8) The point _k p _SC constituting the small circle is converted into a “point in the cross-sectional circle of height _n d _s ” of the “cylindrical array voting unit 125” and voted (B-8-12).
(Scan k (B-8-13)
(9) Through the processing so far, one small circle in which the point at the position _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (157).
(Scan i (B-8-14))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scans _n d _s (B-8-15))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed in all cross-sectional circles of the columnar arrangement.

(12) The “peak extraction unit 126” extracts “the point where the voted intensity is the maximum (peak)” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point” It is. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-section circle” (B-8-16).
(End)
Embodiment B-9. (Measured "3D orientation _{n s0} and the normalized shortest distance _n d _s0 plane" without knowing the visual axis direction _{a xis,} and the binocular parallax sigma)
A case where “small circle transformation by binocular parallax σ (formula (66a))” is used in the above-described measurement, that is, the method of 4.3.3 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. 68 obtained by modifying Embodiment B-8. The following steps (2) to (11) are the same as the corresponding steps in embodiment B-8.

(Start)
(0) “ _Axis parameter scanning unit 121” scans the “visual axis direction parameter a _xis ” in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ). .
( _{Scan axis} )
(1) The “p _axis setting unit 115” sets the “position p _axis on the visual _axis ” to be equal to the parameter a _xis (B-9-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (B-9-4, B -9-5, B-9-17).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (B-9-6, B-9-7, B-9-16).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-9-8).

(5) The positions _i p _R and _i p _L of the left and right cameras are input to the “σ determination unit 128”, and the binocular parallax _i σ (that is, _i p _{L −i} p _R ) is output (B-9) -9).

(6) The four parameters _n d _s , _i τ, _i p _R , and p _axis set as described above are input to the “radius R calculation unit 123”, and the radius _i R and the position _i p _R are output ( B-9-10).

(7) The radius _i R and the position _i p _R are input to the “small circle conversion unit 124” and converted into a “small circle on the spherical surface” having the radius _i R centered on the position _i p _R.
(Scan k)
(8) The point _k p _SC constituting the small circle is converted into a “point in the sectional circle of height _n d _s ” of the “cylindrical array voting unit 125” and voted (B-9-14).
(Scan k (B-9-15))
(9) Through the processing so far, one small circle in which the point at the position _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (157).
(Scan i (B-9-16))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scans _n d _s (B-9-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
( _{Scan axis} (B-9-18))
(12) by the processing up to here, the vote total cross-sectional circle of the "cylinder arrangement group for all of the visual axis direction parameter a _xis" is performed.

(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 126”. As the visual axis direction parameter corresponding to this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (B-9-19).
(End)
Embodiment B-10. (Measures “normalized distance _n d ₀ of point” by binocular parallax σ)
A case where the “measurement method using binocular parallax σ (equation (67c))” is used in the above-described measurement, that is, the method of 4.4.3 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG.

(Start)
(1) The position p _axis on the visual _axis is set to be equal to the visual axis direction a _xis by the “p _axis setting unit 115” (see Embodiment B-1) (B-10-1).

(2) The positions p _R and p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (B-10-2).

(3) The positions p _R and p _L of the left and right cameras are input to the “σ determination unit 128”, and binocular parallax σ (ie, p _L −p _R ) is output (B-10-3).

(4) The three parameters τ, p _R and p _axis set as described above are input to the “point distance calculation unit 127”, and the normalized distance _n d ₀ to the point is output (B-10-4). ). The calculation of _n d ₀ is performed by calculating Expression (67c) in the unit.
(End)
Embodiment C-1. ( "3-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane" measured by the motion parallax tau)
The embodiment shown in FIG. 71 will be described with respect to the above-described measurement, that is, the case of using the motion parallax τ in the method of 2.1. It is performed according to the following flow shown in FIG.

(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, “p _inf setting unit 15” sets “position p _inf after elapse of infinite time” as being equal to the moving direction v (C-1-1).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-1-2, C -1-3, C-1-14).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-1-4, C-1-5, C-1-13).
(Scan i)
(4) The current and next time positions _i p ₀ and _i p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (C-1-6).

(5) The current and next time positions _i p ₀ and _i p ₁ are input to the “τ determination unit 28”, and the motion parallax _i τ (ie, the same as step (5) of the embodiment A-6) _{, i p 1 - i p 0} ) and outputs the (C-1-7).

(6) Consider a circle with a radius r centered in the moving direction v (FIG. 13), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-1-8, C-1). -9, C-1-12).
(Scan r)
(7) over five parameters set by the _{n d s, i p 0,} p inf, i τ, type r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" Te, 2.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _{s- (C-1-10).} This intersection is a component of “a small circle with radius _i R centered at _i p ₀ ” as proved in 2.2.1. This point ^{_{i r n s +, i r}} n s- for "polar coordinates on the sphere (longitude ^{_{i r α s +, i r}} α s-, latitude ^{_{i r β s +, i r}} β s-)" has the formula (29) It is calculated by the following formula.

ⁱ _r β _{s +} = r
ⁱ _r β _s− = r (200a)
_{^{i r α s + = i α}} a + cos -1 (((n d s sin (i τ) / sin (i a + i τ))
_{-Cos (r) cos (i a} )) / (sin (r) -sin (i a)))
_{^{i r α s- = i α a}} -cos -1 (((n d s sin (i τ) / sin (i a + i τ))
_{-Cos (r) cos (i a} )) / (sin (r) -sin (i a)))
(200b)
Here, _i α _a and _i a are the longitude and latitude coordinates of _i p ₀ (see FIG. 14).

(8) ⁱ _r n _s + points constituting the small circle ^above, _i r _n the _s-, in cross section circle of height _n d _s of "columnar array voting unit 224", a polar coordinate (inclination ⁱ _r ^α _{s +, ^{proj, i r α s-, proj}} , + radius vector _{^{i r β s, proj, i}} r β s-, i r n s + a point represented by _{^{proj), proj, i r n}} s-, converted to _proj Vote (C-1-11) —The vote is made by adding “brightness at position _i p ₀ ”. The transformation is generally f () as a projection function,
_{^{i r β s +, Proj =}} f (i r β s +)
_{^{i r β s-, Proj = f}} (i r β s-) (201a)
_{^{i r α s +, Proj =}} i r α s +
_{^{i r α s-, Proj = i}} r α s- (201b)
(See step (7) of Embodiment A-1). In the case of equidistant projection method is _{^{f () = 1, i r}} β s +, Proj, i r β s-, Proj is given by the following equation.

^{_{i r β s +, Proj =}} i r β s +
^{_{i r β s-, Proj = i}} r β s- (201c)
In summary of the above, it constitutes a small circle "points on the sphere ^{_{i r n s +, i r}} n s-" points on "plane in circular cross ^{_{i r n s +, Proj,}} i r n s-, Proj And vote (add) “brightness at position _i p ₀ ” to that point. Each cross-sectional circle can be realized by a register arrangement or a memory arrangement.
(Scan r (C-1-12))
(9) With the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-1-13))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p 0}"} has been converted is drawn.
(Scans _n d _s (C-1-14))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(12) The “peak extraction unit 225” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array (C-1-15) —this maximum point is the “small circle group” Is a place that intersects one point. As the “height coordinate” of the local maximum point, “the standardized shortest distance _n d _s of the plane” is obtained, and as the “coordinate in the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained.
(End)
Embodiment C-2. (Without knowing the "3-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane" direction of movement v of and measured by motion parallax tau)
The embodiment shown in FIG. 73 will explain a case where the above-described measurement, that is, “method of measuring without knowing the moving direction v (2.5)” is performed with respect to 2.1 and using motion parallax τ. It is performed according to the following flow shown in FIG. The following steps (2) to (11) are the same as the corresponding steps in the embodiment C-1.

(Start)
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (lower limit value v _min to upper limit value v _max ) (C-2-1, C 2-2, C-2-17).
(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (C-2-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-2-4, C -2-5, C-2-16).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-2-6, C-2-7, C-2-15).
(Scan i)
(4) Output the current and next time positions _i p ₀ and _i p ₁ from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (C-2-8).

(5) The positions _i p ₀ and _i p ₁ at the current time and the next time are input to the “τ determination unit 28”, and the motion parallax _i τ (ie, the same as step (5) in the embodiment A-5) _{, i p 1 - i p 0} ) and outputs the (C-2-9).

(6) Consider a circle with a radius r centered in the moving direction v (FIG. 13), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-2-10, C-2). -11, C-2-14).
(Scan r)
(7) over five parameters set by the _{n d s, i p 0,} p inf, i τ, type r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" Te, 2.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _s-(C-2-12).

(8) the small circle point constituting the ^{_{i r n s +, i r}} n the _s-, vote and converts the "point in the cross-sectional circle of height _n d _s" of the "cylindrical array voting unit 224" ( C-2-13).
(Scan r (C-2-14))
(9) With the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-2-15))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p 0}"} has been converted is drawn.
(Scans _n d _s (C-2-16))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
(Scan v (C-2-17))
(12) In the process so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all moving direction parameters v”.

(13) From the columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 225”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (C-2-18).
(End)
Embodiment C-3. ("Measures the three-dimensional orientation _{s0 of the} plane and the normalized shortest distance _n d _s0 " by binocular parallax σ)
The case of the above-described measurement, that is, the case of binocular parallax σ in the method of 4.3.1 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. After step (8), if _i p _R and { _i p _R } are replaced with _i p ₀ and { _i p ₀ }, they are the same as in the embodiment C-1.

(Start)
(1) With the “p _axis setting unit 115”, the position p _axis on the visual _axis is set to be equal to the visual axis direction a _xis (see step (1) in Embodiment B-1).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-3-2, C -3-3, C-3-14).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-3-4, C-3-5, C-3-13).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (C-3-6).

(5) The positions _i p _R and _i p _L of the left and right cameras are input to the “σ determination unit 128”, and binocular parallax _i σ (ie, _i p _L − _i p _R ) is output (embodiment B) Refer to step (5) of -6 (C-3-7)).

(6) _Consider a _circle with a radius r centered on the visual axis direction a _xis (FIG. 26), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-3-8, C -3-9, C-3-12).
(Scan r)
(7) is set above the five parameters _{n d s, i p R,} p axis, i σ, type r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" Te, 4.3.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _s-(C-3-10). This intersection is a component of “a small circle with radius _iR centered on _i p _R ” as proved in 4.3.2. This point ^{_{i r n s +, i r}} n s- for "polar coordinates on the sphere (longitude ^{_{i r α +, i r α}} -, latitude ^{_{i r β +, i r β}} -)" is 4.3.1 ( It is calculated by the following formula as described in 4).

ⁱ _r β ₊ = r
ⁱ _r β ₋ = r (210a)
_{^{i r α s + = i α}} c + cos -1 (((n d s sin (i σ) / sin (i c + i σ))
_{-Cos (r) cos (i c} )) / (sin (r) -sin (i c)))
_{^{i r α s- = i α c}} -cos -1 (((n d s sin (i σ) / sin (i c + i σ))
_{-Cos (r) cos (i c} )) / (sin (r) -sin (i c)))
(210b)
(8) The small circle point constituting the _{ⁱ r} n _s ^{+, i} and r _{n s-,} vote and converts the "columnar array voting unit 224" of the "height in the cross section circle of _n d _s" (C- 3-11).
(Scan r (C-3-12))
(9) Through the processing so far, one small circle in which the point at the position _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-3-13))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scans _n d _s (C-3-14))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(12) The “peak extraction unit 225” extracts “the point where the voted intensity becomes the maximum (peak)” in the cylindrical array (C-3-15) —this maximum point is the “small circle group” Is a place that intersects a single point. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle”.
(End)
Embodiment C-4. (Measured "3D orientation _{n s0} and the normalized shortest distance _n d _s0 plane" without knowing the visual axis direction _{a xis,} and the binocular parallax sigma)
The embodiment shown in FIG. 77 is a case where the above-described measurement, that is, “method of measuring without knowing the visual axis direction a _xis (4.3.3)” is performed with respect to 4.3.1 and using binocular parallax σ. explain. It is performed according to the following flow shown in FIG. 78, which is a modification of Embodiment C-3. The following steps (1) to (11) are the same as the corresponding steps in the embodiment C-3.

(Start)
(0) The “visual axis parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scan unit 121”. (C-4-1, C-4-2, C-4-17).
( _{Scan axis} )
(1) The “position on the visual axis p _axis ” is set to be equal to the parameter a _xis by the “p _axis setting unit 115” (C-4-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-4-4, C -4-5, C-4-16).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-4-6, C-4-7, C-4-15).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (C-4-8).

(5) The positions _i p _R and _i p _L at the left and right cameras are input to the “σ determination unit 128”, and the binocular parallax _i σ (that is, _i p _{L −i} p _R ) is output (C-4). -9).

(6) _Consider a _circle with a radius r centered on the visual axis direction a _xis (FIG. 26), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-4-10, C -4-11, C-4-14).
(Scan r)
(7) is set above the five parameters _{n d s, i p R,} p axis, i σ, type r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" Te, 4.3.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _s-(C-4-12).

(8) the small circle point constituting the ^{_{i r n s +, i r}} n the _s-, vote and converts to a point in cross section circle of height _n d _s of "columnar array voting unit 224""(C -4-13).
(Scan r (C-4-14))
(9) Through the processing so far, one small circle in which the point at the position _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-4-15))
(10) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scans _n d _s (C-4-16))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
( _{Scan axis} (C-4-17))
(12) by the processing up to here, the vote total cross-sectional circle of the "cylinder arrangement group for all of the visual axis direction parameter a _xis" is performed.

(13) In the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 225”. As the visual axis direction parameter for this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the plane three-dimensional is used as the “coordinate in the cross-sectional circle”. The direction n _s0 is obtained. (C-4-18).
(End)
Embodiment C-5. (Measurement of three-dimensional orientation n _{s0 of} plane and normalized shortest distance _n d _s0 )
The measurement of posting, that is, the method of 2.1 will be described in the embodiment of FIG. This is performed in the following flow shown in FIG. The following steps (7) to (11) are the same as steps (8) to (12) of the embodiment C-1.
(Start)
(1) The moving direction v is extracted by the “moving direction v extracting unit 14” in the same manner as in step (1) of the embodiment A-1. Next, the “p _inf setting unit 15” sets the “position p _inf after elapse of infinite time” to be equal to the moving direction v (C-5-1).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-5-2, C- 5-3, C-5-13).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-5-4, C-5-5, C-5-12).
(Scan i)
(4) Output the current and next time positions _i p ₀ and _i p ₁ from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ”.

(5) Consider a circle with a radius r centered in the moving direction v (FIG. 13), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-5-7, C− 5-8, C-5-11).
(Scan r)
Input set five parameters _n d _s (6) _{above, i p 0, i p 1} , p inf, the r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" to, 2.1 (4) two intersections ⁱ _r n _s + ^a, and outputs the _i r _{n s-.} This intersection is a component of “a small circle with radius _i R centered at _i p ₀ ” as proved in 2.2.1. This point ^{_{i r n s +, i r}} n s- for "polar coordinates on the sphere (longitude ^{_{i r α s +, i r}} α s-, latitude ^{_{i r β s +, i r}} β s-)" has the formula (29) It is calculated by the following formula.

ⁱ _r β _{s +} = r
ⁱ _r β _s− = r (220a)
_{^{i r α s + = i α}} a + cos -1 (((n d s sin (i b- i a) / sin (i b))
_{-Cos (r) cos (i a} )) / (sin (r) -sin (i a)))
_{^{i r α s- = i α a}} -cos -1 (((n d s sin (i b- i a) / sin (i b))
_{-Cos (r) cos (i a} )) / (sin (r) -sin (i a)))
(220b)
Here, _i α _a and _i a are the longitude and latitude coordinates of _i p ₀ , and _i b is the latitude coordinate of _i p ₁ (see FIGS. 14 and 9).

(7) the small circle point constituting the ^{_{i r n s +, i r}} n the _s-, vote and converts the "point in the cross-sectional circle of height _n d _s" of the "cylindrical array voting unit 224" ( See step (8) in embodiment C-1 (C-5-10)).
(Scan r (C-5-11))
(8) Through the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-5-12))
(9) Through the processing so far, a small circle group in which “all the points in the image { _{ip 0} }” are transformed is drawn in the cross-sectional circle having the height of _n d _s .
(Scans _n d _s (C-5-13))
(10) In the process up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(11) The “peak extraction unit 225” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point” It is. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (C-5-14).
(End)
Embodiment C-6. (Measured without knowing the direction of movement v of the "three-dimensional orientation _{n s0} and the normalized shortest distance _n d _s0 plane")
The embodiment shown in FIG. 81 will be described with respect to the case where the posted measurement, that is, the “method of measuring without knowing the moving direction v (2.5)” is performed for 2.1. It is performed according to the following flow shown in FIG. The following steps (2) to (10) are the same as the corresponding steps in the embodiment C-5.

(Start)
(0) The “moving direction parameter v” is scanned in all possible directions (from the lower limit value V _min to the upper limit value v _max ) by the “v parameter scanning unit 21” (C-6-1, C-6-2, C-6-16).
(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” as being equal to this parameter v (C-6-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-6-4, C -6-5, C-6-15).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-6-6, C-6-7, C-6-14).
(Scan i)
(4) The current and next time positions _i p ₀ and _i p ₁ are output from the “image register 12 at the current time t ₀ ” and the “image register 13 at the next time t ₁ ” (C-6-8).

(5) Consider a circle with a radius r centered in the moving direction v (FIG. 13), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-6-9, C− 6-10, C-6-13).
(Scan r)
Input set five parameters _n d _s (6) _{above, i p 0, i p 1} , p inf, the r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" to, 2.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _s-(C-6-11).

(7) the small circle point constituting the ^{_{i r n s +, i r}} n the _s-, vote and converts the "point in the cross-sectional circle of height _n d _s" of the "cylindrical array voting unit 224" ( Embodiment C-6-12).
(Scan r (C-6-13))
(8) Through the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-6-14))
(9) Through the processing so far, a small circle group in which “all the points in the image { _{ip 0} }” are transformed is drawn in the cross-sectional circle having the height of _n d _s .
(Scan _{n d s (C-6-15)} )
(10) In the process up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} has been converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
(Scan v (C-6-16))
(11) In the processing so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all the moving direction parameters v”.

(12) From the columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 225”. As the movement direction parameter corresponding to this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle” The direction _ns0 is obtained (C-6-17).
(End)
Embodiment C-7. (Measurement of three-dimensional orientation n _{s0 of} plane and normalized shortest distance _n d _s0 )
The posting measurement, that is, the method of 4.3.1 will be described in the embodiment of FIG. It is performed according to the following flow shown in FIG. After step (7), if _i p _R and { _i p _R } are replaced with _i p ₀ and { _i p ₀ }, they are the same as in the embodiment C-5.
(Start)
(1) With the “p _axis setting unit 115”, the position p _axis on the visual _axis is set to be equal to the visual axis direction a _xis (see step (1) in embodiment B-1 (C-7-1)). .

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-7-2, C- 7-3, C-7-13).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-7-4, C-7-5, C-7-12).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (C-7-6).

(5) Consider a circle with a radius r centered on the visual axis direction a _xis (FIG. 26), and scan r from 0 to π / 2 by the “scanning unit 222 with radius r” (C-7-7, C-7-8, C-7-11).
(Scan r)
Parameter _n d _s five set in (6) _{above, i p R, i p L} , p axis, type r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" to, 4.3.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _{s- (C-7-9).} This intersection is a component of “a small circle with radius _iR centered on _i p _R ” as proved in 4.3.2. This point ^{_{i r n s +, i r}} n s- for "polar coordinates on the sphere (longitude ^{_{i r α s +, i r}} α s-, latitude ^{_{i r β s +, i r}} β s-)" has the formula (65c) It is calculated by the following formula.

