JP3343204B2 - Image origin measurement method and its recording medium - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for measuring an image origin of a visual sensor, which is important in performing three-dimensional measurement, from information obtained using a visual sensor as input means.

[0002]

2. Description of the Related Art In general, a center-projection type visual sensor such as a television camera or the like, as shown in FIG.
With a projection center O _P of the point. Here, the principal point of the foot O _S of perpendicular dropped from the projection center O _P to the image projection surface (Princip
al point), which is the “image origin” in the image coordinate system (U, V).

In general, when performing three-dimensional measurement using a visual sensor, sensor calibration is required.
At this time, the parameters that need to be obtained are the focal length of the visual sensor, distortion, the center of the image (image origin), and the position and orientation of the sensor itself in the three-dimensional space (X, Y, Z). Among them, the focal length, distortion, and image center are values unique to the sensor and are called internal parameters. The position and orientation of the sensor itself in the three-dimensional space are values relating to the installation of the sensor and are called external parameters. The camera calibration methods proposed so far generally capture a known test object arranged in a space and indirectly estimate camera parameters according to a preset camera model. But in that case, how to give the initial value,
The setting of the feature points affects the solution, causing a problem that affects the reliability. Further, the behavior of the error between the parameters is not clear, and care must be taken in setting the evaluation function. Therefore, it is desirable to obtain various parameters independently.

As a method of independently obtaining the image origin, a method of measuring the vanishing point direction from a set of parallel lines existing on a plurality of planes not directly facing the camera and obtaining the image origin is generally used.

[0005] Separately, a laser beam is emitted toward the camera, and its reflection is observed to obtain an image origin "au".
There is also a method called “tocollimated laser”.

[0006]

However, in the case of the above-mentioned conventional general camera calibration method, the method of giving initial values and the setting of feature points affect the solution, and there is a problem that reliability is affected. I do. Further, the behavior of the error between the parameters is not clear, and care must be taken in setting the evaluation function. Therefore, it is desirable to obtain various parameters independently, but it is necessary to prepare a special jig for the measuring device.

On the other hand, in the case of the above-described conventional method of independently measuring the vanishing point direction from a set of parallel lines existing on a plurality of planes that do not directly face the camera and obtaining the image origin, the entire image is obtained. The condition that the distortion can be ignored or sufficiently corrected is required. Further, it is necessary to measure the inclination of the set plane in advance with high accuracy, and the measurement requires some experience.

[0008] In addition, a laser beam is emitted toward a camera, and its reflection is observed to obtain an image origin.
In the "limited laser" method, there is a risk of damaging the sensor device itself.

An object of the present invention is to provide a simple and highly accurate calibration method with respect to an image origin which is important when performing three-dimensional measurement using an image.

[0010]

The present invention solves the above-mentioned problems by the following means (1) to ( 5 ).

[0011]

( 1 ) Test that can be moved to an arbitrary position in space
Highly accurate measurement of the test object and the position of the test object in the space
Set up a possible jig and take a picture with the visual sensor to be measured.
Shadow, from the observation result of the test object by the visual sensor.
A first step of moving the test object position in the space
And observation information of the visual sensor and positional information of the test object.
Second stage for estimating and calculating the image origin of the visual sensor from the information
Floor and the first stage and the second stage one or more times
Times, and the values obtained by the above calculation are statistically processed.
A third step of determining the image origin;
Method, wherein in the first step, an end point of any two line segments crossing one verification pixel on the imaging surface of the visual sensor is set as an observation point, and the test object is moved to a point corresponding to the observation point. The second step is to calculate the projection center of the visual sensor from the observation information on the point corresponding to the observation point by the visual sensor and the position information of the point corresponding to the observation point. A second procedure, wherein in the third step, the first procedure and the second procedure are performed once by changing a combination of the two line segments for all verification pixels on the imaging surface of the visual sensor. Or a third procedure to be repeated a plurality of times, a fourth procedure for calculating a standard deviation distribution of the projection center calculated in the third procedure, and a coordinate of a verification pixel having the minimum standard deviation as an image origin. And a fifth step of outputting. Image origin measurement method that.

( 2 ) In the first procedure, a point corresponding to the observation point is displayed as a test object using display means capable of blinking arbitrary coordinates on a plane, and in the third procedure, the observation point is displayed. An image origin measurement method characterized by changing a combination of two line segments by moving a display of a point corresponding to a predetermined pitch at a predetermined pitch.

( 3 ) In the first procedure, a point corresponding to the observation point is displayed as a test object by using display means capable of blinking arbitrary coordinates on a plane. In the third procedure, the display of the display is displayed. An image origin measurement method characterized by changing a combination of two line segments by changing an inclination angle.

