JPH1196375A - Method and device for measuring time series image moving place and recording medium recorded with time series image moving place measuring program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate a finer speed component about each pixel or each graphic element in an area that is an object in a time series image and to measure a moving place of the time series image. SOLUTION: A time series image is inputted and an object area and a movement locus that is included there are extracted from it (steps 401 and 402). The object area is undergone three-dimensional Hough transform, three- dimensional parameter space that is acquired as a result is projected to two-dimensional space, voting is performed about the space to detect a crossing line of a tangential plane of the movement locus, the peak of a distribution of voting values in parameter space that is acquired as a result, and a speed component is calculated from it (steps 402 to 406). About each pixel in the object area, the degree of contribution to the voting value of the peak that is detected in the step 406 is calculated, the speed component that is calculated in the step 406 is allocated to each pixel according to the size, the speed components of each pixel and each frame are collected to configure a speed place, and it is outputted (steps 407 to 409).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for recognizing a motion of an object in a time-series image obtained by a video camera, a weather radar device, remote sensing, or the like.

[0002]

2. Description of the Related Art In monitoring of people and traffic, controlling manufacturing processes, and elucidating and predicting natural phenomena such as weather, etc., it is desired to improve the efficiency and efficiency using recognition processing of time-series images. I have.

Conventionally, there is a method using Hough transform (also called voting) as a method for measuring the movement of an object from a time-series image. As an example, when considering a spatio-temporal image in which time-series images are stacked in the time direction, a curved motion trajectory drawn by an edge or contour of a target included in the image is detected,
There is a method of obtaining a target velocity component from directions of intersections formed by a plurality of different tangent planes tangent to the motion trajectory (Japanese Patent Application Nos. 9-3116, 9-8563, and 9-1).
No. 14577). This method converts a spatiotemporal image into a parameter space that expresses the velocity component (direction and magnitude of velocity) of the target object by Hough transform, detects the peak of the distribution in the parameter space, and detects the target from the coordinate values. Find the velocity component of the object. This technique has the advantage that the most dominant translation speed component in the target area is obtained robustly against noise and occlusion.

[0004]

However, in the prior art, since a relatively large area of the time-series image is required to obtain the local velocity in the time-series image, the motion field (optical flow, velocity There is a problem that the resolution is also spatially and temporally coarse.

An object of the present invention is to provide a time-series image for estimating a finer speed component and measuring a motion field of the time-series image for each pixel or each graphic element in a target area in the time-series image. It is an object of the present invention to provide a motion field measurement method and apparatus, and a recording medium on which a time-series image motion field measurement program is recorded.

[0006]

According to the time-series image motion field measurement method of the present invention, Hough transform or voting is performed on each pixel or graphic element included in a certain area in the time-series image, and as a result, The peak of the voting distribution is detected in the obtained parameter space (voting space), the degree of contribution of each pixel or graphic element to the peak value is calculated, and the velocity component of each pixel or graphic element is determined according to the magnitude. Thus, the motion field (optical flow, velocity field) of the time-series image is obtained.

Therefore, even when an image is deteriorated due to noise or the like, a dominant velocity component in the image can be detected.

The time-series image motion field measuring method according to the present invention, when the time-series image is viewed as a three-dimensional space (spatio-temporal) obtained by adding a time axis to an image plane, The motion trajectory is detected in the form of a curved surface whose edges and contours are drawn in time and space, and the tangent plane tangent to the motion trajectory is detected by three-dimensional Hough transform. Detected by conversion or voting, measure the velocity component of the object, detect the peak of the voting value representing the detected velocity component in the parameter space (voting space) obtained by Hough transformation or voting, and Then, the degree of contribution of each pixel constituting the motion trajectory in the spatiotemporal space is calculated, and the velocity component of each pixel or graphic element is determined based on the magnitude thereof, thereby obtaining the motion field of the time-series image.

Therefore, it is possible to obtain an accurate speed (magnitude, direction) of an object in an image.

The time-series image motion field measuring method according to the present invention further comprises the step of determining the degree of each pixel or graphic element included in the time-series image which contributes to the peak value in the parameter space by Hough transform or voting. In the parameter space where 3D Hough transform voting has been performed on the element,
The voting value of the cell in which the largest voting value is accumulated among the elements (cells) that can contribute to the detected velocity component is obtained as the degree of contribution of the pixel or the graphic element.