ⁱ _r β _{s +} = r
ⁱ _r β _s− = r (230a)
_{^{i r α s + = i α}} c + cos -1 (((n d s sin (i d- i c) / sin (i d))
_{-Cos (r) cos (i c} )) / (sin (r) -sin (i c)))
_{^{i r α s- = i α c}} -cos -1 (((n d s sin (i d- i c) / sin (i d))
_{-Cos (r) cos (i c} )) / (sin (r) -sin (i c)))
(230b)
Here, _i α _c and _i c are the longitude and latitude coordinates of _i p _R , and _i d is the latitude coordinate of _i p _L (see FIGS. 26 and 23).

(7) This point constitutes a small circle ^{_{i r n s +, i r}} n the _s-, vote and converts the "point in the cross-sectional circle of height _n d _s" of the "cylindrical array voting unit 224" ( C-7-10).
(Scan r (C-7-11))
(8) With the processing so far, one small circle whose point _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-7-12))
(9) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scan _n d _s (C-7-13)
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.

(11) The “peak extraction unit 225” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point” It is. The standardized shortest distance _n d _s of the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (C-7-14).
(End)
Embodiment C-8. (Measurement "3-dimensional orientation _{n s0} plane and the normalized shortest distance _n d _s0" a without knowing the visual axis direction _{a xis)}
The case where the above-described measurement, that is, the “method for measuring without knowing the visual axis direction a _xis (4.3.3)” is performed for 4.3.1 will be described with reference to the embodiment of FIG. It is performed according to the following flow shown in FIG. 86 obtained by modifying Embodiment C-7. The following steps (1) to (10) are the same as the steps corresponding to the embodiment C-7.

(Start)
(0) The “visual axis parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scan unit 121”. (C-8-1, C-8-2, C-8-16).
( _{Scan axis} )
(1) The “position on the visual axis p _axis ” is set by the “p _axis setting unit 115” to be equal to the parameter a _xis (C-8-3).

(2) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 221” (C-8-4, C- 8-5, C-8-15).
(Scans _n d _s )
(3) The address i of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max (C-8-6, C-8-7, C-8-14).
(Scan i)
(4) The positions _i p _R and _i p _L in the left and right cameras are output from the “right camera image register 112” and the “left camera image register 113” (C-8-8).

(5) Consider a circle with a radius r centered on the visual axis direction a _xis (FIG. 26), and scan r from 0 to π / 2 by the “scan unit 222 with radius r” (C-8-9, C-8-10, C-8-13).
(Scan r)
Parameter _n d _s five set in (6) _{above, i p R, i p L} , p axis, type r to "small circle components ^{_{i r n s +, i r}} n s- computing unit 223" to, 4.3.1 (4) two intersections ⁱ _r n _s + ^a, _i r _n outputs the _{s- (C-8-11).}

(7) the small circle point constituting the ^{_{i r n s +, i r}} n the _s-, vote and converts the "point in the cross-sectional circle of height _n d _s" of the "cylindrical array voting unit 224" ( C-8-12).
(Scan r (C-8-13))
(8) With the processing so far, one small circle whose point _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s . However, the small circle is converted by the equation (201).
(Scan i (C-8-14))
(9) In the processing up to this point, the height in the cross section circle of _n d _s, small circle group "all the points in the image _{{i p R}"} is converted is drawn.
(Scan _{n d s (C-8-15)} )
(10) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted is drawn. That is, voting is performed on all the cross-sectional circles of the columnar arrangement.
( _{Scan axis} (C-8-16))
(11) In the processing so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all visual axis direction parameters a _xis ”.

(12) From the columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 225”. As the movement direction parameter corresponding to this array, the true movement direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s of the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (C-8-17).
(End)
Embodiment D
Embodiment D-1 (standardization time)
This embodiment will be described with reference to FIG. This is performed according to the following flow shown in FIG.
(1) The moving direction v is extracted in the same manner as in step (1) of the embodiment A-1, and the “p _inf setting unit 15” _assumes that the moving direction v is equal to the moving direction v, and “position p _inf after elapse of infinite time” Is set (D-1-1).
(2) Scan the address i of each point in the image from the lower limit value i _min to the upper limit value i _max by “scanning unit 401 of pixel number i” (D-1-2, D-1-3, D− 1-17).

(Scan i)
(3) by _"i p ₀ cutout unit 402 of the local region image centered on the" pixel _i p ₀ corresponding to "address i from each image of the present time _{t 0} obtained following the time _{t 1} at the camera 11 (See FIG. 1 of JP-A-09-081369 and FIG. 1 of the IEICE Technical Report (Kawakami, Okamoto, vol. IE-19, pp. 31-38, 1996)) the, as shown in the left end of FIG. 159, is cut out for the current time _{t 0} and following the time _{t 1} (D-1-4).
(4) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scanning unit 403 of motion parallax number k” (D-1-5, D-1-6, D-1-16). ). (Scan k)
(5) This number k is a serial number of motion parallax and corresponds to the motion parallax _k τ, that is, the motion vector ( _k τ _{x k} τ _y ) as shown in FIG. This association is performed by the “motion parallax _k τ conversion unit 404”, and the motion parallax _k τ is output (D-1-7).

However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Due to the motion parallax that contradicts this moving direction v, step (10) is skipped (D-1-8).
(6) The “local region image at the current time t ₀ and the next time t ₁ ” and the motion parallax _k τ set above are input to the “motion parallax detection unit 405” (see FIG. 159) to reduce the response intensity. Calculate with the equation (D-1-9).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y) (301)
Here, _i a ₀ (x, y) is the intensity of the (x, y) pixel in the local region image (upper left of FIG. 159) at the current time, and _i a ₁ (x, y) is the local at the next time. This is the intensity of (x, y) pixels in the region image (lower left of FIG. 159).

The calculation of the response intensity is shown as a two-dimensional correlation for the sake of simplicity. For example, the two-dimensional correlation based on the differential absolute value that is normally used in the MPEG2 encoder, the two-dimensional correlation based on the multiplication operation, and the Hough transform. And correlation by inverse Hough transform (“Method of performing two-dimensional correlation and convolution one-dimensionally along the ρ coordinate of the Hough plane,” Kawakami and Okamoto, IEICE Technical Report, vol. IE 96-19, pp. 31- 38, 1996, and Japanese Patent Laid-Open No. 05-165958, Japanese Patent Laid-Open No. 05-165957, Japanese Patent Laid-Open No. 06-044364, Japanese Patent Laid-Open No. 09-08369. Further, a method of detecting speed such as a differential gradient method may be used. That is, this response intensity may be obtained based on the intensity of the pixel. This is the same for all embodiments of the present invention for obtaining response strength.
(7) a normalized time parameter _n t _c by _'n t _c parameters of the scanning unit 16 ", scanning the lower limit _n t _c, the _min limit _n t _c, until _max (D-1-10, D- 1-11, D-1-15).

(Scans _n t _c )
The following is performed in the same process as in the above-described embodiment A-6. However, in the present embodiment, unlike the embodiment A-6, the process of “voting the response intensity of the motion parallax detection unit” is performed in the following step (10).
(8) In the “ _i p ₀ conversion unit 406”, the pixel number i is converted into the pixel _i p ₀ , and the four parameters _n t _c , _k τ, _i p ₀ , and p _inf set as described above are converted into “multi-ratio conversion”. enter the unit 17 ", and outputs the position _{ik p c (D-1-12)} .
(9) the position _ik p _c and polar transformation on a great circle on the sphere by the "polar transformation unit 18", and outputs the position of a point group constituting the great circle _{{ik p GC} (D-} 1 -13).

In supra embodiment _{A-6, ik τ, ik} p c, ik p GC is _{i tau,} i _{p c,} are marked as i _{p GC.}
(10) Vote “response intensity of motion parallax detection unit 405” to “point group on great circle of height _n t _c ” of “cylindrical array voting unit 19” (D-1-14). By the process so far, one great circle in which the point at the position _i p ₀ has been “multi-ratio transformed and pole transformed” is drawn in the cross-sectional circle having the height _n t _c .

  Note that although a cylindrical arrangement is used here, it is not necessary to be a “cylindrical” in all the embodiments of the present invention (including all the drawings), and may be generalized as a three-degree-of-freedom arrangement. In that case, the great circle is a curve corresponding to it.

(Scanning _n t _c (D-1-15))
(Scan k (D-1-16))
(Scan i (D-1-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n t c},} "all the points in the image _{{i p 0}"} is the cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
(12) The “peak extraction unit 20” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized time _n t _c0 until crossing the plane” is obtained, and as the “coordinate in the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (D−1). -18).

Although the normalized time _n t _c is used here, since Δt is constant, an absolute time t _c (that is, _n t _c Δt) may be used. This is the same for all embodiments of the present invention that deal with the normalized time _n t _c .

Embodiment D-2 (standardized time + v unknown)
This embodiment will be described with reference to the block diagram of FIG. This is done in the flow of FIG.

The following steps (2) to (11) are the same as the corresponding steps in the embodiment D-1.
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (D-2-1, D-2-2, D-2-20).

(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” as being equal to this parameter v (D-2-3).
(2) The address of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max by the “scan unit 401 of image number i” (D-2-4, D-2-5, d-2). -19).

(Scan i)
(3) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 11 by the “local region image cutting unit 402 centered on _i p ₀ ” are used. Each local region image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (D-2-6).
(4) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scan unit 403 with motion parallax number k” (D-2-7, D-2-8, D-2-18). ).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-1, the “motion parallax _k τ conversion unit 404” converts this number k into a motion parallax _k τ and outputs it (D-2-9) .

However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Due to the motion parallax that contradicts this moving direction v, the steps up to step (10) are skipped (D-2-10).
(6) The “local region image at the current time t ₀ and the next time t ₁ ” and the motion parallax _k τ set above are input to the “motion parallax detection unit 405” (see FIG. 159) to reduce the response intensity. Calculate with the formula (D-2-11).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(7) The normalization time parameter _n t _c is scanned from the lower limit value _n t _{c, min} to the upper limit value _n t _{c, max} by the “ _n t _c parameter scanning unit 16” (D-2-12, D− 2-13, D-2-17).

(Scans _n t _c )
The following is performed in the same process as in the above-described embodiment A-6. However, in the present embodiment, unlike Embodiment A-6, the process of “voting the response intensity of the motion parallax detection unit” is performed in Step (10).
(8) In the “ _i p ₀ conversion unit 406”, the pixel number i is converted into the pixel _i p ₀ , and the four parameters _n t _c , _k τ, _i p ₀ , and p _inf set as described above are converted into “multi-ratio conversion”. enter the unit 17 ", and outputs the position _{ik p c (D-2-14)} .
(9) the position _ik p _c and polar transformation on a great circle on the sphere by the "polar transformation unit 18", and outputs the position _{{ik p GC}} of the points that constitute the great circle (D-2 -15).
(10) Vote “response intensity of motion parallax detection unit 405” to “point cloud on great circle of height _n t _c ” of “cylindrical array voting unit 19” (D-2-16). By the process so far, one great circle in which the point at the position _i p ₀ has been “multi-ratio transformed and pole transformed” is drawn in the cross-sectional circle having the height _n t _c .

(Scanning _n t _c (D-2-17))
(Scan k (D-2-18))
(Scan i (D-2-19))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n t c},} "all the points in the image _{{i p 0}"} is the cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.

(Scanning v (D-2-20))
(12) In the process so far, voting is performed on all the cross-sectional circles of the “cylindrical array group for all moving direction parameters v”.
(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 20”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point at which the intensity reaches a peak in the array is extracted, the normalization time _n t _c0 until the plane is crossed as the “height coordinate” of the point is obtained, and the coordinate of the plane is obtained as “coordinate in the cross-sectional circle”. A three-dimensional orientation _ns0 is obtained (D-2-21).

Embodiment D-3 (standardized shortest distance)
FIG. 91 is a block diagram of the present embodiment D-3, and FIG. 92 is a flowchart of the embodiment D-3.
(1) The moving direction v is extracted in the same manner as in step (1) of the embodiment A-1, and the “p _inf setting unit 15” _assumes that the moving direction v is equal to the moving direction v, and “position p _inf after elapse of infinite time” Is set (D-3-1).
(2) The address “i” of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max by the “scanning unit 401 of pixel number i” (D-3-2, D-3-3, D− 3-17).

(Scan i)
(3) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 11 by the “local region image cutting unit 402 centered on _i p ₀ ” are used. Each local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (D-3-4).
(4) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scanning unit 403 of the motion parallax unit k” (D-3-5, D-3-6, D-3-16). ).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-1, the “motion parallax _k τ conversion unit 404” converts this number k into a motion parallax _k τ and outputs it (D-3-7). . However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Due to the motion parallax that contradicts this moving direction v, the steps up to step (10) are skipped (D-3-8).
(6) Similarly to step (6) in Embodiment D-1, “local parallax images at current time t ₀ and next time t ₁ ” and motion parallax _k τ are expressed as “motion parallax detection unit 405” (see FIG. 159). ) And the response intensity is calculated by the following equation (D-3-9).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(7) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22”.

(Scans _n d _s )
The following is performed in the same processing as in the above-described embodiment A-8. However, in this embodiment, unlike Embodiment A-8, a process of “voting the response intensity of the motion parallax detection unit” is performed in Step (10).
(8) In the “ _i p ₀ conversion unit 406”, the pixel number i is converted into the pixel _i p ₀ , and the four parameters _n d _s , _k τ, _i p ₀ , and p _{inf set as described above} are used in the embodiment A- In the same manner as in step (6) of FIG. 8, the “radius R calculation unit 23” is input, and the radius _ik R and the position _i p ₀ are output (D-3-12). In the unit 23, the radius _ik R is calculated by the following equation.

^{_{ik R = cos -1 (n d}} s sin k τ / sin (i a + k τ))
(9) In the same manner as in step (7) of Embodiment A-8, the radius _ik R and the position _i p ₀ are input to the “small circle conversion unit 24”, and the radius _ik centered on the position _i p ₀ is entered. It is converted into a “small circle of a spherical surface” of R (D-3-13).
(10) In the same manner from Step embodiment A-8 (8) and (9), the "columnar array voting unit 25" for "height _n d point group on the small circle of _s" Step (6) The response intensity of the motion parallax detection unit 405 is voted (D-3-14). By the processing so far, one small circle obtained by converting the point at the position _i p ₀ into a small circle is drawn in a cross-sectional circle having a height of _n d _s .

(Scanning _n d _s (D-3-15))
(Scan k (D-3-16))
(Scan i (D-3-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} it is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
(12) The “peak extraction unit 26” extracts “the point where the voted intensity is maximized” in the cylindrical arrangement—this maximum point is “the place where the small circles intersect at one point”. The maxima as "height coordinate""normalized shortest distance _{n d} s0 to _plane" is determined, the 3-dimensional orientation n _s0 plane as "the coordinates of the cross section circle" is obtained (D-3- 18).

Embodiment D-4 (standardized shortest distance + v unknown)
This embodiment will be described with reference to the block diagram of FIG. This is executed according to the following flow shown in FIG.

The following steps (2) to (11) are the same as the corresponding steps in the embodiment D-3.
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (D-4-1, D-4-2, D-4-20).

(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (D-4-3).
(2) The address of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 401 of pixel number i” (D-4-4, D-4-5, D-4). -19).

(Scan i)
(3) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 11 by the “local region image cutting unit 402 centered on _i p ₀ ” are used. Each local region image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (D-4-6).
(4) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scan unit 403 of motion parallax number k” (D-4-7, D-4-8, D-4-18). ).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-1, the “motion parallax _k τ conversion unit 404” converts this number k into motion parallax _k τ and outputs it (D-4-9). . However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Due to the motion parallax that contradicts this moving direction v, step (10) is skipped (D-4-10).
(6) In the same manner as in step (6) of Embodiment D-1, “local image of current time t ₀ and next time t ₁ ” and motion parallax _k τ are expressed as “motion parallax detection unit 405” (FIG. 159). And the response intensity is calculated by the following equation (D-4-11).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(7) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (D-4-12, D -4-13, D-4-17).

(Scans _n d _s )
The following is performed in the same processing as in the above-described embodiment A-8. However, in this embodiment, unlike Embodiment A-8, in step (10), a process of “voting the response intensity of the motion parallax detection unit” is performed.
(8) In the “conversion unit 406”, the pixel number i is converted into a pixel, and the four parameters _n d _s , _k τ, _i p ₀ , and p _{inf set above} are used as the step (6) of the embodiment A-8. In the same manner as in (2), the “radius R calculation unit 23” is input, and the radius _ik R and the position _i p ₀ are output (D-4-14). In the unit 23, the radius _ik R is calculated by the following equation.

^{_{ik R = cos -1 (n d}} s sin k τ / sin (i a + k τ))
(9) In the same manner as in step (7) of Embodiment A-8, the radius _ik R and the position _i p ₀ are input to the “small circle conversion unit 24”, and the radius _ik centered on the position _i p ₀ is entered. It is converted into a “spherical circle small circle” of R (D-4-15).
(10) In the same manner from Step embodiment A-8 (8) and (9), the "columnar array voting unit 25" for "height _n d point group on the small circle of _s" Step (6) The response intensity of the motion parallax detection unit 405 is voted (D-4-16). By the processing so far, one small circle obtained by converting the point at the position _i p ₀ into a small circle is drawn in a cross-sectional circle having a height of _n d _s .

(Scans _n d _s (D-4-17))
(Scan k (D-4-18))
(Scan i (D-4-19))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} it is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.

(Scanning v (D-4-20))
(12) In the process so far, voting is performed on all cross sections of the “cylindrical array group for all moving direction parameters v”.
(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 26”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the standardized shortest distance _n d _s to the plane is used as the “height coordinate” of the point, and the plane 3 is used as the “coordinate in the cross-sectional circle”. The dimension orientation _ns0 is obtained (D-4-21).

Embodiment D-5 (stereo + normalized distance)
FIG. 95 is a block diagram showing the embodiment D-5, and FIG. 96 is its flowchart.
(1) As in the above-described embodiment B-1, the “p _axis setting unit 115” sets the position p _axis on the visual _axis as equal to the visual axis direction a _xis (D-5-1).
(2) The “scan unit 421 of pixel number i” scans the address i of each point in the image from the lower limit value i _min to the upper limit value i _max (D-5-2, D-5-3, D- 5-3-17).

(Scan i)
(3) by "clipping unit 422 of the local region image centered on the _i p _R", the local region image centered on the _"i p _R for the address i", as shown at the left end of FIG. 167, right and left It cuts out from each image obtained with the cameras 412 and 413 (D-5-4).
(4) The “binocular parallax number k scanning unit 423” scans the binocular parallax number k from the lower limit value k _min to the upper limit value k _max (D-5-5, D-5-6, D-5). -16).

(Scan k)
(5) This number k is a serial number of binocular parallax and corresponds to binocular parallax _k σ, that is, a parallax vector ( _k σ _{x k} σ _y ) as shown in FIG. This association is performed by the “binocular parallax _k σ conversion unit 424”, and the binocular parallax _k σ is output (D-5-7).