( 4 ) In the third step, an image origin measuring method is characterized in that a combination of two line segments for all verification pixels is automatically changed.

( 5 ) The above (1), (2), (3),
(4 ) A recording medium characterized by recording a program for causing a computer to execute the procedure in any one of the image origin measurement methods according to (4 ) on a computer-readable medium.

According to the present invention, a test object which can be moved to an arbitrary position or display means which can blink arbitrary coordinates on a plane is provided, and the position of the test object or the plane is determined based on the position coordinates of the test object and the observation information of the visual sensor. By updating the blinking coordinates above and measuring and verifying the image origin at all coordinates on the image plane, it is possible to measure and verify the image origin at any position on the image plane, making it easy to measure the image origin with high accuracy To be able to In addition, the position of the test object can be automatically updated, so that a user who has no prior knowledge can easily perform the calibration of the image origin.

[0018]

Embodiments of the present invention will be described below in detail with reference to the drawings.

An embodiment of the present invention will be described by taking calibration of a CCD camera as an example. In the present embodiment, a liquid crystal display is used as a test object, and “Japanese Patent Application No. 6-60658” by the present applicant is used as a calculation method. The calculation method in this embodiment is as follows.

FIG. 1 is a diagram for explaining the projection of an elliptical cone in this embodiment. First, in a space, a cone having the projection center as the vertex and the optical axis as the central axis is considered. In FIG.
π _S is an imaging surface, and π _L is a display unit of the display unit. [pi _S and [pi _L becomes a cut surface of a cone whose apex the projection center O _P, the trajectory is a circle and an ellipse. In the case of an ideal pinhole model camera, the cones on the imaging surface side and the display unit side have the same apex angle, but they do not always agree because they are affected by distortion. Therefore, in the present invention, lens aberration and aspect difference, which are main factors of distortion, are taken into consideration. The lens aberration is proportional to the image origin O _S in the radial direction, and is a difference in the apex angle of the cone. The effect of the aspect difference appears as expansion and contraction of π _{L in} the U-axis direction. Therefore,
The cone on the imaging surface side is transformed into an elliptical cone on the display unit side. Considering the above characteristics, only the fact that “the central axis of the cone is the optical axis” is the geometrically invariant feature. This is a conic section to be cut by any plane containing the optical axis, Hikarihe means that the bisector of the corner whose vertices O _P.

Here, it is assumed that the image origin is determined by some method. The end points of the line segment having the image origin as the bisecting point on the imaging surface are obtained, and the corresponding points on the display unit are set to P _L1 and P _L2 . O _P , P _L1 , P
FIG. 2 shows the triangle connecting _L2 . Where π
_L is a display unit surface, [pi _P is any plane containing the optical axis, the O _L is the intersection of the optical axis and the display unit side. From the geometrical characteristics of the optical axis becomes the angle bisectors of O _P. O _L , P _L1 , and P _L2 are parameters that can be measured, and the position of O _P and the slope of π _P are unknown.

In the triangle (ABC) shown in FIG. 3, considering the property of the bisector of the corner of the vertex C, the intersection of the bisector and the side AB is defined as O, and the relationship AC: BC = AO: BO is obtained. order but satisfied, C is _{(x a x b / (x} a + x b),
0) as the center and radius _{r = x a x b / (} x a + x b) of the Ap
ollonius circle. When this is applied to three-dimensional space, O _P position as shown in FIG. 4 (C position) P
_It can be constrained on a sphere centered on the extension of _L1 (A position) and P _L2 (B position). This sphere will be referred to as an Apollonius sphere.

[0023] The Apollonius spheres is determined each independently When finding the secondary line segment that passes through O _L, can two spheres g _1, g ₂ is set as shown in FIG. Therefore, the existence range of O _P becomes the g _1, g ₂ of the line of intersection C _P. The presence area of the optical axis is also the NoTadashi plane to [pi _L include C _P can be limited to [pi _P.

The centers of g ₁ and g ₂ are _defined as O _g1 (x ₁ , y ₁ ) and O _g2
(X _2, y ₂₎ to the C _P of the center _{O C (x 3, y 3} ) is as follows.

X ₃ = (y ₂ −y ₁ ) (x ₁ y ₂ −x ₂ y ₁ ) / {(x ₁ −x ₂ ) ² + (y ₁ −y ₂ ) ² } (1) y ₃ = (X ₁ −x ₂ ) (x ₁ y ₂ −x ₂ y ₁ ) / {(x ₁ −x ₂ ) ² + (y ₁ −y ₂ ) ² } (2) FIG. 6 shows an Apollonius sphere. two line segment l _1, l ₂ of the crossing angle theta _L, the π angle apparent on _S theta _S
This shows the relationship.