Further, the degree of contribution of each pixel or graphic element included in the time-series image to the peak value of the voting value in the parameter space used for detecting the velocity component is compared with a certain threshold value. By giving the velocity component obtained from the peak position of the voting value in the parameter space only to the pixels or graphic elements having the degree of contribution of each pixel, even in a situation where noise or the like is added, each pixel can be stably and accurately set. Alternatively, the velocity component of the graphic element is obtained, and a fine motion field of the time-series image can be obtained.

According to the time-series image motion field measuring apparatus of the present invention, when a time-series image is viewed as a three-dimensional space (time-space) obtained by adding a time axis to an image plane, an edge of an object included in the image and Means for detecting a curved motion trajectory whose contour is drawn in space-time, and detecting a tangent plane tangent to the motion trajectory 3
Dimensional Hough transform means, means for detecting the direction of an intersection line intersecting the different tangent planes by Hough transform or voting, and means for detecting the velocity detected in a parameter space (voting space) obtained by Hough transform or voting. Means for detecting the peak of the voting value indicating the direction, means for calculating the degree of contribution of each pixel constituting the motion trajectory in space and time to this peak, and the velocity component of each pixel or graphic element according to its magnitude Means for determining

In the parameter space in which three-dimensional Hough transform voting has been performed on each pixel or graphic element, the largest voting value among elements (cells) that can contribute to the detected velocity component is accumulated. Means for searching for cells and calculating the voting value as the degree to which each pixel or graphic element contributes to the peak value in the parameter space by Hough transform or voting.

Further, in the time-series motion field measuring apparatus according to the present invention, the degree of contribution of each pixel or graphic element included in the time-series image to the peak value of the voting value in the parameter space used for detecting the velocity component is determined. A means is provided for comparing the threshold values and providing a velocity component obtained from the peak position of the voting value in the parameter space only to pixels or graphic elements having a degree of contribution equal to or greater than the threshold value. Therefore, even in a situation where noise or the like is added, the speed component of each pixel or graphic element can be obtained stably and accurately, and a fine motion field of a time-series image can be obtained.

[0015]

Next, embodiments of the present invention will be described with reference to the drawings.

Referring to FIG. 1, a time-series image motion field measuring apparatus according to an embodiment of the present invention includes an input unit 100 and a processing unit 20.
0 and the output unit 300.

The processing unit 200 extracts a motion field measurement target area from the time-series image input by the input unit 100, and as a set of pixels in the spatio-temporal image memory 251 which is a three-dimensional array, A region extraction unit 201 that forms a motion trajectory drawn by an edge or a contour, a three-dimensional Hough transform unit 202 that detects a distribution of a tangent plane of the motion trajectory, and a distribution of a tangent plane obtained by the three-dimensional Hough transform are stored. A two-dimensional voting space memory 252, a projection unit 203 for projecting the distribution in the three-dimensional voting space memory 252 onto a two-dimensional plane, and a two-dimensional method for storing the resulting tangential plane when viewed in the normal direction A line parameter space memory 253, an intersection voting unit 204 for obtaining a distribution of the direction of the intersection by a combination of a plurality of tangent planes represented by the three-dimensional normal parameter space memory 253, The intersection line parameter space memory 254 for storing the intersection line distribution obtained as a result of the above, and the peak of the voting distribution in the intersection line parameter space memory 254 is detected, and the target velocity component in the target time-series image region is determined. The speed component detection unit 205 to be obtained and the contribution calculation unit 21 for calculating the degree of contribution to the peak value detected by the speed component detection unit 205 for each pixel in the spatiotemporal image memory 251
1, a contribution determining unit 212 that evaluates the magnitude of the contribution of each pixel, and determines whether to allocate the speed component determined by the speed component detection unit 205, and a speed component assigned to each pixel. It is composed of a speed field construction unit 213 that composes the entire target region and composes a speed field.

Next, the processing flow of the apparatus shown in FIG. 1 will be described with reference to the flowchart shown in FIG. A time-series image is input (step 401), a target area and a motion trajectory included therein are extracted from the time-series image (step 402), and a three-dimensional Hough transform is performed on the target area (step 40).
3), project the resulting three-dimensional parameter space into a two-dimensional space (step 404), and perform voting for detecting the intersection of the tangent plane of the motion trajectory in this space (step 405). The peak of the resulting distribution of voting values in the parameter space is detected, and the velocity component is calculated therefrom (step 406). Then, for each pixel in the target area, the voting value of the peak detected in step 406 is obtained. Is calculated (step 40).
7) The velocity component obtained in step 406 is assigned to each pixel according to the magnitude (step 408), and the velocity field is constructed by combining the velocity components of each pixel and each frame and output (step 409).