However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (ie, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (10) (D-5-8) .
(6) The “local region images of the right and left cameras” and the binocular parallax _k σ set above are input to the “binocular parallax detection unit 425” (FIG. 167), and the response intensity is calculated by the following equation. (D-5-9).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y) (302)
Here, _i a _R (x, y) is the intensity of the (x, y) pixel in the local region image (upper left of FIG. 167) of the right camera, and _i a _L (x, y) is the local of the left camera. This is the intensity of (x, y) pixels in the region image (lower left of FIG. 167).

Although the calculation of the response intensity is shown as a two-dimensional correlation for the sake of simplicity, the two-dimensional correlation based on the absolute difference, which is standard for MPEG2 encoders, the two-dimensional correlation based on a multiplication operation, and the inverse of the Hough transform. Correlation by Hough transform (“Method of performing two-dimensional correlation and convolution one-dimensionally along the ρ coordinate of the Hough plane,” Kawakami and Okamoto, IEICE Technical Report, vol. IE96-19, pp.31-38, 1996, and Japanese Patent Laid-Open No. 05-165958, Japanese Patent Laid-Open No. 05-165957, Japanese Patent Laid-Open No. 06-044364, Japanese Patent Laid-Open No. 09-08369, and the differential gradient method. Can be obtained based on the intensity of the pixel.
(7) The standardized distance parameter _n d _c is scanned from the lower limit value _n d _{c, min} to the upper limit value _n d _{c, max} by the “ _n d _c parameter scanning unit 116” (D-5-10, D− 5-11, D-5-15).

(Scans _n d _c )
The following is performed in the same process as in the above-described embodiment B-6. However, in the present embodiment, unlike the embodiment B-6, the process of “voting the response intensity of the binocular parallax detection unit” is performed in step (10).
(8) In the “ _i p _R conversion unit 426”, the pixel number i is converted into the pixel _i p _R , and the four parameters _n d _c , _k σ, _i p _R , and p _axis set as described above are converted into “multiple ratios”. enter the conversion unit 117 ", and outputs the position _{ik p c (D-5-12)} .
(9) the position _ik p _c and polar transformation on a great circle on the sphere by the "polar transformation unit 118", and outputs the position of a point group constituting the great circle _{{ik p GC} (D-} 5 -13).
(10) to the "great circle on the point group of the height of the _n d _c" of the "cylindrical array voting unit 119" vote "response strength of binocular disparity detection unit" (D-5-14). By the process so far, one great circle whose point _i p _R is “cross ratio transformed and pole transformed” is drawn in a sectional circle having a height of _n d _c .

(Scan _{n d s (D-5-15)} )
(Scan k (D-5-16))
(Scan i (D-5-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d c},} the "all points _{{i p} R} in the _image" are cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
(12) “Peak extraction unit 120” extracts “the point where the voted intensity becomes the maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. The “normalized distance _n d _c0 ” to the plane is obtained as the “height coordinate” of this local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (D-5-18). ).

Although the normalized distance _n d _c is used here, the absolute distance d _c (that is, _n d _c Δx _LR ) may be used because Δx _LR is constant. This point, the present invention is the same for all embodiments dealing with normalized distance _n d _c.

Embodiment D-6 (stereo + normalized distance + a _xis unknown)
FIG. 97 is a block diagram showing the embodiment D-6, and FIG. 98 is its flowchart.

The following steps (2) to (11) are the same as the corresponding steps in the embodiment D-5.
(0) “ _Axis parameter scanning unit 121” scans the “visual axis direction parameter a _xis ” in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ). (D-6-1, D-6-2, D-6-20).

( _{Scan axis} )
(1) The “position on the visual axis p _axis ” is set to be equal to the parameter a _xis by the “p _axis setting unit 115” (D-6-3).
(2) The address “i” of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max by the “scanning unit 421 of pixel number i” (D-6-4, D-6-5, D−). 6-19).

(Scan i)
(3) by "clipping unit 422 of the local region image centered on the _i p _R", the local region image centered on the "pixel _i p _R corresponding to the address i", as shown at the left end of FIG. 167, Cut out from the images obtained by the right and left cameras 412 and 413 (D-6-6).
(4) “Binocular parallax number k scanning unit 423” scans the binocular parallax number k of the binocular parallax detection unit from the lower limit value k _min to the upper limit value k _max (D-6-7, D-6) -8, D-6-18).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-5, this “ _k conversion parallax _k σ conversion unit 424” converts this number k into binocular parallax _k σ and outputs it (D-6-6). 9).

However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (ie, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (10) (D-6-10) .
(6) The “local region images of the right and left cameras” and the binocular parallax _k σ set above are input to the “binocular parallax detection unit 425” (FIG. 167), and the response intensity is calculated by the following equation. (D-6-11).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(7) The normalized distance parameter _n d _c is scanned from the lower limit value _n d _{c, min} to the upper limit value _n d _{c, max} by the “ _n d _c parameter scanning unit 116” (D-6-12, D− 6-13, D-6-17).

(Scans _n d _c )
The following is performed in the same process as in the above-described embodiment B-6. However, in this embodiment, unlike the embodiment B-6, a process of “voting the response intensity of the binocular parallax detection unit” is performed in step (10).
(8) In the “ _i p _R conversion unit 426”, the pixel number i is converted to the pixel _i p _R , and the four parameters _n d _c , _k σ, _i p _R , and p _axis set as described above are converted into the “multi-ratio conversion”. enter the unit 117 'outputs the position _{ik p c (D-6-14)} .
(9) the position _ik p _c and polar transformation on a great circle on the sphere by the "polar transformation unit 118", and outputs the position of a point group constituting the great circle _{{ik p GC} (D-} 6 -15).
(10) to the "cylindrical array voting unit 119" of the "height _n d a great circle on the point group of the _c", to vote "response strength of binocular disparity detection unit 425" (D-6-16). By the process so far, one great circle whose point _i p _R is “cross ratio transformed and pole transformed” is drawn in a sectional circle having a height of _n d _c .

(Scan _{n d c (D-6-17)} )
(Scan k (D-6-18))
(Scan i (D-6-19))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d c},} the "all points _{{i p} R} in the _image" are cross-ratio conversion and polar transformed great circles group be painted. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.

( _{Scan axis} ((D-6-20))
(12) by the processing up to here, the vote total cross-sectional circle of the "cylinder arrangement group for all of the visual axis direction parameter a _xis" is performed.
(13) From the columnar array group, a “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 120” (D-6-21). As the movement direction parameter for this array, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized distance _n d _c0 to the plane is used as the “height coordinate” of the point, and 3 of the plane is used as the “coordinate in the cross-sectional circle”. A dimension orientation _ns0 is determined.

Embodiment D-7 (stereo + standardized shortest distance)
FIG. 99 is a block diagram showing the embodiment D-7, and FIG. 100 is a flowchart thereof.
(1) As in the above-described embodiment B-1, the “p _axis setting unit 115” sets the position p _axis on the visual _axis to be equal to the visual axis direction a _xis (D-7-1).
(2) The “scan unit 421 of image number i” scans the address i of each point in the image from the lower limit value i _min to the upper limit value i _max (D-7-2, D-7-3, D- 7-17).

(Scan i)
(3) by "clipping unit 422 of the local region image centered on the _i p _R", the local region image centered on the "pixel _i p _R corresponding to the address i", as shown at the left end of FIG. 167, Cut out from the images obtained by the right and left cameras 412 and 413 (D-7-4).
(4) The “binocular parallax number k scanning unit 425” scans the binocular parallax number k from the lower limit value k _min to the upper limit value k _max (D-7-5, D-7-6, D-7). -16).
(5) In the same manner as in step (5) of Embodiment D-5, “number b parallax number k scanning unit 425” converts this number k into binocular parallax _k σ and outputs it (D-7) -7). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (ie, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (10) (D-7-8) .
(6) In the same manner as in Step (6) of Embodiment D-5, the “local region images of the right and left cameras” and the binocular parallax _k σ are input to the “binocular parallax detection unit 425” (FIG. 167). Then, the response intensity is calculated by the following equation (D-7-9).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(7) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (D-7-10, D -7-11, D-7-15).

(Scans _n d _s )
The following is performed in the same process as in the above-described embodiment B-8. However, in this embodiment, unlike Embodiment B-8, the process of “voting the response intensity of the binocular parallax detection unit” is performed in Step (10).
(8) In the “ _i p _R conversion unit”, the pixel number i is converted into the pixel _i p _R , and the four parameters _n d _s , _k σ, _i p _R , and p _axis set as described above are converted into the embodiment B- Similarly to step (6) of FIG. 8, the “radius R calculation unit 123” is input, and the radius _ik R and the position _i p _R are output (D-7-12). In the unit 123, the radius _ik R is calculated by the following equation.

^{_{ik R = cos -1 (n d}} s sin k σ / sin (i c + k σ))
(9) In the same manner as in step (7) of the embodiment B-8, the radius _ik R and the position _i p _R are input to the “small circle conversion unit 124”, and the radius _ik about the position _i p _R is the center. It is converted into a “small circle on the spherical surface” of R (D-4-13).
(10) In the same manner as Steps (8) to (9) of Embodiment B-8, the “point group on the small circle with the height _n d _s ” of the “cylindrical array voting unit 125” is changed to Step (6). The response intensity of the binocular parallax detection unit 425 is voted (D-7-14). By the processing so far, one small circle obtained by converting the point at the position _i p _R into a small circle is drawn in a cross-sectional circle having a height of _n d _s .

(Scans _n d _s (D-7-15))
(Scan k (D-7-16))
(Scan i (D-7-17))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
(12) The “peak extraction unit 126” extracts “the point where the voted intensity is maximized” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point”. The “normalized distance _n d _s0 ” to the plane is obtained as the “height coordinate” of the local maximum point, and the three-dimensional azimuth n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (D-7-18). ).

Embodiment D-8 (stereo + normalized shortest distance + a _xis unknown)
FIG. 101 is a block diagram of D-8 of this embodiment, and FIG. 102 is its flowchart.

The following steps (2) to (11) are the same as the corresponding steps in the embodiment D-7.
(0) “ _Axis parameter scanning unit 121” scans the “visual axis direction parameter a _xis ” in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ). (D-8-1, D-8-2, D-8-20).

( _{Scan axis} )
(1) The “p _axis setting unit 115” sets “position p _axis on the visual _axis ” as equal to the parameter a _xis (D-8-3).
(2) The “scan unit 421 of pixel number i” scans the address i of each point in the image from the lower limit value i _min to the upper limit value i _max (D-8-4, D-8-5, D− 8-19).

(Scan i)
(3) by "clipping unit 422 of the local region image centered on the _i p _R", the local region image centered on the "pixel _i p _R corresponding to the address i", as shown at the left end of FIG. 167, Cut out from the images obtained by the right and left cameras 412 and 413 (D-8-6).
(4) The “binocular parallax number k scanning unit 423” scans the binocular parallax number k from the lower limit value k _min to the upper limit value k _max (D-8-7, D-8-8, D-8). -18).

(Scan k)
(5) In the same manner as in step (5) of Example D-5, the “binocular parallax _k σ conversion unit 424” converts this number k into binocular parallax _k σ and outputs it (D-8) -9). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (10) (D-8-10) .
(6) In the same manner as in Step (6) of Embodiment D-5, the “local region images of the right and left cameras” and the binocular parallax _k σ are input to the “binocular parallax detection unit 425” (FIG. 167). Then, the response intensity is calculated by the following equation (D-8-11).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(7) The standardized distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (D-8-12, D− 8-13, D-8-17).

(Scans _n d _s )
The following is performed in the same process as in the above-described embodiment B-8. However, in this embodiment, unlike Embodiment B-8, the process of “voting the response intensity of the binocular parallax detection unit” is performed in Step (10).
(8) In the “ _i p _R conversion unit 426”, the pixel number i is converted into the pixel _i p _R , and the four parameters _n d _s , _k σ, _i p _R , and p _{axis set as described above} are used in the embodiment B- Similarly to step (6) of FIG. 8, the data is input to the “radius R calculation unit 123”, and the radius _ik R and the position _i p _R are output (D-8-14). In the unit 123, the radius _ik R is calculated by the following equation.

^{_{ik R = cos -1 (n d}} s sin k σ / sin (i c + k σ))
(9) In the same manner as in step (7) of the embodiment B-8, the radius _ik R and the position _i p _R are input to the “small circle conversion unit 124”, and the radius _ik about the position _i p _R is the center. It is converted into a “small circle on the spherical surface” of R (D-8-15).
(10) In the same manner as Steps (8) to (9) of Embodiment B-8, the “point group on the small circle with the height _n d _s ” of the “cylindrical array voting unit 125” is changed to Step (6). The response intensity of the binocular parallax detection unit 425 is voted (D-8-16). With the processing so far, a small circle whose point at the position _i p _R is converted into a small circle is drawn in a cross-sectional circle whose height is _n d _s .

(Scans _n d _s (D-8-17))
(Scan k (D-8-18))
(Scan i (D-8-19))
(11) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
( _{Scan axis} )
(12) by the processing up to here, the vote total cross-sectional circle of the "cylinder arrangement group for all of the visual axis direction parameter a _xis" is performed.
(13) From the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 126”. As the visual axis direction parameter for this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 to the plane is used as the “height coordinate” of the point, and the plane 3 The dimension orientation _ns0 is obtained (D-8-21).

Embodiment D-9 (standardized shortest distance)
FIG. 103 is a block diagram showing the configuration of the embodiment D-9, and FIG. 104 is a flowchart showing the operation thereof.
(1) The moving direction v is extracted in the same manner as in step (1) of the embodiment A-1 described above, and the “p _inf setting unit 15” determines that the moving direction v is equal to the moving direction v, “position p after infinite time has elapsed. _inf "is set (D-9-1).
(2) Scan the address i of each point in the image from the lower limit value i _min to the upper limit value i _max by “scanning unit 401 of pixel number i” (D-9-2, D-9-3, D− 9-19).

(Scan i)
(3) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 11 by the “local region image cutting unit 402 centered on _i p ₀ ” are used. Each local region image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (D-9-4).
(4) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scan unit 403 of motion parallax number k” (D-9-5, D-9-6, D-9-18). ).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-1, the “motion parallax _k τ conversion unit 404” converts “this number k into motion parallax _k τ and outputs it (D-9-7). However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Skips to step (11) due to motion parallax that contradicts this moving direction v (D-9-8).
(6)) Similar to step (6) of Embodiment D-1, “local parallax images at current time t ₀ and next time t ₁ ” and motion parallax _k τ are expressed as “motion parallax detection unit 405” (FIG. 159). The response intensity is calculated by the following equation (D-9-9).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(7) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (D-9-10, D -9-11, D-9-17).

(Scans _n d _s )
(8) Consider a circle with a radius r centered in the moving direction v (FIG. 13), and scan r from 0 to π / 2 (D-9-12, D-9-13, D-9-16). .

(Scan r)
(9) In the “ _i p ₀ conversion unit 406”, the pixel number i is converted into a pixel, and the five parameters _n d _s , _i p ₀ , p _inf , _k τ, r set as described above are converted into “small circle constituent elements”. ^{_{i r n s +, i r}} n s- of input to the calculation unit 223 ", the two intersection points described with reference to FIG. ^{_{13 i r n s +, i}} r n outputs the _{s- (D-9-14} ).
(10) the point constituting the small circle ⁱ _r n _s ^+, a _i r _{n s-,} converted to "points in circular section of height _n d _s" of the "cylindrical array voting unit 224". Next, the “response intensity of the motion parallax detection unit” calculated in step (6) is voted for that point (D-9-15).

(Scan r (D-9-16))
(11) Through the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s .

(Scans _n d _s (D-9-17))
(Scanning k (D-9-18))
(Scan i (D-9-19))
(12) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} it is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
(13) The “peak extraction unit 225” extracts “the point where the voted intensity is maximized” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point”. The maxima as "height coordinate""normalized shortest distance _{n d} s0 to _plane" is determined, the 3-dimensional orientation n _s0 plane as "the coordinates of the cross section circle" is obtained (D-9- 20).

Although the normalized shortest distance _n d _s is used here, since Δx is constant, the shortest distance d _s (that is, _n d _s Δx) may be used. This point is the same for all embodiments of the present invention that handle the normalized shortest distance _n d _s .

Embodiment D-10 (standardized shortest distance + v unknown)
FIG. 105 is a block diagram showing the configuration of the present embodiment D-10, and FIG. 106 is a flowchart showing the operation thereof.

The following steps (2) to (11) are the same as the corresponding steps in embodiment D-9.
(0) The “moving direction parameter v” is scanned by the “v parameter scanning unit 21” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (D-10-1, D-10-2, D-10-22).

(Scan v)
(1) The “p _inf setting unit 15” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (D-10-3).
(2) The address of each point in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 401 of pixel number i” (D-10-4, D-10-5, D-10) -21).

(Scan i)
(3) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 11 by the “local region image cutting unit 402 centered on _i p ₀ ” are used. Each local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (D-10-6).
(4) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scan unit 403 of motion parallax number k” (D-10-7, D-10-8, D-10-20). ).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D, the “motion parallax number _k τ conversion unit 404” converts this number k into a motion parallax _k τ and outputs it (D-10-9). However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Due to the motion parallax that contradicts this moving direction v, the steps up to step (11) are skipped (D-10-10).
(6) Similarly to step (6) in Embodiment D-1, “local parallax images at current time t ₀ and next time t ₁ ” and motion parallax _k τ are expressed as “motion parallax detection unit 405” (see FIG. 159). ) And the response intensity is calculated by the following equation (D-10-11).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(7) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 22” (D-10-12, D -10-13, D-10-19).

(Scans _n d _s )
The following is performed in the same process as in the above-described embodiment C-2. However, in this embodiment, unlike Embodiment C-2, a process of “voting the response intensity of the motion parallax detection unit” is performed in Step (10).
(8) Consider a circle with a radius r centered in the moving direction v (FIG. 13), and scan r from 0 to π / 2 (D-10-14, D-10-15, D-10-18). .

(Scan r)
(9) In the “ _i p ₀ conversion unit 406”, the pixel number i is converted to the pixel _i p ₀ , and the five parameters _n d _s , _i p ₀ , p _inf , _k τ, r set above are changed to “small”. circular components ⁱ _r n _s ^+, enter the _i r _{n s-calculation} unit 223 ", with reference to Figure 13 two intersections ⁱ _r n _s + ^explained, _i r _n outputs the _s-(D- 10-16).
(10) the small circle point constituting the ⁱ _r n _s ^+, a _i r _{n s-,} converted to "points in circular section of height _n d _s" of the "cylindrical array voting unit 224", to the point Vote the “response intensity of motion parallax detection unit” calculated in step (6) (D-10-17).

(Scan r (D-10-18))
(11) Through the processing so far, one small circle in which the point at the position _i p ₀ has been transformed is drawn in the cross-sectional circle having a height of _n d _s .

(Scan _{n d s (D-10-19)} )
(Scan k (D-10-20))
(Scan i (D-10-21))
(12) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p 0}"} it is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.

(Scanning v (D-10-22))
(13) In the process so far, voting is performed for all the cross-sectional circles of the “cylindrical array group for all moving direction parameters v”.
(14) In the above-described columnar array group, extraction is performed by the “specific columnar array” “peak extraction unit 225” having the maximum intensity. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 to the plane is set as the “height coordinate” of the point, and 3 of the plane is set as the “coordinate in the cross-sectional circle”. The dimension orientation _ns0 is obtained (D-10-23).