Here, let us consider a case where the apparent angle θ _S is 90 ° in order to simplify the calculation. Let the intersection of π _P and π _{L be} l _P, and the angle between l ₁ and l _P is θ ₁ , and the angle between l ₂ and l _P is θ
Assuming that ₂ , tan θ ₁ = (x ₁ y ₃ −x ₃ y ₁ ) / (x ₁ x ₃ + y ₁ y ₃ ) (3) tan θ ₂ = (x ₂ y ₃ −x ₃ y ₂ ) / (x _{2 x 3 + y 2 y 3} ) a (4). The angle θ _P formed by l _P and the optical axis l _O on π _P is uniquely determined as θ _P = sin ^-1 (√ (tan θ ₁ tan θ ₂ )) (5). Thus the position of the O _P at ^{_{x P = 2x 3 cos 2 θ}} P (6) y P = 2y 3 cos 2 θ P (7) z P = 2cosθ P sinθ P √ (x 3 2 + y 3 2) (8) , Given coordinates (FIG. 7). The above calculation method is verified for all coordinates (pixel positions) on the image according to the flowchart shown in FIG.

Here, the procedure of the processing of this embodiment will be described with reference to the flowchart of FIG. The processing in this embodiment is performed according to the procedures (1) to (10) in the figure.

(1) All pixels on the liquid crystal display are subjected to processing such as blinking and the observation area in the camera field of view is determined, and then the starting coordinate point for obtaining the standard deviation distribution of (8) is determined.

(2) Starting coordinate point on image plane (verification pixel)
Is set as an observation point, and the corresponding point is drawn on the liquid crystal display. As a drawing method, when drawing a corresponding point of an observation point, when an arbitrary area of the liquid crystal display is blinked, if the change is observed at the observation point, the area is set as a candidate area, and then a candidate area is set. Is divided into two, the candidate regions are similarly limited, and this process is repeated a plurality of times to determine corresponding points.

(3) Position information of the corresponding point of the two lines,
The projection center is calculated by the above calculation method using the observation information of two lines by the camera.

(4) The combination of two line segments with the verification pixel as a bisecting point is changed, and drawing is performed in the same manner as described above. here,
Any two line segments crossing the verification pixel are used, and the observation point is changed at regular intervals around the verification pixel.

(5) Steps (3) and (4) are repeated until all combinations are completed.

(6) After completion of all combinations, the verification pixel (image origin candidate) is moved.

(7) Steps (3) to (6) are repeated until scanning of all verification pixels is completed. Note that it is also possible to scan all the verification pixels for one two-segment and repeat this for all two-segment combinations.

(8) The standard deviation distribution of the results obtained in the processes (3) to (6) is calculated, and the change in the camera projection center position for each verification pixel is described as the standard deviation.

(9) Search for and output a coordinate position at which the standard deviation distribution has the minimum value.

(10) Since the definition of the image origin is the coordinates of the foot when a perpendicular is dropped from the camera projection center to the imaging surface,
The coordinate position having the minimum standard deviation is output as the image origin.

When the above-described method is applied in a state where there is an error in the setting of the image origin, the Apolloniu which does not satisfy the condition for bisecting the apex angle by the axis and is set from an arbitrary two line segments.
The line of intersection of the s-sphere means not stable. Therefore, for each point in the entire image area, Ap
ollonius sphere. Apol led at this time
FIG. 9 is a graph showing the stability of the Lonius ball intersection. From this result, it can be seen that the variance σ of the projection center position is the smallest at the image origin, and increases as the distance from the image origin increases. Therefore, an arbitrary starting point is determined, and a variance value in the vicinity of the starting point 8 is searched, and the search pixel is moved in a direction in which the variance value becomes smallest. At the moved point, the surrounding area is similarly searched, and the same scanning is repeated. Then, when the search pixel does not move even when the surrounding area is searched, the end condition is set. In this case, a point where the variance is the smallest is found, and the image origin of the camera system is uniquely determined. Therefore, by searching for a point at which the variance value of the projection center position is minimized, the image origin of the camera system can also be searched.

The method of this embodiment is based on the premise that the liquid crystal display is not parallel to the camera image plane. The simulation results for the accuracy of estimating the tilt of the liquid crystal display (test target) with respect to the camera and the projection center position indicate
It has been found that the test object does not affect the accuracy if the inclination is 20 ° to 70 ° with respect to the camera. Also, as a result of examining the relationship between the magnitude of the vertex angle of the projection cone and the estimation accuracy of the projection center position, it was found that sufficient accuracy was obtained when the vertex angle of the cone was 60 ° or less.