Embodiment D-11 (stereo + standardized shortest distance)
FIG. 107 is a block diagram showing the configuration of the present embodiment D-11, and FIG. 108 is a flowchart showing its operation.
(1) As in the above-described embodiment B-1, the “p _axis setting unit 115” sets the position p _axis on the visual _axis as equal to the visual axis direction a _xis (D-11-1).
(2) The “scanning unit 421 of pixel number i” scans the address i of each point in the image from the lower limit value i _min to the upper limit value i _max (D-11-2, D-11-3, D− 11-19).

(Scan i)
(3) by "clipping unit 422 of the local region image centered on the _i p _R", the local region image centered on the _"i p _R for the address i", as shown at the left end of FIG. 167, right and left Cut out from each image obtained by the cameras 412 and 413 (D-11-4).
(4) The “binocular parallax number k scanning unit 423” scans the binocular parallax number k from the lower limit value k _min to the upper limit value k _max (D-11-5, D-11-6, D-11). -18).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-5, the “binocular parallax _k σ conversion unit 424” converts this number k into binocular parallax _k σ and outputs it (D-11) -7). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (10) (D-11-8) .
(6) In the same manner as in step (6) of Embodiment D, the “local region images of the right and left cameras” and the binocular parallax _k σ are input to the “binocular parallax detection unit 425” (see FIG. 167). The response intensity is calculated by the following equation (D-11-9).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(7) The standardized distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (D-11-10, D− 11-11, D-11-17).

(Scans _n d _s )
The following is performed in the same process as in the above-described embodiment C-3. However, in the present embodiment, unlike the embodiment C-3, a process of “voting the response intensity of the binocular parallax detection unit” is performed in step (10).
(8) _Consider a _circle with a radius r centered on the visual axis direction a _xis (FIG. 26), and scan r from 0 to π / 2 (D-11-12, D-11-13, D-11-). 16).

(Scan r)
(9) In the “ _i p _R conversion unit 426”, the pixel number i is converted into the pixel _i p _R , and the five parameters _n d _s , _i p ₀ , p _inf , _k σ, r set above are changed to “small”. circular components ⁱ _r n _s ^+, enter the _i r _{n s-calculation} unit 223 ", the two intersection points ⁱ _r n _s + has been described with reference to FIG. ^26, _i r _n outputs the _s-(D- 11-14).
(10) the point constituting the small circle ⁱ _r n _s ^+, a _i r _{n s-,} converted to "points in circular section of height _n d _s" of the "cylindrical array voting unit 224". Next, the “response intensity of the binocular parallax detection unit” calculated in step (6) is voted for that point (D-11-15).

(Scan r (D-11-16))
(11) Through the processing so far, one small circle in which the point at the position _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s .

(Scans _n d _s (D-11-17))
(Scan k (D-11-18))
(Scan i (D-11-19))
(12) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.
(13) The “peak extraction unit 225” extracts “the point where the voted intensity is maximized” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point”. The maxima as "height coordinate""normalized shortest distance _n d _s0 to plane" is determined, the 3-dimensional orientation _{n s0} plane as "the coordinates of the cross section circle" is obtained (D-11- 20).

Although the normalized shortest distance _n d _s is used here, since Δx _LR is constant, the shortest distance d _s (that is, _n d _s Δx _LR ) may be used. This point is the same for all embodiments of the present invention that handle the normalized shortest distance.

Embodiment D-12 (stereo + normalized shortest distance + a _xis unknown)
FIG. 109 is a block diagram showing the configuration of the present embodiment D-12, and FIG. 110 is a flowchart showing the operation thereof.

The following steps (2) to (11) are the same as the corresponding steps in embodiment D-11.
(0) “ _Axis parameter scanning unit 121” scans the “visual axis direction parameter a _xis ” in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ). (D-12-1, D-12-2, D-12-22).
(1) The “p _axis setting unit 115” sets “position p _axis on the visual _axis ” as equal to the parameter a _xis (D-12-3).

( _{Scan axis} )
(2) The “scan unit 421 of pixel number i” scans the address i of each point in the image from the lower limit value i _min to the upper limit value i _max (D-12-4, D-12-5, D− 12-21).

(Scan i)
(3) by "clipping unit 422 of the local region image centered on the _i p _R", the local region image centered on the "pixel _i p _R corresponding to the address i", as shown at the left end of FIG. 167, Cut out from the images obtained by the right and left cameras 412 and 413 (D-12-6).
(4) The “binocular parallax number k scanning unit 423” scans the binocular parallax number k from the lower limit value k _min to the upper limit value k _max (D-12-7, D-12-8, D-12). -20).

(Scan k)
(5) In the same manner as in step (5) of Embodiment D-5, the “binocular parallax _k σ conversion unit 425” converts this number k into binocular parallax _k σ and outputs it (D-12) -9). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (10) (D-12-10) .
(6) “Local area images of right and left cameras” and binocular parallax _k σ are input to “Binocular parallax detection unit 425” (see FIG. 167) in the same manner as in Step (6) of Embodiment D-5. Then, the response intensity is calculated by the following equation (D-12-11).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(7) The standardized shortest distance parameter _n d _s is scanned from the lower limit value _n d _{s, min} to the upper limit value _n d _{s, max} by the “ _n d _s parameter scanning unit 122” (D-12-12, D -12-13, D-12-19).

(Scans _n d _s )
The following is performed in the same process as in the above-described embodiment C-4. However, in the present embodiment, unlike the embodiment C-4, a process of “voting the response intensity of the binocular parallax detection unit” is performed in step (10).
(8) _Consider a _circle with a radius r centered on the visual axis direction a _xis (FIG. 26), and scan r from 0 to π / 2 (D-12-14, D-12-15, D-12-). 18).

(Scan r)
(9) In the “ _i p _R conversion unit 426”, the pixel number i is converted to the pixel _i p _R , and the five parameters _n d _s , _i p _R , p _inf , _k σ, r set above are changed to “small” circular components ⁱ _r n _s ^+, enter the _i r _{n s-calculation} unit 223 ", the two intersection points ⁱ _r n _s + has been described with reference to FIG. ^26, _i r _n outputs the _s-(D- 12-16).
(10) the small circle point constituting the ⁱ _r n _s ^+, a _i r _{n s-,} converted to "points in circular section of height _n d _s" of the "cylindrical array voting unit 224", to the point Vote “response intensity of binocular parallax detection unit” calculated in step (6) (D-12-17).

(Scan r (D-12-18))
(11) Through the processing so far, one small circle in which the point at the position _i p _R has been transformed is drawn in a cross-sectional circle having a height of _n d _s .

(Scanning _n d _s (D-12-19))
(Scan k (D-12-20))
(Scan i (D-12-21))
(12) In the processing up to this point, in the cross-sectional circle of all heights _{{n d s},} small circle group "all the points in the image _{{i p R}"} is converted small circle is drawn. That is, voting is performed for all the cross-sectional circles in the columnar arrangement.

( _{Scan axis} (D-12-22))
(13) In the process so far, voting is performed on all the cross-sectional circles of the “column array group for all visual axis direction parameters a _xis ”.
(14) “Peak extraction unit 225” extracts the “specific cylinder arrangement” having the maximum intensity from the above-described cylinder arrangement group. As the visual axis direction parameter for this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 to the plane is set as the “height coordinate” of the point, and 3 of the plane is set as the “coordinate in the cross-sectional circle”. The dimension orientation _ns0 is obtained (D-12-23).

Embodiment E
Embodiment E-1 (standardization time)
FIG. 111 is a block diagram showing the configuration of the present embodiment E-1, and FIGS. 112 and 113 are flowcharts showing the operation thereof.
(Α) _ij τ table (FIG. 163) —that is, a table for retrieving and outputting motion parallax _ij τ from the pixel number i of the input image and the element number j of the cylindrical array (FIG. 111 ( A), see FIG. 112)
(Α1) The moving direction v is extracted in the same manner as in step (1) of the above-described embodiment A-1, and the “p _inf setting unit 501” determines that the moving direction v is equal to the moving direction v, “position p after infinite time has elapsed. _inf "is set (E-1α-1).
(Α2) The “scanning unit 502 of pixel number i” scans the address i of each point _i p ₀ in the image from the lower limit value i _min to the upper limit value i _max (E-1α-2, E-1α-3 , E-1α-10).

(Scan i)
(Α3) The “scan unit 503 of element number j” scans the address j of the element (n _sj , _n t _cj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-1α-4, E-1α-5, E-1α-9).

(Scan j)
(Α4) “ _i p ₀ output unit 504”, “n _sj output unit 505”, and “ _n t _cj output unit 506” allow the pixels _i p ₀ in the image corresponding to these addresses (i, j) and The cylindrical array elements (n _sj , _n t _cj ) are output (E-1α-6).
(Α5) The four parameters n _sj , _n t _cj , _i p ₀ , and p _inf set above are input to the “ _ij τ table creation unit 507”, and the motion is performed by the method shown in (A-1) of the appendix. The parallax _ij τ is calculated (E-1α-7).

To become complicated here, following movement direction _(movement direction from i _{p 0} to _i p _1, i.e. _ij phi) is omitted like. First, _ij φ is obtained from P ₁ calculated in (A) of the appendix, and this also needs to be the contents of the _ij τ table, and the motion vector ( _ij τ) to be detected in the following step (β5) _It is necessary to vote only when the direction tan ⁻¹ ( _ij τ _y / _ij τ _x ) in the _x , _ij τ _y ) and this _ij φ match, but these are omitted. This is also omitted in the following embodiments involving similar processing.

The motion parallax _ij τ is not limited to the method shown in Appendix (A-1), and may be calculated by any method. This is the same in all the embodiments described to calculate motion parallax by the method shown in Appendix (A-1).

Further, in the present embodiment, an _ij τ table is created, and _ij τ is extracted from the created _ij τ table in actual voting. However, it is not necessary to create the _ij τ table in advance. _Ij τ may be calculated when performing the above operation.
(Α6) This motion parallax _ij τ is stored as contents corresponding to the address (i, j) of the _ij τ table (FIG. 163) (E-1α-8).

(Scan j (E-1α-9))
(Scan i (E-1α-10))
Thus, the _ij τ table (FIG. 163) is obtained.
(Β) Using the _ij τ table created above, the plane orientation n _s0 and the normalized time _n t _c0 are detected (see FIGS. 111 and 113).
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 552 of pixel number i” (E-1β-1, E-1β-2) , E-1β-10).

(Scan i)
(Β2) Similar to step (3) in Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 551 by the “local region image extraction unit 554 centered on _i p ₀ ”. Each local region image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ for the current time t ₀ and the next time t ₁ (E-1β-3).
( _Β3 ) The address j of the element (n _sj , _n t _cj ) in the columnar array is scanned from the lower limit value j _min to the upper limit value j _max by the “scan unit 553 of element number j” (E-1β-4, E-1β-5, E-1β-9).

(Scan j)
(Β4) These addresses (i, j) are input to the _ij τ table 555 (FIG. 163), and the motion parallax _ij τ is output (E-1β-6).
(Β5) In the same manner as in step (6) of Embodiment D-1, “local image of current time t ₀ and next time t ₁ ” and motion parallax _k τ are determined as “motion parallax detection unit 556” (FIG. 159). And the response intensity is calculated by the following equation (E-1β-7).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- ij τ x, y- ij τ y)
(Β6) Vote this response intensity to the elements (n _sj , _n t _cj ) in the “cylindrical array voting unit 557” (E-1β-8).

(Scan j (E-1β-9))
(Scan i (E-1β-10))
(Β7) “Peak extraction unit 558” extracts “the point at which the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized time _n t _c0 until crossing the plane” is obtained, and as the “coordinate in the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (E-1β -11).

Embodiment E-2 (normalization time + v unknown)
FIG. 114 is a block diagram showing the configuration of the present embodiment E-2, and FIGS. 115 and 116 are flowcharts showing the operation thereof.
(Α) _ij τ table (FIG. 164) for all moving directions {v} (ie, search and output motion parallax _ij τ from pixel number i of input image, element number j of cylindrical array, and moving direction v) To create a table to do (see FIG. 114 (A), FIG. 115)
The steps from (α2) to (α6) below are the same as the corresponding steps in embodiment E-1 (α).
(Α0) “V-parameter scanning unit 508” scans “movement direction parameter v” in all possible directions (from lower limit value v _min to upper limit value v _max ) (E-2α−1, E-2α-2, E-2α-13).

(Scan v)
(Α1) “p _inf setting unit 501” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (E-2α-3).
(Α2) The “scanning unit 502 of pixel number i” scans the address i of each point _i p ₀ in the image from the lower limit value i _min to the upper limit value i _max (E-2α-4, E-2α-5 , E-2α-12).

(Scan j)
(Α3) The “scan unit 503 of element number j” scans the address j of the element (n _sj , _n t _cj ) in the cylindrical array from the lower limit j _min to the upper limit j _max (E-2α-6, E-2α-7, E-2α-11).

(Scan j)
(Α4) “ _i p ₀ output unit 504”, “n _sj output unit 505”, and “ _n t _cj output unit 506” allow the pixels _i p ₀ in the image corresponding to these addresses (i, j) and The cylindrical array elements (n _sj , _n t _cj ) are output (E-2α-8).
(Α5) The four parameters n _sj , _n t _cj , _i p ₀ , and p _inf set above are input to the “ _ij τ table creation unit 507”, and the motion parallax is performed according to the method shown in Appendix (A-1). _ij τ is calculated (E-2α-9).
(Α6) This motion parallax _ij τ is stored as content corresponding to the address (i, j) of the _ij τ table (FIG. 164) (E-2α-10).

(Scan j (E-2α-11))
(Scan i (E-2α-12))
(Α7) By the processing so far, an _ij τ table for each moving direction v is obtained.

As described above, the _ij τ table (FIG. 164) for all the moving directions {v} is obtained.
(Β) Using the _ij τ table created above, the plane orientation n _s0 and the normalized time _n t _c0 are detected (see FIGS. 114B and 116).
The steps from (β1) to (β6) below are the same as the corresponding steps in embodiment E-1.
(Β0) “V-parameter scanning unit 559” scans the “movement direction parameter v” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (E-2β-1, E-2β-2, E-2β-13).

(Scan v)
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 552 of pixel number i” (E-2β-3, E-2β-4 , E-2β-12).

(Scan i)
(Β2) Similar to step (3) in Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 551 by the “local region image extraction unit 554 centered on _i p ₀ ”. Each local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (E-2β-5).
( _Β3 ) The address j of the element (n _sj , _n t _cj ) in the cylindrical array is scanned from the lower limit value j _min to the upper limit value j _max by the “scan unit 553 of element number j” (E-2β-6, E-2β-7, E-2β-11).

(Scan j)
(Β4) These addresses (i, j) and moving direction v are input to the _ij τ table 555 (FIG. 164), and motion parallax _ij τ is output (E-2β-8).
(Β5) Similarly to step (6) of Embodiment D-1, “local image of current time t ₀ and next time t ₁ ” and motion parallax _ij τ are expressed as “motion parallax detection unit 556” (see FIG. 159). ) And the response intensity is calculated by the following equation (E-2β-9).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- ij τ x, y- ij τ y)
(Β6) Vote this response intensity to the elements (n _sj , _n t _cj ) in the “cylindrical array voting unit 557” (E-2β-10).

(Scan j (E-2β-11))
(Scan i (E-2β-12))
(Β7) In the process so far, voting is performed for all the elements in the cylindrical array.

(Scanning v (E-2β-13))
(Β8) In the process so far, voting is performed for all elements of the cylindrical array group for all the movement direction parameters v.
(Β9) Among the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 558”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point at which the intensity reaches a peak in the array is extracted, the normalization time _n t _c0 until the plane is crossed as the “height coordinate” of the point is obtained, and the coordinate of the plane is obtained as the “coordinate in the cross-sectional circle”. A three-dimensional orientation _ns0 is obtained (E-2β-14).

Embodiment E-3 (standardized shortest distance)
FIG. 117 is a block diagram showing the configuration of the embodiment E-3, and FIGS. 118 and 119 are flowcharts thereof.
(Α) _ij τ table (FIG. 163) —that is, a table for retrieving and outputting motion parallax _ij τ from the pixel number i of the input image and the element number j of the cylindrical array (FIG. 117 ( A), see FIG. 118)
(Α1) The moving direction v is extracted in the same manner as in step (1) of the above-described embodiment A-1, and the “p _inf setting unit 501” determines that the moving direction v is equal to the moving direction v, “position p after infinite time has elapsed _inf "is set (E-3α-1).
(Α2) The “scan unit 502 with pixel number i” scans the address i of each point _i p ₀ in the image from the lower limit value i _min to the upper limit value i _max . (E-3α-2, E-3α-3, E-3α-10).

(Scan i)
( _Α3 ) The address j of the element (n _sj , _n d _sj ) in the cylindrical array is scanned from the lower limit value j _min to the upper limit value j _max by the “scan unit 503 of element number j”.

(Scan j)
(Α4) “ _i p ₀ output unit 504”, “n _sj output unit 505”, and “ _n d _sj output unit 509” cause pixel _i p ₀ and cylinder in the image corresponding to these addresses (i, j) The array element (n _sj , _n d _sj ) is output (E-3α-6).
Four parameters _n sj set in (.alpha.5) _{above, n d sj, i p 0} , the _{p inf} enter the _"ij tau table making unit 510 ', motion parallax in the manner shown in Appendix (A-2) _ij τ is calculated (E-3α-7).

The motion parallax _ij τ is not limited to the method shown in Appendix (A-2), and may be calculated by any method. This is the same in all the embodiments described to calculate motion parallax by the method shown in Appendix (A-2).
(Α6) This motion parallax _ij τ is stored as contents corresponding to the address (i, j) of the _ij τ table (FIG. 163) (E-3α-8).

(Scan j (E-3α-9))
(Scan i (E-3α-10))
Thus, the _ij τ table (FIG. 163) is obtained.
(Β) Using the _ij τ table created above, the plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected (see FIGS. 117 (B) and 119).
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 552 of pixel number i” (E-3β-1, E-3β-2) , E-3β-10).

(Scan i)
(Β2) Similar to step (3) in Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 551 by the “local region image extraction unit 554 centered on _i p ₀ ”. Each local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ for the current time t ₀ and the next time t ₁ (E-3β-3).
(Β3) “Address j of element (n _sj , _n d _sj ) in the columnar array is scanned from lower limit value j _min to upper limit value j _max by“ scanning unit 553 of element number j ”(E-3β-4) , E-3β-5, E-3β-9).

(Scan j)
(Β4) These addresses (i, j) are input to the _ij τ table 565 (FIG. 163), and the motion parallax _ij τ is output (E-3β-6).
(Β5) In the same manner as in step (6) of Embodiment D-1, “local image of current time t ₀ and next time t ₁ ” and motion parallax _ij τ are set as “motion parallax detection unit 566” (see FIG. 159). ) And the response intensity is calculated by the following equation (E-3β-7).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- ij τ x, y- ij τ y)
(Β6) Vote this response strength to the elements (n _sj , _n d _sj ) in the “cylindrical array voting unit 567” (E-3β-8).

(Scan j (E-3β-9))
(Scan i (E-3β-10))
(Β7) “Peak extraction unit 568” extracts “the point at which the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, the “normalized shortest distance _n d _s0 until it crosses the plane” is obtained, and as the “coordinate within the cross-sectional circle”, the three-dimensional orientation _{ns 0} of the plane is obtained (E− 3β-11).

Embodiment E-4 (standardized shortest distance + v unknown)
120 is a block diagram showing the configuration of the present embodiment E-4, and FIGS. 121 and 122 are flowcharts showing the operation thereof.
(Α) _ij τ table (FIG. 164) for all moving directions {v} (ie, search and output motion parallax _ij τ from pixel number i of input image, element number j of cylindrical array, and moving direction v) To create a table to do (see FIG. 120 (A), FIG. 121)
The steps from (α2) to (α6) below are the same as the corresponding steps in embodiment E-3 (α).
(Α0) “V-parameter scanning unit 508” scans “movement direction parameter v” in all possible directions (lower limit value v _min to upper limit value v _max ) (E-4α-1, E-4α-2, E-4α-13).

(Scan v)
(Α1) “P _inf setting unit 501” sets “position p _inf after infinite time has elapsed” to be equal to this parameter v (E-4α-3).
(Α2) The “scanning unit 502 of pixel number i” scans the address i of each point _i p ₀ in the image from the lower limit value i _min to the upper limit value i _max (E-4α-4, E-4α-5 , E-4α-12).

(Scan i)
(Α3) The “scan unit 503 of element number j” scans the address j of the element (n _sj , _n d _sj ) in the cylindrical array from the lower limit j _min to the upper limit j _max (E-4α-6, E-4α-7, E-4α-11).

(Scan j)
(Α4) “ _i p ₀ output unit 504”, “n _sj output unit 505”, and “ _n d _sj output unit 509” cause pixel _i p ₀ and cylinder in the image corresponding to these addresses (i, j) The array element (n _sj , _n d _sj ) is output (E-4α-8).
Four parameters _n sj set in (.alpha.5) _above, n _{d sj,} a i _{p 0, p inf} enter the _"ij tau table making unit 510 ', movement in the manner shown in appendix (A-2) The parallax _ij τ is calculated (E-4α-9).
(Α6) This motion parallax _ij τ is stored in the contents corresponding to the address (i, j) in the _ij τ table (FIG. 164) (E-4α-10).

(Scan j (E-4α-11))
(Scan i (E-4α-12))
(Α7) By the processing so far, an _ij τ table for each moving direction v is obtained.

(Scanning v (E-4α-13))
From the above: an _ij τ table (FIG. 164) for all moving directions {v} is obtained.
(Β) Using the _ij τ table created above, the plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected (see FIGS. _120B and 122).
The steps from (β1) to (β6) below are the same as the corresponding steps in embodiment E-3.
(Β0) “V-parameter scanning unit 559” scans the “movement direction parameter v” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (E-4β-1, E-4β-2, E-4β-13)
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 552 of pixel number i” (E-4β-3, E-4β-4 , E-4β-12).

(Scan i)
(Β2) Similar to step (3) in Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 551 by the “local region image extraction unit 554 centered on _i p ₀ ”. A local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (E-4β-5).
(Β3) The “scan unit 553 of element number j” scans the address j of the element (n _sj , _n d _sj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-4β-6, E-4β-7, E-4β-11).

(Scan j)
(Β4) These addresses (i, j) are input to the _ij τ table (FIG. 164), and the motion parallax _ij τ is output (E-4β-8).
(Β5) Similarly to step (6) of Embodiment D-1, “local image of current time t ₀ and next time t ₁ ” and motion parallax _ij τ are expressed as “motion parallax detection unit 566” (FIG. 159). And the response intensity is calculated by the following equation (E-4β-9).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- ij τ x, y- ij τ y)
(Β6) This response intensity is voted for “elements (n _sj , _n d _sj ) in the cylindrical array voting unit 567 (E-4β-10).

(Scan j (E-4β-11))
(Scan i (E-4β-12))
(Β7) In the process so far, voting is performed for all the elements in the cylindrical array.

(Scanning v (E-4β-13))
(Β8) In the process so far, voting is performed for all elements of the cylindrical array group for all the movement direction parameters v.
(Β9) Among the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 568”. As the movement direction parameter for this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 to the plane is set as the “height coordinate” of the point, and 3 of the plane is set as the “coordinate in the cross-sectional circle”. The dimension orientation _ns0 is obtained (E-4β-14).

Embodiment E-5 (stereo + normalized distance)
FIG. 123 is a block diagram showing the configuration of the present embodiment E-5, and FIGS. 124 and 125 are flowcharts showing the operation thereof.
(Α) _ij σ table (FIG. 171) —That is, a table for retrieving and outputting motion parallax _ij σ from the pixel number i of the input image and the element number j of the cylindrical array (FIG. 123 ( A), see FIG. 124)
(Α1) As in the above-described embodiment B-1, the “p _axis setting unit 511” sets the position _paxis on the visual _axis as equal to the visual axis direction _axis (E-5α-1).
(Α2) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 512 of pixel number i” (E-5α-2, E-5α-3 , E-5α-10).

(Scan i)
( _Α3 ) The address j of the element (n _sj , _n d _cj ) in the cylindrical array is scanned from the lower limit value j _min to the upper limit value j _max by “scanning unit 503 of element number j” (E-5α-4, E-5α-5, E-5α-9).

(Scan j)
(Α4) “ _i p _R output unit 514”, “n _sj output unit 515”, and “ _n d _cj output unit 516” correspond to these addresses (i, j) and pixels _i p _R and The cylindrical array elements (n _sj , _n d _cj ) are output (E-5α-6).
Four parameters _n sj set in (.alpha.5) _{above, n d cj, i p R} , by entering the _{p axis} to _"ij sigma table making unit 517 ', both in the method shown in appendix (B-1) The eye parallax _ij σ is calculated (E-5α-7).

The binocular parallax _ij σ is not limited to the method shown in Appendix (B-1), and may be calculated by any method. This point is the same in all the embodiments described to calculate binocular parallax by the method shown in Appendix (B-1).
(Α6) This binocular disparity _ij σ is stored as contents corresponding to the address (i, j) of the _ij σ table (FIG. 171) (E-5α-8).

(Scan j (E-5α-9))
(Scan i (E-5α-10))
Thus, the _ij σ table (FIG. 171) is obtained.
(Β) Using the _ij σ table created above, the plane orientation n _s0 and the normalized distance _n d _c0 are detected (see FIGS. _123B and 125).
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 572 of pixel number i” (E-5β-1, E-5β-2) , E-5β-10).

(Scan i)
The cutout unit 574 "of the local region image centered at _(.beta.2) i _{p R,} from the image obtained by the right camera 561 and the left camera 562, centered on the" address i pixels corresponding to _i p _R " As shown in the left end of FIG. 167, the local area image is cut out for the right and left cameras (E-5β-3).
(Β3) The “scan unit 573 of element number j” scans the address j of the element (n _sj , _n d _cj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-5β-4, E-5β-5, E-5β-9).

(Scan j)
(Β4) These addresses (i, j) are input to the _ij σ table 575 (FIG. 171), and binocular parallax _ij σ is output (E-5β-6).
(Β5) “Local area images of right and left cameras” and binocular parallax _ij σ are input to “binocular parallax detection unit 576” (FIG. 167) in the same manner as in step (6) of embodiment D-5. Then, the response intensity is calculated by the following equation (E-5β-7).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- ij σ x, y- ij σ y)
(Β6) Vote this response intensity to the elements (n _sj , _n d _cj ) in the “cylindrical array voting unit 577” (E-5β-8).

(Scan j (E-5β-9))
(Scan i (E-5β-10))
(Β7) “Peak extraction unit 578” extracts “a point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “a place where the great circles intersect at one point” It is. The “normalized distance _n d _c0 ” to the plane is obtained as the “height coordinate” of this local maximum point, and the three-dimensional orientation n _s0 of the plane is obtained as the “coordinate in the cross-sectional circle” (E-5β-11). ).

Embodiment E-6 (stereo + normalized distance + a _xis unknown)
FIG. 126 is a block diagram showing the configuration of the present embodiment E-6, and FIGS. 127 and 128 are flowcharts showing the operation thereof.
(Α) _ij σ table for all visual axis directions {a _xis } (FIG. 172) _-that is, pixel number i of the input image, element number j of the cylindrical array, and binocular parallax _ij σ from visual axis direction a _xis A table for searching for and outputting data (see FIGS. 126A and 127)
The steps from (α2) to (α6) below are the same as the corresponding steps in embodiment E-5 (α).
(.Alpha.0) by 'scanning unit 518 of _{a xis} Parameters ", over all directions that may (lower limit _{a xis,} from _min limit _{a xis,} until _max), scans the" visual axis direction parameter _{a xis} " (E-6α-1, E-6α-2, E-6α-13).

( _{Scan axis} )
(Α1) The “p _axis setting unit 511” sets the “position p _axis on the visual _axis ” as equal to the parameter a _xis (E-6α-3).
(Α2) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 512 of pixel number i” (E-6α-4, E-6α-5 , E-6α-12).

(Scan i)
(Α3) The “scan unit 513 of element number j” scans the address j of the element (n _sj , _n d _cj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-6α-6, E-6α-7, E-6α-11).

(Scan j)
(Α4) “ _i p _R output unit 514”, “n _sj output unit 515”, and “ _n d _cj output unit 516” correspond to these addresses (i, j) and pixels _i p _R and cylindrical array elements _{(n sj,} n _{d cj)} outputs the (E-6α-8).
Four parameters _n sj set in (.alpha.5) _{above, n d cj, i p R} , by entering the _{p axis} to _"ij sigma table making unit 517 ', both in the method shown in appendix (B-1) The eye parallax _ij σ is calculated (E-6α-9).
(Α6) This binocular disparity _ij σ is stored as content corresponding to the address (i, j) of the _ij σ table (FIG. 172) (E-6α-10).

(Scan j (E-6α-11))
(Scan i (E-6α-12))
(Α7) With the processing so far, an _ij σ table for each visual axis direction a _xis is obtained.

( _{Scan axis} (E-6α-13))
_Thus , the _ij σ table (FIG. 172) for all visual axis directions {a _xis } is obtained.
(Β) Using the _ij σ table created above, the plane orientation n _s0 and the normalized shortest distance _n d _c0 are detected (see FIGS. 126B and 128).
The steps from (β1) to (β6) below are the same as the corresponding steps in embodiment E-5.
( _Β0 ) The “visual axis direction parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scanning unit 579”. (E-6β-1, E-6β-2, E-6β-13).

( _{Scan axis} )
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 572 of pixel number i” (E-6β-3, E-6β-4) , E-6β-12).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 574 of the local region image centered on the "from each image obtained by the right camera 561 and the left camera 562, and around the" pixel _i p _R corresponding to the address i " The local region image to be cut is cut out for the right and left cameras as shown at the left end of FIG. 167 (E-6β-5).
(Β3) The “scan unit 573 of element number j” scans the address j of the element (n _sj , _n d _cj ) in the cylindrical array from the lower limit j _min to the upper limit j _max (E-6β-6, E-6β-7, E-6β-11).

(Scan j)
(Β4) These addresses (i, j) and the visual axis direction a _xis are _input to the _ij σ table 575 (FIG. 172), and binocular parallax _ij σ is output (E-6β-8).
(Β5) “Local area images of right and left cameras” and binocular parallax _ij σ are input to “binocular parallax detection unit 576” (FIG. 167) in the same manner as in step (6) of embodiment D-5. Then, the response intensity is calculated by the following equation (E-6β-9).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- ij σ x, y- ij σ y)
(Β6) Vote this response intensity to the elements (n _sj , _n d _cj ) in the “cylindrical array voting unit 577” (E-6β-10).

(Scan j (E-6β-11))
(Scan i (E-6β-12))
(Β7) In the process so far, voting is performed for all the elements in the cylindrical array.

(A _xis scan (E-6β-13))
(Beta 8) in the processing up to this point, the vote on all the elements of the cylindrical array groups for all of the visual axis direction parameter a _xis performed.
(Β9) The “specific cylinder arrangement” having the maximum intensity is extracted by the “peak extraction unit 578” from the cylinder arrangement group. As the visual axis direction parameter for this array, the true movement direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized distance _n d _c0 to the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (E-6β-14).

Embodiment E-7 (stereo + standardized shortest distance)
FIG. 129 is a block diagram showing the configuration of the present embodiment E-7, and FIGS. 130 and 131 are flowcharts showing the operation.
(Α) _ij σ table (FIG. 171) —that is, a table for searching for and outputting binocular parallax _ij σ from the pixel number i of the input image and the element number j of the cylindrical array (FIG. 129) (See (A), Fig. 130)
(Α1) As in the above-described embodiment B-1, the “p _axis setting unit 511” sets the position p _axis on the visual _axis as equal to the visual axis direction a _xis (E-7α-1).
(Α2) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 512 of pixel number i” (E-7α-2, E-7α-3 , E-7α-10).

(Scan i)
(Α3) The “scan unit 513 of element number j” scans the address j of the element (n _sj , _n d _sj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-7α-4, E-7α-5, E-7α-9).

(Scan j)
(Α4) “ _i p _R output unit 514”, “n _sj output unit 515”, and “ _n d _sj output unit 519” allow the pixels _i p _R in the image corresponding to these addresses (i, j) and The cylindrical array elements (n _sj , _n d _sj ) are output (E-7α-6).
Four parameters _n sj set in (.alpha.5) _{above, n d sj, i p R} , by entering the _{p axis} to _"ij sigma table making unit 520 ', both in the method shown in appendix (B-2) The eye parallax _ij σ is calculated (E-7α-7).

The binocular parallax _ij σ is not limited to the method shown in Appendix (B-2), and may be calculated by any method. This is the same in all the embodiments described to calculate binocular parallax by the method shown in Appendix (B-2).
(Α6) This binocular disparity _ij σ is stored as content corresponding to the address (i, j) of the _ij σ table (FIG. 171) (E-7α-8).

(Scan j (E-7α-9))
(Scan i (E-7α-10))
Thus, the _ij σ table (FIG. 171) is obtained.
(Β) Using the _ij σ table created above, the plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected (see FIGS. 129 (B) and 131).
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 572 of pixel number i” (E-7β-1, E-7β-2) , E-7β-10).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 574 of the local region image centered on the "from each image obtained by the right camera 561 and the left camera 562, and around the" pixel _i p _R corresponding to the address i " The local region image to be cut is cut out for the right and left cameras as shown at the left end of FIG. 167 (E-7β-3).
(Β3) The “scan unit 573 of element number j” scans the address j of the element (n _sj , _n d _sj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-7β-4, E-7β-5, E-7β-9).
(Β4) These addresses (i, j) are input to the _ij σ table 585 (FIG. 171), and binocular parallax _ij σ is output (E-7β-6).
(Β5) “Local area images of right and left cameras” and binocular parallax _ij σ are input to “binocular parallax detection unit 586” (FIG. 167) in the same manner as in step (6) of embodiment D-5. Then, the response intensity is calculated by the following equation (E-7β-7).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- ij σ x, y- ij σ y)
(Β6) Vote this response intensity to the elements (n _sj , _n d _sj ) in the “cylindrical array voting unit 587” (E-7β-8).
(Β7) “Peak extraction unit 588” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. The "normalized shortest distance to the plane _{n d s0"} maximum points as "height coordinate" is obtained, also 3-dimensional orientation n _s0 plane as "the coordinates of the cross section circle" is obtained (E-7.beta. 11).

Embodiment E-8 (stereo + normalized shortest distance + a _xis unknown)
FIG. 132 is a block diagram showing the configuration of the present embodiment E-8, and FIGS. 133 and 134 are flowcharts showing the operation thereof.
(Α) _ij σ table for all visual axis directions {a _xis } (FIG. 172) _-that is, pixel number i of the input image, element number j of the cylindrical array, and binocular parallax _ij σ from visual axis direction a _xis A table for searching for and outputting data (see FIGS. 132A and 133)
The steps from (α2) to (α6) below are the same as the corresponding steps in embodiment E-7 (α).
(.Alpha.0) by 'scanning unit 518 of _{a xis} Parameters ", all directions that may (lower limit _{a xis,} the upper limit _{a xis} from _{min, max)} over scans the" visual axis direction parameter _{a xis} "( E-8α-1, E-8α-2, E-8α-13).

( _{Scan axis} )
(Α1) The “p _axis setting unit 511” sets the “position p _axis on the visual _axis ” to be equal to the parameter a _xis (E-8α-3).
(Α2) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 512 of pixel number i” (E-8α-4, E-8α-5 , E-8α-12).

(Scan i)
(Α3) By “scanning unit 513 of element number j”, address j of element (n _sj , _n d _sj ) in the cylindrical array is scanned from lower limit value j _min to upper limit value j _max (E-8α-6, E-8α-7, E-8α-11).

(Scan j)
(Α4) “ _i p _R output unit 514”, “n _sj output unit 515”, and “ _n d _sj output unit 519” allow the pixels _i p _R in the image corresponding to these addresses (i, j) and The cylindrical array elements (n _sj , _n d _sj ) are output (E-8α-8).
Four parameters _n sj set in (.alpha.5) _{above, n d sj, i p R} , by entering the _{p axis} to _"ij sigma table making unit 520 ', both in the method shown in appendix (B-2) The eye parallax _ij σ is calculated (E-8α-9).
(Α6) This binocular disparity _ij σ is stored as content corresponding to the address (i, j) of the _ij σ table (FIG. 172) (E-8α-10).

(Scan i (E-8α-11))
(Scan j (E-8α-12))
(Α7) With the processing so far, an _ij σ table for each visual axis direction a _xis is obtained.

( _{Scan axis} )
_Thus , the _ij σ table (FIG. 172) for all visual axis directions {a _xis } is obtained.
(Β) The plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected using the _ij σ table created above (see FIGS. _132B and 134).
The steps from (β1) to (β6) below are the same as the corresponding steps in Embodiment D-7.
( _Β0 ) The “visual axis direction parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scanning unit 579”. (E-8β-1, E-8β-2, E-8β-13).

( _{Scan axis} )
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 572 of pixel number i” (E-8β-3, E-8β-4) , E-8β-12).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 574 of the local region image centered on the "from each image obtained by the right camera 561 and the left camera 562, and around the" pixel _i p _R corresponding to the address i " The local region image to be cut is cut out for the right and left cameras as shown at the left end of FIG. 167 (E-8β-5).
(Β3) The “scan unit 573 of element number j” scans the address j of the element (n _sj , _n d _sj ) in the cylindrical array from the lower limit value j _min to the upper limit value j _max (E-8β-6, E-8β-7, E-8β-11).

(Scan j)
(Β4) These addresses (i, j) are input to the _ij σ table 585 (FIG. 172), and binocular parallax _ij σ is output (E-8β-8).
(Β5) “Local area images of right and left cameras” and binocular parallax _ij σ are input to “binocular parallax detection unit 586” (FIG. 167) in the same manner as in step (6) of embodiment D-5. Then, the response intensity is calculated by the following equation (E-8β-9).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- ij σ x, y- ij σ x)
(Β6) Vote this response intensity to the elements (n _sj , _n d _sj ) in the “cylindrical array voting unit 587” (E-8β-10).

(Scan j (E-8β-11))
(Scan i (E-8β-12))
(Β7) In the process so far, voting is performed for all elements in the “cylindrical array voting unit 587”.

( _{Scan axis} (E-8β-13))
(Beta 8) in the processing up to this point, the vote on all the elements of the cylindrical array groups for all of the visual axis direction parameter a _xis performed.
(Β9) Among the above-mentioned columnar array group, a “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 588”. As the visual axis direction parameter for this array, the true movement direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 to the plane is set as the “height coordinate” of the point, and 3 of the plane is set as the “coordinate in the cross-sectional circle”. dimensional orientation _{n s0} is determined (E-8β-14).