It should be noted that the effects of the present invention can be obtained by using other calculation methods such as an algorithm for calculating a generally used perspective projection transformation matrix instead of the calculation method shown in the above embodiment. Can be

In the above embodiment, an example is shown in which a test object and a display unit whose pixel position is specified with high precision are used. However, the present invention is not limited to this, and in general, any position in space is used. And a jig capable of measuring the position of the test object in the space with high accuracy. In addition, the example in which the observation point is moved at a predetermined pitch to obtain the standard deviation distribution of the projection center position has been described. However, by changing the inclination of the display surface of the display unit in the range of 20 ° to 70 °, a plurality of times can be obtained. The standard deviation distribution of the projection center position may be obtained by performing observation.

The processing procedure in the method described in the above embodiment, for example, the processing procedure shown in the flowchart of FIG. 8 can be appropriately executed by a computer, and the computer is caused to execute the procedure. Can be recorded on a computer-readable medium, such as a floppy disk or a CD-ROM, and distributed.

[0043]

As described above, according to the present invention,
A test object movable to an arbitrary position or display means capable of blinking arbitrary coordinates on a plane is provided, and the position of the test object or blinking coordinates on the plane is updated based on the position coordinates of the test object and observation information of the visual sensor. Since the image origin is measured and verified at every coordinate on the image plane, the image origin can be measured and verified at an arbitrary position on the image plane, so that the image origin can be measured with high accuracy. In addition, if the position of the test object is automatically updated, it is possible for a user who has no prior knowledge to easily perform the calibration of the image origin.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating elliptical cone projection according to an embodiment of the present invention.

FIG. 2 is a view showing a conical cross section in FIG. 1;

FIG. 3 is a diagram for explaining a bisecting condition of a corner in the embodiment.

FIG. 4 is a diagram for explaining an Apollonius sphere in the embodiment.

FIG. 5 is a diagram for explaining an intersection of Apollonius spheres in the embodiment.

FIG. 6 is a diagram showing a relationship between an intersection angle θ _{L of} two line segments l ₁ and l ₂ in which an Apollonius sphere is set and an apparent angle θ _S in the embodiment.

FIG. 7 is a diagram for explaining a result of deriving a projection center position in the embodiment.

FIG. 8 is a flowchart showing a processing procedure in the embodiment.

FIG. 9 is a graph showing the stability of the Apollonius ball intersection in this embodiment.

FIG. 10 is a diagram for explaining an image origin.

[Explanation of symbols]

O _P ... projection center O _S ... image origin [pi _S ... imaging plane [pi _L ... display unit of the display unit (display unit surface) [pi _P ... arbitrary plane O _L ... intersection of the optical axis and the display unit plane including the optical axis

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-6-102013 (JP, A) JP-A-61-277012 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01B 11/00-11/30

Claims (5)

(57) [Claims] A test pair movable to an arbitrary position in space.
The position of the elephant and the test object in the space can be measured with high accuracy
Set up a jig and shoot with the visual sensor to be measured
Before the observation result of the test object by the visual sensor.
A first step of moving the position of the test object in the space
From the observation information of the visual sensor and the position information of the test object.
Second step of estimating and calculating the image origin of the visual sensor
And repeating the first step and the second step one or more times.
And calculate the image origin by statistical processing of the values obtained in the above calculation.
Determining the image origin.
Therefore, in the first step, a test object is moved to a point corresponding to the observation point, with an end point of any two line segments crossing one verification pixel on the imaging surface of the visual sensor as an observation point. A step of calculating a projection center of the visual sensor from observation information on the point corresponding to the observation point by the visual sensor and position information of the point corresponding to the observation point in the second step.
In the third step, the combination of the two segments is changed for all the verification pixels on the imaging surface of the visual sensor, and the first procedure and the second procedure are performed one or more times. A third procedure to be repeated twice, a fourth procedure for calculating the standard deviation distribution of the projection center calculated in the third procedure, and outputting, as the image origin, the coordinates of the verification pixel having the minimum standard deviation. It has a fifth procedure, and images origin measurement method you wherein a. 2. In the first procedure, a point corresponding to the observation point is displayed using a display means capable of blinking arbitrary coordinates on a plane as a test object, and in the third procedure, the observation point is assigned to the observation point. The image origin measuring method according to claim 1 , wherein a combination of two line segments is changed by moving a display of a corresponding point at a predetermined pitch. 3. In the first step, a point corresponding to the observation point is displayed using a display means capable of blinking arbitrary coordinates on a plane as a test object. In the third step, the display tilts. The image origin measuring method according to claim 1 , wherein a combination of two line segments is changed by changing an angle. The method according to claim 4, wherein the third step, the image origin measurement according to all validation automatically changing two segment combination of the pixel, claim 1, 2, 3, characterized in that Law. 5. The method of claim 1, 2, 3, a program for executing the procedure in the image origin measurement method according to the computer in any one of 4, the computer is recorded on a medium readable, characterized in that recoding media.