Embodiment F
Embodiment F-1
FIG. 135 is a block diagram showing the configuration of the present embodiment F-1, and FIGS. 136 and 137 are flowcharts showing the operation thereof.

Embodiment F-1 (standardization time)
(Α) Converting the _ij τ table (FIG. 163) of Embodiment E-1 into a { _ik j} table (see FIGS. 135 (A) and 136)
(Α1) The _ij τ table 555 (see FIGS. 111 and 163) is created (F-1α-1) by the processing of the embodiment E-1α (see FIG. 112), and the { _ik j} table conversion unit 601 The _ij τ table (that is, motion parallax τ) is rewritten to the motion parallax number k based on the correspondence relationship of FIG. 160 (F-1α-2). As a result, the _ij τ table in FIG. 163 is rewritten to the “ _ij k table (middle stage in FIG. 135A)” with the address (i, j) and the content _ij k.
(Α2) Next, the _ij k table is rearranged by the { _ik j} table conversion unit 601 to create a table with the address (i, k) and the content “element number j of the cylindrical array” (F -1α-3). As described in the above-described embodiment A-6, the arbitrary address (i, k)-that is, the pixel having the position _i p ₀ and the motion parallax _k τ--is converted into a “cylindrical array” by cross ratio conversion and polar conversion In other words, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 165).
(Α3) As described above, when an arbitrary address (i, k) is designated, a { _ik j} table (FIG. 165) in which an element number group { _ik j} of a cylindrical array is output can be created.
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized time _n t _c0 are detected (see FIGS. 135B and 137).
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 652 of pixel number i” (F-1β-1, F-1β-2) , F-1β-12).

(Scan i)
(Β2) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 651 by the “local region image cutting unit 654 centered on _i p ₀ ” Each local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (F-1β-3).
(Β3) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scan unit 653 of motion parallax number k” (F-1β-4, F-1β-5, F-1β-11). ).

(Scan k)
(Beta4) by 'motion parallax _k tau conversion unit 655 ", the number k, in the same manner as in Embodiment D-1 step (5) into motion parallax _{k τ (F-1β-6} ). However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Step (β7) is skipped (F-1β-7) due to motion parallax that contradicts the moving direction v.
(Β5) The “local region image at the current time t ₀ and the next time t ₁ ” and the motion parallax _k τ set above are input to the “motion parallax detection unit 656” (FIG. 159), and the embodiment D-1 In the same manner as in step (6), the response intensity is calculated by the following equation (F-1β-8).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(Β6) The address (i, k) is input to the { _ik j} table 602, and the element number group { _ik j} of the cylindrical array is output (F-1β-9).
(Β7) Vote the response intensity calculated in step (β5) to the element number group of the cylinder array corresponding to this number group { _ik j} in the “cylindrical array voting unit 657” (F-1β-10) .

(Scan k (F-1β-11))
(Scan i (F-1β-12))
(Β8) “Peak extraction unit 658” extracts “the point at which the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the great circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “normalized time _n t _c0 until crossing the plane” is obtained, and as the “coordinate in the cross-sectional circle”, the three-dimensional orientation n _s0 of the plane is obtained (F−1β). -13).

Embodiment F-2 (standardized time + v unknown)
FIG. 138 is a block diagram showing the configuration of the present embodiment F-2, and FIGS. 139 and 140 are flowcharts showing the operation thereof.
(Α) Creation of { _ik j} table (FIG. 166) for all moving directions {v} (see FIGS. 138A and 139)
(Α1) The _ij τ table 555 (see FIGS. 114 and 164) is created (F-2α-1) by the processing of the embodiment E-2α (see FIG. 115), and the { _ik j} table conversion unit 601 The _ij τ table 555 (that is, motion parallax τ) is rewritten to the motion parallax number k based on the correspondence relationship in FIG. 160 (F-2α-2). As a result, the _ij τ table in FIG. 164 is rewritten to “ _ij k table (middle stage in FIG. 138 (A))” with the address (i, j) and the content _ij k.
(Α2) Next, the _ij k table is rearranged by the { _ik j} table conversion unit 601 to create a table with the address (i, k) and the content “element number j of the cylindrical array” (F -2α-3). As described in the above-described embodiment A-6, the arbitrary address (i, k)-that is, the pixel having the position _i p ₀ and the motion parallax _k τ--is converted into a “cylindrical array” by cross ratio conversion and polar conversion. In other words, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 166).
(Α3) As described above, when an arbitrary address (i, k) and a moving direction v are designated, a { _ik j} table (FIG. 166) in which an element number group { _ik j} of a cylindrical array is output can be created.
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized time _n t _c0 are detected (see FIGS. 138 (B) and 140).
The steps from (β1) to (β7) below are the same as the corresponding steps in Embodiment F-1.
(Β0) “V-parameter scanning unit 659” scans “movement direction parameter v” in all possible directions (from lower limit value v _min to upper limit value v _max ) (F-2β-1, F-2β-2, F-2β-15).

(Scan v)
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 652 of pixel number i” (F-2β-3, F-2β-4 , F-2β-14).

(Scan i)
(Β2) Similar to step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 654 by the “local region image cutting unit 654 centered on _i p ₀ ” Each local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (F-2β-5).
(Β3) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scan unit 653 of motion parallax number k” (F-2β-6, F-2β-7, F-2β-13). ).

(Scan k)
(Beta4) in the same manner as in Embodiment E-1 of step (5), the "motion parallax _k tau conversion unit 653" and converts the number k to the motion parallax _{k τ (F-2β-8} ). However, when the motion parallax _k τ, that is, the direction of the vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Step (β7) is skipped (F-2β-9) due to motion parallax inconsistent with direction v.
(Β5) The “local region image at the current time t ₀ and the next time t ₁ ” and the motion parallax _k τ set above are input to the “motion parallax detection unit 656” (see FIG. 159), and the embodiment D- In the same manner as in step (6) of 1, the response intensity is calculated by the following equation (F-2β-10).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(Β6) The address (i, k) and the moving direction v are input to the { _ik j} table 602, and the element number group { _ik j} of the cylindrical array is output (F-2β-11).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to this number group { _ik j} in the “column array voting unit 657” (F-2β-12).

(Scan k (F-2β-13))
(Scan i (F-2β-14))
(Β8) In the process so far, voting is performed for all elements in the cylindrical array.

(Scanning v (F-2β-15))
(Β9) In the process so far, voting is performed for all elements in the cylindrical array for all the moving direction parameters v.
(Β10) Among the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 658”. As the movement direction parameter corresponding to this array, the true movement direction v ₀ is obtained. Further, when a point at which the intensity reaches a peak in the array is extracted, the normalization time _n t _c0 until the plane is crossed as the “height coordinate” of the point is obtained, and the coordinate of the plane is obtained as the “coordinate in the cross-sectional circle”. A three-dimensional orientation _ns0 is obtained (F-2β-16).

Embodiment F-3 (standardized shortest distance)
FIG. 141 is a block diagram showing the configuration of the present embodiment F-3, and FIGS. 142 and 143 are flowcharts showing the operation thereof.
(Α) Converting the _ij τ table (FIG. 163) of Embodiment E-3 into a { _ik j} table (see FIGS. 141 (A) and 142)
(Α1) The contents 565 (see FIGS. 117 and 163) of the _ij τ table are created (F-3α-1) by the processing of the embodiment E-3α (see FIG. 118), and the { _ik j} table conversion unit 611. Thus, the _ij τ table (that is, motion parallax τ) is rewritten to motion parallax number k based on the correspondence relationship of FIG. 160 (F-3α-2). As a result, the _ij τ table in FIG. 163 is rewritten to the “ _ij k table (middle stage in FIG. 141A)” with the address (i, j) and the content _ij k.
([Alpha] 2) Next, by their _{ik j} table conversion unit 611, and reordering the _ij k table address (i, k) contents to create a table of "element number j of the column sequence" (F -3α-3). As described in the above-described embodiment A-8, the arbitrary address (i, k)-that is, the pixel with the position _i p ₀ and the motion parallax _k τ- Since all the above points are connected to each other, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 165).
(Α3) As described above, when an arbitrary address (i, k) is designated, a { _ik j} table (FIG. 165) in which an element number group { _ik j} of a cylindrical array is output can be created.
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected (see FIGS. 141B and 143).
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 652 of pixel number i” (F-3β-1, F-3β-2) , F-3β-12).

(Scan i)
(Β2) In the same manner as in step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 651 by the “local region image cutting unit 654 centered on _i p ₀ ” A local region image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (F-3β-3).
(Β3) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “motion parallax k scanning unit 653” (F-3β-4, F-3β-5, F-3β-11). .

(Scan k)
(Β4) The “motion parallax _k τ conversion unit 665” converts this number k into the motion parallax _k τ in the same manner as in step (5) of Embodiment D-1 (F-3β-6). However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Step (β7) is skipped (F-3β-7) due to motion parallax that contradicts the moving direction v.
(Β5) The “local region image at the current time t ₀ and the next time t ₁ ” and the motion parallax _k τ set above are input to the “motion parallax detection unit 666” (FIG. 159), and the embodiment D-1 In the same manner as in step (6), the response intensity is calculated by the following equation (F-3β-8).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(Β6) The address (i, k) is input to the { _ik j} table 612, and the element number group { _ik j} of the cylindrical array is output (F-3β-9).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to the number group { _ik j} in the “column array voting unit 667” (F-3β-10).

(Scan k (F-3β-11))
(Scan i (F-3β-12))
(Β8) “Peak extraction unit 668” extracts “a point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “a place where small circles intersect at one point” It is. As the “height coordinate” of this local maximum point, “the normalized shortest distance _n d _s0 to the plane” is obtained, and as the “coordinate in the cross-section circle”, the three-dimensional orientation _{ns 0} of the plane is obtained (F-3β−). 13).

Embodiment F-4 (standardized shortest distance + v unknown)
FIG. 144 is a block diagram showing the configuration of the present embodiment F-4, and FIGS. 145 and 146 are flowcharts showing the operation thereof.
(Α) The _ik τ table (FIG. 164) of the embodiment E-4 is converted into the { _ik j} table (see FIGS. 144 (A) and 145).
(Α1) The _ij τ table 565 (see FIGS. 120 and 164) is created (F-4α-1) by the processing of the embodiment E-4α (see FIG. 121), and the { _ik j} table conversion unit 611 The _ij τ table (that is, motion parallax τ) is rewritten to the motion parallax number k based on the correspondence relationship in FIG. 160 (F-4α-2). As a result, the _ij τ table in FIG. 164 is rewritten to the “ _ij k table (middle stage in FIG. 144 (A))” with the address (i, j) and the content _ij k.
([Alpha] 2) Next, by their _{ik j} table conversion unit 611, and reordering the _ij k table address (i, k) contents to create a table of "element number j of the column sequence" (F -4α-3). As described in the above-described embodiment A-8, the arbitrary address (i, k)-that is, the pixel having the position _i p ₀ and the motion parallax _k τ-- Since all the points on the circle are connected, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 166).
(Α3) As described above, when the arbitrary address (i, k) and the moving direction v are designated, the { _ik j} table 612 (FIG. 166) in which the element number group { _ik j} of the cylindrical array is output can be created.
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected (see FIGS. 144B and 146).
The steps from (β1) to (β7) below are the same as the corresponding steps in Embodiment F-3.
(Β0) “V-parameter scanning unit 659” scans the “movement direction parameter v” in all possible directions (from the lower limit value v _min to the upper limit value v _max ) (F-4β-1, F-4β-2, F-4β-15).

(Scan v)
(Β1) The address i of each point _i p ₀ in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 652 of pixel number i” (F-4β-3, F-4β-4) , F-4β-14).

(Scan i)
(Β2) Similarly to step (3) of Embodiment D-1, the current time t ₀ and the next time t obtained by the camera 651 by the “local region image extraction unit 654 centered on _i p ₀ ” A local area image centered on “pixel _i p ₀ corresponding to address i” is cut out from each image of ₁ at the current time t ₀ and the next time t ₁ (F-4β-5).
(Β3) The motion parallax number k is scanned from the lower limit value k _min to the upper limit value k _max by the “scanning unit 653 of motion parallax number k” (F-4β-6, F-4β-7, F-4β-13). ).

(Scan k)
(Β4) The “motion parallax _k τ conversion unit 665” converts this number k into the motion parallax _k τ in the same manner as in step (5) of Embodiment D-1 (F-4β-8). However, when the motion parallax _k τ, that is, the direction of the motion vector ( _k τ _{x k} τ _y ) is different from the “direction from _i p ₀ to v (ie, p _inf )” in FIG. Step (β7) is skipped (F-4β-9) because of motion parallax contradicting the moving direction v.
(Β5) The “local region image at the current time t ₀ and the next time t ₁ ” and the motion parallax _k τ set above are input to the “motion parallax detection unit 666” (FIG. 159), and the embodiment D-1 In the same manner as in step (6), the response intensity is calculated by the following equation (F-4β-10).

Response intensity _{= Σ x Σ y i a 0} (x, y) i a 1 (x- k τ x, y- k τ y)
(Β6) The address (i, k) and the moving direction v are input to the { _ik j} table 612, and the element number group { _ik j} of the cylindrical array is output (F-4β-11).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to the number group { _ik j} in the “column array voting unit 667” (F-4β-12).

(Scan k (F-4β-13))
(Scan i (F-4β-14))
(Β8) In the process so far, voting is performed for all elements in the cylindrical array.

(Scanning v (F-4β-15))
(Β9) In the process so far, voting is performed for all elements in the cylindrical array for all the moving direction parameters v.
(Β10) Among the above-mentioned columnar array group, a “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 668”. As the movement direction parameter corresponding to this array, the true movement direction v ₀ is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 until the plane crosses the plane as the “height coordinate” of the point, and the plane as the “coordinate in the cross-sectional circle” 3D orientation _{n s0} is calculated (F-4β-16).

Embodiment F-5 (stereo + normalized distance)
FIG. 147 is a block diagram showing the configuration of the present embodiment F-5, and FIGS. 148 and 149 are flowcharts showing the operation thereof.
(Α) Converting the _ij σ table (FIG. 171) of Embodiment E-5 into a { _ik j} table (see FIGS. 147 (A) and 145).
(Α1) The _ij σ table 575 (see FIGS. 123 and 171) is created (F-5α-1) by the processing of the embodiment E-5α (see FIG. 124), and “{ _ik j} table conversion unit 621”. Thus, the _ij σ table (that is, binocular parallax σ) is rewritten to the binocular parallax number k based on the correspondence relationship of FIG. 168 (F-5α-2). As a result, the _ij σ table of FIG. 171 is rewritten to “ _ij k table (middle stage of FIG. 147 (A))” with the address (i, j) and the content _ij k.
(Α2) Next, the { _ik j} table conversion unit 621 ”rearranges this _ij k table to create a table 622 whose address is (i, k) and whose content is“ element number j of a cylindrical array ”. (F-5α-3). As described in the above-described embodiment B-6, the arbitrary address (i, k)-that is, the pixel having the position _i p _R and the binocular parallax _k σ- Since all the points on the great circle in the array are connected ”, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 173).
(Α3) As described above, when an arbitrary address (i, k) is designated, a { _ik j} table (FIG. 173) in which an element number group { _ik j} of a cylindrical array is output can be created.
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized distance _n d _c0 are detected (see FIGS. 147 (B) and 149).
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 672 with pixel number i” (F-5β-1, F-5β-2) , F-5β-12).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 674 of the local region image centered on the "from each image obtained by the right camera 661 and the left camera 662, and around the" pixel _i p _R corresponding to the address i " The local region image to be cut is cut out for the right and left cameras as shown at the left end of FIG. 167 (F-5β-3).
(Β3) “Binocular parallax number k scanning unit 673” scans binocular parallax number k from lower limit value k _min to upper limit value k _max (F-5β-4, F-5β-5, F-5β -11).

(Scan k)
(Β4) “Binocular parallax _k σ conversion unit 675” converts this number k into binocular parallax _k σ in the same manner as in step (5) of Embodiment D-5 (F-5β-6). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (β7) (F-5β- 7).
(Β5) The “local region images of the right and left cameras” and the binocular parallax _k σ set as described above are input to the “binocular parallax detection unit 676” (FIG. 167), and the steps of Embodiment D-5 ( In the same manner as 6), the response intensity is calculated by the following equation (F-5β-8).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(Β6) The address (i, k) is input to the { _ik j} table 622, and the element number group { _ik j} of the cylindrical array is output (F-5β-9).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to the number group { _ik j} in the “column array voting unit 677” (F-5β-10).

(Scan k (F-5β-11))
(Scan i (F-5β-12))
(Β8) “Peak extraction unit 678” extracts “a point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “a place where the great circles intersect at one point” It is. The maxima as "height coordinate""normalized shortest distance to the plane _{n d c0"} is obtained, also 3-dimensional orientation n _s0 plane is obtained as "coordinates in the cross-sectional circle" (F-5.beta. 13).

Embodiment F-6 (stereo + normalized distance + a _axis unknown)
FIG. 150 is a block diagram showing the configuration of the present embodiment F-6, and FIGS. 151 and 152 are flowcharts showing the operation thereof.
(Α) _Creation of { _ik j} table (FIG. 166) for all visual axis directions {a _xis } (see FIGS. _150A and 151)
(Α1) The _ij σ table 575 (see FIGS. 126 and 172) is created by the processing of the embodiment E-6α (see FIG. 127) (F-6α-1), and the { _ik j} table conversion unit 621 The _ij σ table (that is, binocular parallax σ) is rewritten to the binocular parallax number k based on the correspondence relationship in FIG. 168 (F-6α-2). As a result, the _ij σ table in FIG. 172 is rewritten to “ _ij k table (middle stage in FIG. 150A)” with the address (i, j) and the content _ij k.
(Α2) Next, the { _ik j} table conversion unit 621 rearranges the _ij k table to create a table whose address is (i, k) and whose content is “element number j of cylindrical array” (F -6α-3). As described in the above-described embodiment B-6, the arbitrary address (i, k)-that is, the pixel having the position _i p _R and the binocular parallax _k σ-- Since the element numbers j are connected to all points on the great circle of the cylinder array, the element numbers j are _collected and represented by an element number group { _ik j} (see FIG. 174).
(Α3) As described above, when an arbitrary address (i, k) and a visual axis direction a _xis are designated, a { _ik j} table (FIG. 174) in which an element number group { _ik j} of a cylindrical array is output can be created. .
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized distance _n d _c0 are detected (see FIGS. 150B and 152).
The steps from (β1) to (β7) below are the same as the corresponding steps in Embodiment F-5.
( _Β0 ) The “visual axis parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scanning unit 679”. (F-6β-1, F-6β-2, F-6β-15).

( _{Scan axis} )
(Β1) The “scanning unit 672 with pixel number i” scans the address i of each point _i p _R in the image from the lower limit value i _min to the upper limit value i _max (F-6β-3, F-6β-4) , F-6β-14).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 674 of the local region image centered on the "from each image obtained by the right camera 661 and the left camera 662, and around the" pixel _i p _R corresponding to the address i " The local region image to be cut is cut out for the right and left cameras as shown at the left end of FIG. 167 (F-6β-5).
(.Beta.3) by 'scanning unit 673 of the binocular parallax numbers k ", the binocular parallax number k scanning from the lower limit value _{k min} until the lower limit value _{k max (F-6β-6} , F-6β-7, F-6β -13).
(Β4) In the same manner as in step (5) of Embodiment D-5, the “binocular parallax _k σ conversion unit 675” converts this number k into binocular parallax _k σ (F-6β-8). . However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (β7) (F-6β- 8).
(Β5) The “local region images of the right and left cameras” and the binocular parallax _k σ set as described above are input to the “binocular parallax detection unit 676” (FIG. 167), and the steps of Embodiment D-5 ( In the same manner as in 6), the response intensity is calculated by the following equation (F-6β-10).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(Β6) The address (i, k) and the visual axis direction a _xis are input to the { _ik j} table 622, and the element number group { _ik j} of the cylindrical array is output (F-6β-11).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to the number group { _ik j} in the “column array voting unit 677” (F-6β-12).

(Scan k (F-6β-13))
(Scan i (F-6β-14))
(Β8) In the process so far, voting is performed for all elements in the cylindrical array.

(Convert _axis (F-6β-15))
(Β9) In the processes so far, voting is performed for all elements in the cylindrical array for all visual axis direction parameters a _xis .
(Β10) Among the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 678”. As the visual axis direction parameter corresponding to this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized distance _n d _c0 to the plane is used as the “height coordinate” of the point, and the three-dimensional plane is used as the “coordinate in the cross-sectional circle”. The direction _ns0 is obtained (F-6β-16).

Example F-7 (stereo + standardized shortest distance)
FIG. 153 is a block diagram showing the configuration of the present embodiment F-7, and FIGS. 154 and 155 are flowcharts showing the operation thereof.
(Α) Converting the _ij σ table (FIG. 171) of Embodiment E-7 into a { _ik j} table (see 153 (A) and FIG. 154)
(Α1) The _ij σ table 585 (see FIGS. 129 and 171) is created by the processing of the embodiment E-7α (see FIG. 130), and (F-7α-1), “{ _ik j} table conversion unit 631. The _ij σ table (that is, the binocular parallax σ) is rewritten to the binocular parallax number k based on the correspondence relationship in FIG. 168 (F-7α-2). As a result, the _ij σ table of FIG. 171 is rewritten to “ _ij k table (middle stage of FIG. 153A)” with the address (i, j) and the content _ij k.
(Α2) Next, the “{ _ik j} table conversion unit 631” rearranges the _ij k table to create a table 632 having the address (i, k) and the content “element number j of the cylindrical array”. (F-7α-3). As described in the above-described embodiment B-8, the arbitrary address (i, k)-that is, the pixel having the position _i p _R and the binocular parallax _k σ-- Since all the points on the small circle are connected to each other, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 173).
(Α3) As described above, when an arbitrary address (i, k) is designated, a { _ik j} table (FIG. 173) in which an element number group { _ik j} of a cylindrical array is output can be created.
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized distance _n d _s0 are detected (see FIGS. 153 (B) and 155).
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 672 with pixel number i” (F-7β-1, F-7β-2) , F-7β-12).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 674 of the local region image centered on the "from each image obtained by the right camera 661 and the left camera 662, and around the" pixel _i p _R corresponding to the address i " As shown at the left end of FIG. 167, the local region image to be cut out is cut out for the right and left cameras (F-7β-3).
(Β3) “Binocular parallax number k scanning unit 673” scans binocular parallax number k from lower limit value k _min to upper limit value k _max (F-7β-4, F-7β-5, F-7β -11).

(Scan k)
(Β4) “Binocular parallax _k σ conversion unit 685” converts this number k into binocular parallax _k σ in the same manner as in step (5) of Embodiment D-5 (F-7β-6). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (β7) (F-7β- 7).
(Β5) The “local region images of the right and left cameras” and the binocular parallax _k σ set as described above are input to the “binocular parallax detection unit 686” (FIG. 167), and the steps of Embodiment D-5 ( In the same manner as in 6), the response intensity is calculated by the following equation (F-7β-8).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(Β6) The address (i, k) is input to the { _ik j} table 632, and the element number group { _ik j} of the cylindrical array is output (F-7β-9).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to the number group { _ik j} in the “column array voting unit 687” (F-7β-10).

(Scan k (F-7β-11))
(Scan i (F-7β-12))
(Β8) “Peak extraction unit 688” extracts “the point where the voted intensity becomes a maximum (peak)” in the cylindrical array—this maximum point is “the place where the small circles intersect at one point” It is. The maxima as "height coordinate""normalized shortest distance to the plane _{n d s0"} is obtained, also 3-dimensional orientation n _s0 plane is obtained as "coordinates in the cross-sectional circle" (F-7.beta. 13).

Embodiment F-8 (stereo + standardized shortest distance + a _axis unknown)
FIG. 156 is a block diagram showing the configuration of the present embodiment F-8, and FIGS. 157 and 158 are flowcharts showing the operation thereof.
(Α) The _ik σ table (FIG. 172) of Embodiment E-8 is converted into a { _ik j} table (see FIGS. 156 (A) and 157).
(Α1) The _ij σ table 585 (see FIGS. 132 and 172) is created by the processing of the embodiment E-8α (see FIG. 133), and (F-8α-1), the { _ik j} table conversion unit 631 ” Thus, the _ij σ table (that is, binocular parallax σ) is rewritten to the binocular parallax number k based on the correspondence relationship in FIG. 168 (F-8α-2). As a result, the _ij σ table of FIG. 172 is rewritten to “ _ij k table (middle stage of FIG. 156 (A))” with the address (i, j) and the content _ij k.
(Α2) Next, the { _ik j} table conversion unit 631 ”sorts this _ij k table to create a table 632 whose address is (i, k) and whose content is“ element number j of cylindrical array ”. (F-8α-3). As described in the above-mentioned embodiment B-8, the arbitrary address (i, k)-that is, the pixel having the position _i p _R and the binocular parallax _k σ-- Since all the points on the small circle are connected, the element number j is a set and is represented by an element number group { _ik j} (see FIG. 174).
(Α3) As described above, when an arbitrary address (i, k) and a visual axis direction a _xis are designated, a { _ik j} table (FIG. 174) in which an element number group { _ik j} of a cylindrical array is output can be created. .
(Β) Using the { _ik j} table created above, the plane orientation n _s0 and the normalized shortest distance _n d _s0 are detected (see FIGS. 156B and 158).
The steps from (β1) to (β7) below are the same as the corresponding steps in Embodiment F-7.
( _Β0 ) The “visual axis parameter a _xis ” is scanned in all possible directions (from the lower limit value a _{xis, min} to the upper limit value a _{xis, max} ) by the “a _xi parameter scanning unit 679”. (F-8β-1, F-8β-2, F-8β-15).

( _{Scan axis} )
(Β1) The address i of each point _i p _R in the image is scanned from the lower limit value i _min to the upper limit value i _max by “scanning unit 672 with pixel number i” (F-8β-3, F-8β-4). , F-8β-14).

(Scan i)
(.Beta.2) by _'i p _R cutout unit 674 of the local region image centered on the "from each of the images obtained by the right camera 661 and the left camera 662, and around the" pixel _i p _R corresponding to the address i " The local region image to be cut is cut out for the right and left cameras as shown at the left end of FIG. 167 (F-8β-5).
(Β3) “Binocular parallax number k scanning unit 673” scans binocular parallax number k from lower limit value k _min to upper limit value k _max (F-8β-6, F-8β-7, F-8β -13).

(Scan k)
(Β4) “Binocular parallax _k σ conversion unit 685” converts this number k into binocular parallax _k σ in the same manner as in step (5) of Embodiment D-5 (F-8β-8). However, when the binocular parallax _k σ, that is, the direction of the parallax vector ( _k σ _{x k} σ _y ) is different from the “direction from _i p _R to a _xis (that is, p _axis )” in FIG. since the binocular parallax inconsistent with this visual axis direction _{a xis,} skip to step (β7) (F-8β- 9).
(Β5) The “local region images of the right and left cameras” and the binocular parallax _k σ set as described above are input to the “binocular parallax detection unit 686” (FIG. 167), and the steps of Embodiment D-5 ( Similar to 6), the response intensity is calculated by the following equation (F-8β-10).

Response intensity _{= Σ x Σ y i a R} (x, y) i a L (x- k σ x, y- k σ y)
(Β6) The address (i, k) and the visual axis direction a _xis are input to the { _ik j} table 632, and the element number group { _ik j} of the cylindrical array is output (F-8β-11).
(Β7) The response intensity calculated in step (β5) is voted for the element group of the cylinder array corresponding to the number group { _ik j} in the “column array voting unit 687” (F-8β-12).

(Scan k (F-8β-13))
(Scan i (F-8β-14))
(Β8) In the process so far, voting is performed for all elements in the cylindrical array.

( _{Scan axis} (F-8β-15))
(Β9) In the processes so far, voting is performed for all elements in the cylindrical array for all visual axis direction parameters a _xis .
(Β10) Among the above-described columnar array group, the “specific columnar array” having the maximum intensity is extracted by the “peak extraction unit 688”. As the visual axis direction parameter corresponding to this arrangement, the true visual axis direction _axis0 is obtained. Further, when a point where the intensity reaches a peak in the array is extracted, the normalized shortest distance _n d _s0 to the plane is set as the “height coordinate” of the point, and 3 of the plane is set as the “coordinate in the cross-sectional circle”. dimensional orientation _{n s0} is calculated (F-8β-16).
Appendix Method for calculating motion parallax and binocular parallax (A) Motion parallax (A-1) Normalization time _n t _c , plane orientation n _s , current time position p ₀ , infinite time position p _inf (moving direction v), Two methods for calculating the motion parallax τ are described below.

(A-1-1) Method 1: Method by cross ratio and polar conversion (FIG. 161)
Step 1: Normalization time _n t _c , plane orientation n _s , current time position p ₀ , and infinite time position p _inf are set.

Step 2: Since determining p _c of the _{p c} is the intersection of and _{"p 0} and the great circle passing through the _{p inf"} "Polar of _{n s"} (FIG. 161 (A)), is determined by the following equation. Incidentally, in the following equation "vector product of _{n s"} is due to polar transformation of the above 1.3.2.

p _c = [[p ₀ × p _inf ] × n _s ] / | [p ₀ × p _inf ] × n _s | (Appendix-1a)
Step 3: Determine τ or cos ⁻¹ (p ₀ p ₁ )
_n t _{c, p} inf, during _{p 0,} p _1, p _c is
_{n t c = {p inf p} 0 p 1 p c} ( Appendix -2a)
(See equation (12a)). Expressing p _inf , p ₀ , p ₁ , and _pc as central angles (1.3.1 as described above, FIG. 161 (B)) and substituting them into the formula (Appendix-2a), the following formula is obtained (formula (16a )reference).

_n t _c = (sin (a + τ) / sin (τ))
/ (Sin (x) / sin (x-a)) (Appendix-2b)
Solving this equation for τ, τ = tan ⁻¹ (sina sin (x−a)
_{/ (N t c sinx-cosa} sin (x-a)) ( Appendix -3)
The motion parallax τ was determined. Here, a and x are calculated by the following equations.

a = cos ⁻¹ (p _inf p ₀ ) (Appendix-4a)
_{^{x = cos -1 (p inf p}} c) ( Appendix -4b)
It should be noted that, in case you want to find the next time position p ₁ is,
p ₁ = cosτξ + sinτη (Appendix-5a)
Can be calculated. ξ and η are given by the following equations.

ξ = p ₀ (Appendix-5b)
η = [[p ₀ × p _inf ] × p ₀ ] / | [p ₀ × p _inf ] × p ₀ | (Appendix-5c)
(A-1-2) Method 2: Alternative method It can also be determined by the following method.

Step 1: Normalization time _n t _c , plane orientation n _s , current time position p ₀ , and infinite time position p _inf are set.

Step 2: Arbitrary unit movement distance Δx is set, and the distance d _c from the camera center to the plane—this distance is Vt _{c in} FIG.

d _c = _n t _c Δx (Appendix -6)
Step 3: orientation distance determines the plane of the _{d c} in _{n s.} A point P ₀ extending from the camera center in the direction of p ₀ and intersecting with the plane is calculated.

Step 4: Calculate a point P ₁ that has moved the point P ₀ in the direction of p _inf by Δx.

Step 5: p _{1 obtained} by normalizing the point P ₁ with the following equation is a position on the spherical surface at the next time.

p ₁ = P ₁ / | P ₁ | (Appendix-7)
Step 6: can determine a motion parallax τ in the following equation and the position p ₁ from the current time position p _0. Even if the unit movement distance Δx arbitrarily set in step 2 is changed, this motion parallax is not affected.

τ = cos ⁻¹ (p ₀ p ₁ ) (Appendix-8)
(A-2) normalized shortest distance _n d _s when normalized shortest distance _n d _s, planar orientation _{n s,} the current time position _{p 0,} an infinite time position _{p inf} (movement direction v), calculates the motion parallax τ Two types of methods are described below.

(A-2-1) Method 1: Method by small circle transformation (FIG. 162)
Step 1: Standardized shortest distance _n d _s , plane orientation n _s , current time position p ₀ , and infinite time position p _inf are set.

  Step 2: The radius R of the small circle transformation is determined by the following equation (2.2 described above, FIG. 162 (A)).

R = cos ⁻¹ (p ₀ n _s ) (Appendix-11)
Step 3: Determination of τ, that is, cos ⁻¹ (p ₀ p ₁ ) As described in 2.2, there is a relationship of the following equation (FIG. 162 (B)).
_{cosR = n d s / (p} inf p 0 p 1)
_{= N d s sinτ / sin (} a + τ) ( Appendix -12)
Solving this equation for ^{τ τ = tan -1 (sina cosR} / (n d s -cosa cosR))
(Appendix-13)
The motion parallax τ was determined.

Incidentally, if you want to find the next time position p ₁ can be calculated by the following equation as for formula (Appendix -5).

p ₁ = cosτξ + sinτη (Appendix-14)
(A-2-2) Method 2: Alternative method It can also be determined by the following method.

Step 1: Standardized shortest distance _n d _s , plane orientation n _s , current time position p ₀ , and infinite time position p _inf are set.

Step 2: An arbitrary unit moving distance Δx is set, and the shortest distance d _s (FIG. 12) from the camera center to the plane is calculated by the following equation.

d _s = _n d _s Δx (Appendix-15)
Step 3: orientation shortest distance _{n s} is determined that the plane of the _{d s.} A point P ₀ extending from the camera center in the direction of p ₀ and intersecting with the plane is calculated.

Step 4: Calculate a point P ₁ that has moved the point P ₀ in the direction of p _inf by Δx.

Step 5: p _{1 obtained} by normalizing the point P ₁ with the following equation is a position on the spherical surface at the next time.

p ₁ = P ₁ / | P ₁ | (Appendix-16)
Step 6: can determine a motion parallax τ in the following equation and the position p ₁ from the current time position p _0. Even if the unit movement distance Δx arbitrarily set in step 2 is changed, this motion parallax is not affected.

τ = cos ⁻¹ (p ₀ p ₁ ) (Appendix-17)
(B) the binocular disparity (B-1) normalized distance _n d when the normalized distance _{c n} d _c, planar orientation _{n s,} the right camera image position _{p R,} position on the visual axis connecting the left and right cameras _{p axis} ( Two methods for calculating the binocular parallax σ from the moving direction a _xis ) are described below.

(B-1-1) Method 1: Method by cross ratio and polar conversion (FIG. 169)
Step 1: Normalized distance _n d _c , plane orientation n _s , right camera image position p _R , and position p _axis on the visual axis connecting the left and right cameras are set.

Step 2: Since determining p _c of the _{p c} is the intersection of the the _{"p R} and the great circle passing through the _{p axis" "n s} of polar" (FIG. 169 (A)), is determined by the following equation. The “vector product with n _s ” in the following expression is caused by the aforementioned 4.2.3 polar transformation.

p _c = [[p _R × p _axis ] × n _s ] / | [p _R × p _axis ] × n _s | (Appendix-21a)
Step 3: Determine σ or cos ⁻¹ (p _R p _L )
_{n d c, p axis, p} R, p L, between the _{p c} is
_{n d c = {p axis p} R p L p c} ( Appendix -22A)
(See 4.2). _{p axis, p R, p L} , represents the center angle _{p c} (the aforementioned 4.2, FIG. 169 (B)), becomes the following equation is substituted into Equation (Appendix -22A) (formula (60a) see ).

_{n d c = (sin (c} + σ) / sin (σ))
/ (Sin (x) / sin (x-c)) (Appendix-22b)
When this equation is solved for τ, σ = tan ⁻¹ (sinc sin (x−c)
_{/ (N d c sinx-cosc} sin (x-c)) ( Appendix -23)
Thus, the binocular parallax σ was determined. Here, c and x are calculated by the following equations.

c = cos ⁻¹ (p _axis p _R ) (Appendix -24a)
_{^{x = cos -1 (p axis p}} c) ( Appendix -24b)
If you want to find the position p _L in the left camera image,
p _L = cosσξ + sinση (Appendix-25a)
Can be calculated. ξ and η are given by the following equations.

ξ = p _R (Appendix-25b)
_{η = [[p R × p} axis] × p R] / | [p R × p axis] × p R | ( Appendix -25c)
(B-1-2) Method 2: Alternative method It can also be determined by the following method.

Step 1: Normalized distance _n d _c , plane orientation n _s , right camera image position p _R , and position p _axis on the visual axis connecting the left and right cameras are set.

Step 2: Set the distance [Delta] x _LR between any of the right and left cameras, it computes the distance d _{c (FIG.} 22) by the following equation from the center of the camera to the plane.

d _c = _n d _c Δx _LR (Appendix -26)
Step 3: orientation distance determines the plane of the _{d c} in _{n s.} Extending from the camera center in the direction of the p _R, to calculate the P _R point of intersection with the plane.

Step 4: The point _{P R} to calculate the direction [Delta] x _LR only in moved _{P L} of _{p axis.}

Step 5: p _{L obtained} by normalizing this point P _L by the following equation is the left camera image position.

p _L = P _L / | P _L | (Appendix -27)
Step 6: can determine a binocular parallax σ by the following equation from the position p _L and the current time position p _R. Even changing the left and right inter-camera distance [Delta] x _LR set arbitrarily in step 2, the binocular disparity is not affected.

σ = cos ⁻¹ (p _R p _L ) (Appendix -28)
(B-2) Normalized shortest distance _n d _s Normalized shortest distance _n d _s , plane orientation n _s , right camera image position p _R , position p _axis on the visual axis connecting the left and right cameras (visual axis direction a Two methods for calculating the binocular parallax σ from _xis ) will be described below.

(B-2-1) Method 1: Method by small circle transformation (FIG. 170)
Step 1: A normalized shortest distance _n d _s , a plane orientation n _s , a right camera image position p _R , and a position p _axis on the visual axis connecting the left and right cameras are set.

  Step 2: The radius R of the small circle transformation is determined by the following equation (4.3 described above, FIG. 170 (A)).

^{_{R = cos -1 (p R n}} s) ( Appendix -31)
Step 3: Determination of σ, that is, cos ⁻¹ (p _R p _L ) As described in 4.3, there is a relationship of the following equation (FIG. 170 (B)).
_{cosR = n d s / (p} axis p R p L)
_{= N d s sinσ / sin (} c + σ) ( Appendix -32)
Solving this equation for ^{σ σ = tan -1 (sinc cosR} / (n d s -cosc cosR))
(Appendix -33)
Thus, the binocular parallax σ was determined.

Incidentally, if you want to find the next time position p _L can be calculated by the following equation as for formula (Appendix -25).

p _L = cosσξ + sinση (Appendix -34)
(B-2-2) Method 2: Alternative Method It can also be determined by the following method.

Step 1: A normalized shortest distance _n d _s , a plane orientation n _s , a right camera image position p _R , and a position p _axis on the visual axis connecting the left and right cameras are set.

Step 2: Set any lateral camera distance [Delta] x _LR, to calculate the shortest distance d _s by the following equation from the center of the camera to the plane.

_{d s = n d s Δx LR} ( Appendix -35)
Step 3: orientation shortest distance _{n s} is determined that the plane of the _{d s.} Extending from the camera center in the direction of the p _R, to calculate the P _R point of intersection with the plane.

Step 4: The point _{P R} to calculate the direction [Delta] x _LR only in moved _{P L} of _{p axis.}

Step 5: p _{L obtained} by normalizing this point P _L by the following equation is the left camera image position.

p _L = P _L / | P _L | (Appendix -36)
Step 6: The binocular parallax σ can be determined from the position p _L and the right camera image position p _R by the following equation. Even if changing the left and right camera distance [Delta] x _LR set arbitrarily in step 2, the binocular disparity is not affected.

σ = cos ⁻¹ (p _R p _L ) (Appendix -37)

It is explanatory drawing of an optical flow pattern. It is a figure which shows the general view of the computer system employ | adopted as one Embodiment of the image measuring device of this invention. It is a hardware block diagram of the computer system shown in FIG. It is a figure which shows a mode that a plane moves. It is explanatory drawing of the principle which measures the three-dimensional azimuth | direction of a plane. It is a figure which shows a mode that a spherical surface is moved. It is explanatory drawing of the principle which measures the distance to a point. It is explanatory drawing of the principle which measures the time until it crosses a plane. It is a figure which shows the definition of a center angle. It is explanatory drawing of the principle which measures normalization time by a cylinder arrangement | sequence. It is a figure which shows the result of having performed computer simulation of normalization time measurement. It is a figure which shows the relationship between time to cross a plane, and the shortest distance. It is explanatory drawing of the principle which measures the shortest distance to a plane. It is a schematic diagram for small circle conversion. It is a figure which shows the relationship between the distance to a point, and the shortest distance to a plane. It is a figure which shows the geometrical meaning of small circle transformation. It is explanatory drawing of the principle which measures the normalization shortest distance by a cylinder arrangement | sequence. It is a table top view of the standardized shortest distance. It is a figure which shows the equivalence of a camera movement and a plane movement. It is a figure which shows the relationship between the polar transformation on a spherical surface, and the polar transformation on a plane. It is a figure which shows the relationship between a spherical camera image and a plane camera image. It is explanatory drawing of the measurement principle of the distance until it crosses a plane in a visual axis direction. It is a figure which shows the definition of a center angle. It is explanatory drawing of the principle which measures "the normalization distance until it crosses a plane in a visual axis direction" by cylinder arrangement | sequence. It is a figure which shows the relationship between "the normalization distance until crossing a plane in a visual axis direction" and "the shortest distance to a plane". It is principle explanatory drawing which measures the shortest distance to a plane. It is explanatory drawing of the principle which measures the normalization shortest distance by a cylinder arrangement | sequence. It is a block diagram of embodiment A-1. It is a flowchart of Embodiment A-1 of Claim 4. It is explanatory drawing regarding embodiment A-1. It is a block diagram of embodiment A-2. It is a flowchart of embodiment A-2. It is a block diagram of embodiment A-3. It is a flowchart of embodiment A-3. It is explanatory drawing regarding embodiment A-3. It is a block diagram of embodiment A-4. It is a flowchart of embodiment A-4. It is a block diagram of embodiment A-5. It is a flowchart of embodiment A-5. It is a block diagram of embodiment A-6. It is a flowchart of embodiment A-6. It is a block diagram of Embodiment A-7. It is a flowchart of embodiment A-7. It is a block diagram of embodiment A-8. 10 is a flowchart of Embodiment A-8. It is a block diagram of Embodiment A-9. 10 is a flowchart of Embodiment A-9. FIG. 11 is a block diagram of Embodiment A-10. 11 is a flowchart of Embodiment A-10. It is a block diagram of embodiment B-1. It is a flowchart of Embodiment B-1. It is a block diagram of embodiment B-2. It is a flowchart of embodiment B-2. It is a block diagram of embodiment B-3. It is a flowchart of embodiment B-3. It is explanatory drawing regarding Embodiment B-3. It is a block diagram of embodiment B-4. It is a flowchart of embodiment B-4. It is a block diagram of Embodiment B-5. It is a flowchart of embodiment B-5. It is a block diagram of Embodiment B-6. It is a flowchart of embodiment B-6. It is a block diagram of Embodiment B-7. It is a flowchart of embodiment B-7. It is a block diagram of Embodiment B-8. 10 is a flowchart of Embodiment B-8. FIG. 10 is a block diagram of Embodiment B-9. 10 is a flowchart of Embodiment B-9. FIG. 11 is a block diagram of Embodiment B-10. 11 is a flowchart of Embodiment B-10. It is a block diagram of Embodiment C-1. It is a flowchart of Embodiment C-1. It is a block diagram of embodiment C-2. It is a flowchart of Embodiment C-2. It is a block diagram of embodiment C-3. It is a flowchart of Embodiment C-3. It is a block diagram of embodiment C-4. It is a flowchart of Embodiment C-4. It is a block diagram of Embodiment C-5. It is a flowchart of Embodiment C-5. It is a block diagram of Embodiment C-6. It is a flowchart of Embodiment C-6. It is a block diagram of Embodiment C-7. It is a flowchart of Embodiment C-7. It is a block diagram of Embodiment C-8. 10 is a flowchart of Embodiment C-8. It is a block diagram of Embodiment D-1. It is a flowchart of Embodiment D-1. It is a block diagram of embodiment D-2. It is a flowchart of Embodiment D-2. It is a block diagram of Embodiment D-3. It is a flowchart of Embodiment D-3. It is a block diagram of Embodiment D-4. It is a flowchart of Embodiment D-4. It is a block diagram of Embodiment D-5. It is a flowchart of embodiment D-5. It is a block diagram of Embodiment D-6. It is a flowchart of Embodiment D-6. It is a block diagram of Embodiment D-7. It is a flowchart of Embodiment D-7. It is a block diagram of Embodiment D-8. 10 is a flowchart of Embodiment D-8. It is a block diagram of Embodiment D-9. 10 is a flowchart of Embodiment D-9. It is a block diagram of Embodiment D-10. It is a flowchart of Embodiment D-10. It is a block diagram of Embodiment D-11. It is a flowchart of Embodiment D-11. It is a block diagram of Embodiment D-12. 14 is a flowchart of Embodiment D-12. It is Embodiment E-1 block diagram. It is a flowchart of Embodiment E-1. It is a flowchart of Embodiment E-1. It is a block diagram of embodiment E-2. It is a flowchart of embodiment E-2. It is a flowchart of embodiment E-2. It is a block diagram of embodiment E-3. It is a flowchart of embodiment E-3. It is a flowchart of embodiment E-3. It is a block diagram of Embodiment E-4. It is a flowchart of embodiment E-4. It is a flowchart of embodiment E-4. It is a block diagram of Embodiment E-5. It is a flowchart of embodiment E-5. It is a flowchart of embodiment E-5. It is a block diagram of Embodiment E-6. It is a flowchart of embodiment E-6. It is a flowchart of embodiment E-6. It is a block diagram of Embodiment E-7. It is a flowchart of embodiment E-7. It is a flowchart of embodiment E-7. It is a block diagram of Embodiment E-8. 10 is a flowchart of Embodiment E-8. 10 is a flowchart of Embodiment E-8. It is a block diagram of embodiment F-1. It is a flowchart of Embodiment F-1. It is a flowchart of Embodiment F-1. It is a block diagram of embodiment F-2. It is a flowchart of embodiment F-2. It is a flowchart of embodiment F-2. It is a block diagram of embodiment F-3. It is a flowchart of Embodiment F-3. It is a flowchart of Embodiment F-3. It is a block diagram of embodiment F-4. It is a flowchart of Embodiment F-4. It is a flowchart of Embodiment F-4. It is a block diagram of Embodiment F-5. It is a flowchart of Embodiment F-5. It is a flowchart of Embodiment F-5. It is a block diagram of Embodiment F-5. It is a flowchart of Embodiment F-6. It is a flowchart of Embodiment F-6. It is a block diagram of Embodiment F-7. It is a flowchart of Embodiment F-7. It is a flowchart of Embodiment F-7. It is a block diagram of Embodiment F-8. 10 is a flowchart of Embodiment F-8. 10 is a flowchart of Embodiment F-8. It is a block diagram of a motion parallax detection unit. It is the figure which showed the response | compatibility with motion parallax _k (tau) and a parallax vector ( _k (tau) _x , _k (tau) _y ). It is explanatory drawing of the calculation method of motion parallax (tau) by a cross ratio and polar transformation. It is explanatory drawing of the calculation method of motion parallax (tau) by small circle transformation. It is a schematic diagram of an _ij τ table. It is a schematic diagram of an _ij τ table. It is a schematic diagram of a { _ik j} table. It is a schematic diagram of a { _ik j} table. It is a block diagram of a binocular parallax detection unit. It is the figure which showed the response | compatibility with binocular parallax _k (sigma) and a parallax vector ( _k (sigma) _x , _k (sigma) _y ). It is explanatory drawing of the calculation method of binocular parallax (tau) by a cross ratio and polar transformation. It is explanatory drawing of the calculation method of binocular parallax (tau) by small circle transformation. It is a schematic diagram of an _ij σ table. It is a schematic diagram of an _ij σ table. It is a schematic diagram of a { _ik j} table. It is a schematic diagram of a { _ik j} table.

Explanation of symbols

11 Camera 12 Image register at current time t ₀ 13 Image register at next time t ₁ 14 Extraction unit in moving direction v 15 p _inf setting unit 16 _n t _c parameter scanning unit 17 multi-ratio conversion unit 18 polar conversion unit 19 cylindrical array Voting unit 20 Peak extraction unit 21 Scan unit for v parameter 22 Scan unit for _n d _s parameter 23 Calculation unit for radius R 24 Small circle conversion unit 25 Cylindrical array voting unit 26 Peak extraction unit 27 Point distance calculation unit 28 τ determination unit 112 right camera image register 113 left camera image register 115 p _axis setting unit 116 _n d _c parameters of the scanning unit 117 cross-ratio conversion unit 118 pole conversion unit 119 cylindrical array voting unit 120 peak Unloading unit 121 a _xis parameters of the scanning unit 122 _n d _s parameters of the scanning unit 123 radius R of the computation unit 124 a small circle conversion unit 125 cylindrical array voting unit 126 peak extraction unit 127-point distance calculation unit 128 sigma determining unit 221 _n d _s the scanning unit 223 small circle component parameters of the scanning unit 222 radius ^{_{r i r n s +, i}} r n s- calculation unit 224 cylindrical array voting unit 225 peak extraction unit 300 computer system 301 body 302 CRT display 303 keyboard 304 mouse 310 Bus 311 Central processing unit (CPU)
312 RAM
313 Magnetic disk controller 314 Floppy disk driver 315 MO driver 316 Mouse controller 317 Keyboard controller 318 Display controller 319 Communication board 320 Image input board 401 Scan unit 402 of pixel number i Local area image extraction unit centered on 402 _i p _o 403 scanning unit with motion parallax number k 404 motion parallax _k τ conversion unit 405 motion parallax detection unit 406 _i p _o conversion unit 410 communication line 421 scanning unit 422 _i p _R with local area image centered on pixel number i 423 Binocular parallax _k sigma conversion unit 424 Binocular parallax _k sigma conversion unit 425 Binocular disparity detection unit 501 p _inf conversion unit 502,552 pixel number i Scanning unit査units 503,553 element number j 504 _i p _o output unit 505 n _sj output unit 506 _n t _cj output unit 507,510 _ij τ table creation unit 508,559 v parameter of the scanning unit 509 _n d _sj output unit 511 p _axis setting unit 512, 572 Scan unit with pixel number i 513, 573 Scan unit with element number j 514 _i p _R output unit 515 n _sj output unit 516 _n d _cj output unit 517, 520 _ij σ table creation unit 518, 579 a _xis parameter scanning unit 519 _n d _sj output unit 554 _i p _o local area image extraction unit 555, 565 _ij τ table 556 motion parallax detection unit 55 7,567 cylindrical array tomographic units 558,568 peak detection unit 561 right camera 562 left camera 574 _i p of the local region image centered on the _R cut unit 575,585 _ij σ table 576,586 Binocular disparity detection unit 577,587 Cylindrical array voting unit 578, 588 Peak extraction unit 601, 611 { _ik j} Table conversion unit 602, 612 { _ik j} table 651 Camera 652 Scan unit with pixel number i 653 Scan unit with motion parallax number k 654 _i p ₀ running the center to cut unit 655,665 motion parallax _k tau conversion unit 656,666 motion parallax detection unit 657,667 columnar array voting unit 658 and 668 peak extraction unit 659 v parameter of the local region image Units 621 and 631 _{ik j} topical around the scanning unit 674 _i p _R of the table conversion unit 622 and 632 _{ik j} table 661 right camera 662 scan unit 673 Binocular parallax number k of the left camera 672 pixel number i Region image segmentation unit 675, 685 Binocular parallax _k σ conversion unit 676, 686 Binocular parallax detection unit 677, 687 Cylindrical array voting unit 678, 688 Peak extraction unit 679 a _xi parameter scanning unit

Claims (11)

Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the visual axis direction connecting the two observation points is v, and the position of the infinity point on the straight line including the measurement points and extending in the same direction as the two visual axes is p. _{When it is axis} ,
Using a single ratio determined by the three positions p _axis , p _R , and p _L (p _axis p _R p _L ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the measurement plane A method for measuring an image, comprising: obtaining a physical quantity indicating a shortest distance between one of the two observation points. For the single ratio (p _axis p _R p _L ) or an operation equivalent to the single ratio, instead of the two measurement positions p _R and p _L when the measurement point is observed from the two observation points, the measurement point, the measurement position p _R when observed from one of the observation point of said two observation point, of the measurement point, two measurement positions when observed from the two observation points p _R, The image measurement method according to claim 1, wherein a calculation using binocular parallax σ which is a difference in position between p _L is included. As each of the positions p _axis , p _R , and p _L , each position projected on a spherical surface is adopted, and as a physical quantity that indicates the shortest distance, one of the measurement plane and the two observation points is observed. When the shortest distance between the points is d _s and the distance between the two observation points is Δx _LR ,
_n d _s = d _s / Δx _LR
In adopting the normalized shortest distance _n d _s represented,
A first step of setting the normalized shortest distance _n d _s as a parameter;
Using the normalized shortest distance _n d _s set in the first step and a simple ratio (p _axis p _R p _L ) or an operation equivalent to the simple ratio,
R = cos ⁻¹ ( _n d _s / (p _axis p _R p _L ))
A second step for obtaining a radius R defined by the relational expression or a relational expression equivalent to the relational expression,
A third step of obtaining a small circle with a radius R centered on the measurement position of the measurement point when observed from one of the two observation points;
For a plurality of measurement points in the measurement space, the second step and the third step are repeated a plurality of times while changing the parameters in the first step, and then
A plurality of small circles that intersect at the intersections by obtaining intersections between the small circles when the plurality of small circles obtained while repeating the first to third steps a plurality of times are drawn in the small circle drawing space. The fourth step of obtaining the orientation n _{so of the} measurement plane including a plurality of measurement points corresponding to and / or the normalized shortest distance _n d _s0 for the measurement plane is performed. Image measurement method. Measurement points appearing on the image have intensity information,
In the third step, the small circle is obtained, and the intensity of the measurement point corresponding to the small circle at each point on the locus of the small circle when the obtained small circle is drawn in the small circle drawing space. Voting for the value corresponding to
Instead of obtaining the intersection in the fourth step, the voting of the maximum point is obtained by obtaining a maximum point at which the value by voting becomes a maximum value while repeating the first to third steps a plurality of times. orientation n _so the measurement plane including a plurality of measurement points corresponding to the plurality of small circles joined _to, and / or, characterized in that it is a step of obtaining a normalized shortest distance _n d _s0 for the measurement plane according Item 4. The image measurement method according to Item 3. Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. When the measurement position is p _R , p _L , the measurement point, and the position of the infinity point on the straight line extending in the same direction as the visual axis direction v connecting the two observation points is p _axis ,
The three positions _{p axis,} using p R, _{p L,} single-ratio determined by the _{(p axis p R p L)} or single-ratio and equivalent operation, and the measuring point, one of the two observation points An image measurement method characterized in that a physical quantity indicating a distance between one observation point is obtained. For the single ratio (p _axis p _R p _L ) or an operation equivalent to the single ratio, instead of the two measurement positions p _R and p _L when the measurement point is observed from each of the two observation points, the measurement point, the measurement position p _R when observed from one of the observation point of said two observation point, the measurement point, the two two measurement positions p _R when observed from the observation point 6. An image measurement method according to claim 5, wherein a calculation using a binocular parallax σ which is a difference in position between p _L and p _L is included. As a physical quantity which indicates the distance, the distance between one point of observation of the measurement point and the two observation points d _0, when the distance between the each other two observation point was set to [Delta] x _LR ,
_n d ₀ = d ₀ / Δx _LR
In the normalized distance _n d ₀ represented adopted, the normalized distance _n d _0,
n d ₀ = (p _axis p _R p _L )
The image measurement method according to claim 5, wherein the image measurement method is obtained using a relational expression of the above or a relational expression equivalent to the relational expression. Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the direction of the visual axis connecting the two observation points is v, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the two visual axes. When it is p _axis
Using a single ratio determined by the three positions p _axis , p _R , and p _L (p _axis p _R p _L ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the measurement plane An image measurement apparatus comprising: a calculation unit that obtains a physical quantity that indicates the shortest distance between one of the two observation points. Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. When the measurement position is p _R , p _L , the measurement point, and the position of the point at infinity on a straight line extending in the same direction as the visual axis direction v connecting the two observation points is p _axis ,
The three positions _{p axis,} using p R, _{p L,} single-ratio determined by the _{(p axis p R p L)} or single-ratio and equivalent operation, and the measuring point, one of the two observation points An image measuring apparatus comprising a calculation unit that obtains a physical quantity that indicates a distance from one observation point. Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from the two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. The measurement positions are p _R and p _L , the direction of the visual axis connecting the two observation points is v, and the position of an infinite point on a straight line that includes the measurement points and extends in the same direction as the two visual axes. When it is p _axis
Using a single ratio determined by the three positions p _axis , p _R , and p _L (p _axis p _R p _L ) or an operation equivalent to the single ratio, the orientation of the measurement plane and / or the measurement plane An image measurement program storage medium for storing an image measurement program for obtaining a physical quantity indicating the shortest distance between one of the two observation points. Each of the arbitrary measurement points in the measurement space, as seen from an image when the measurement space is viewed from two predetermined observation points in the predetermined measurement space, is observed from each of the two observation points. When the measurement position is p _R , p _L , the measurement point, and the position of the infinity point on the straight line extending in the same direction as the visual axis direction v connecting the two observation points is p _axis ,
The three positions _{p axis,} using p R, _{p L,} single-ratio determined by the _{(p axis p R p L)} or single-ratio and equivalent operation, and the measuring point, one of the two observation points An image measurement program storage medium characterized by storing an image measurement program for obtaining a physical quantity that indicates a distance between one observation point.

Images