Next, an example of the operation of the processing section 200 will be specifically described.

In the target area extracting section 201, the input section 100
From the time-series image input as a target, a local region to be measured for the velocity field is extracted. The time-series image obtained in this way is stored as I (i, j, t). The target region extraction unit 201 further constructs a motion trajectory of the target edge or contour in the image in space and time as three-dimensional volume data.

As an example, calculating a difference between frames of a time-series image and constructing a motion trajectory as a spatio-temporal difference image D (x, y, t) using its positive value, negative value or absolute value. Can be. Then, the spatiotemporal difference image D (x,
(y, t) is stored in the spatiotemporal image memory 251. In the example using positive values,

[0022]

(Equation 1) , The spatiotemporal difference image D (x, y, t) can be calculated, and a columnar motion trajectory having an edge or contour in the image as a bottom curve can be extracted. The magnitude of the gray value of the spatiotemporal difference image D (x, y, t) is substantially proportional to the discontinuity of the spatial distribution of the brightness of edges and contours in the image and the amount of motion. It should be noted that a method of extracting a motion trajectory other than those described above can also be used.

Next, the three-dimensional Hough transform means 202 detects the distribution of a tangent plane which can be in contact with the motion trajectory in the spatiotemporal space by Hough transform or voting, and determines the frequency distribution thereof in a three-dimensional voting space which is a three-dimensional array. Obtained in memory 252.

As shown in FIG. 3, a point (x
_i , y _i , t _i ) is defined by three parameters (θ,
φ, ρ)

[0025]

(Equation 2) Can be expressed as Here, (θ, φ) represents the normal direction of the plane, and ρ represents the shortest distance from the origin to the plane. The space spanned by the three parameters representing the plane will be referred here the plane parameter space S _p. From equation (2), one point (x _i , y _i , z _i ) in the three-dimensional space corresponds to one curved surface in the plane parameter space as S ₁ in FIG.

[0026] In practice, the plane parameter space S _p, small gap (Δθ, Δφ, Δρ) is discretized, which is stored in the three-dimensional voting space memory 252 is a three-dimensional array. Here, the elements of the array are called cells.

Next, using voting, the distribution of the tangent plane of the motion trajectory in the target area represented as the spatiotemporal difference image D is obtained as the value of the cell in the parameter space. Specifically, space-time difference image D (i, j, t) for all the pixels in, computes a curved surface represented by the formula (2), the cells in the parameter space S _p where this curved surface passes The value is increased by the value of the pixel D (i, j, t) in the spatiotemporal difference image. This process is called voting. After voting for all pixels, the total value of cell voting in the plane parameter space S _p (θ, φ, ρ) is defined as the intensity of the tangent plane of the motion trajectory having the parameters (θ, φ, ρ). Therefore, if the distribution of the vote value in the parameter space S _p forms a peak, the coordinates of the peak, tangent plane of movement trajectories included in the time-space can be represented. Note that methods other than those described above can also be used.

The projection unit 203 includes a three-dimensional Hough transform unit 202
Search for the maximum value of the voting values accumulated in the cell in the ρ direction for each (θ, φ) in the plane parameter space S _p (θ, φ, σ) obtained in the three-dimensional voting space memory 252 by Is stored in the two-dimensional normal parameter space memory 253 which is a two-dimensional array. This (θ,
The space spanned by φ) is called a normal parameter space S _N (θ, φ).

[0029]

(Equation 3) This processing has a function of integrating the distribution of the tangent plane of the motion trajectory drawn in the space and time by the contour and the edge in the target area into the distribution when viewed in the normal direction invariant in time and position.

Next, the intersection line voting section 204
And a two-dimensional normal parameter space memory 253
The parameter space that represents the distribution of the direction of the intersecting line formed by a plurality of tangent planes that can be in contact with the motion trajectory is input with the distribution of the motion trajectory drawn in the space and time by the contours and edges in the target area stored in It converted to be called intersection line parameter space S _L).

Here, as shown in FIG. 4, the direction of the line of intersection is defined by an angle α and an xy plane (image plane) formed with the x axis when the line of intersection passing through the origin is projected onto the xy plane. Using the angle β,

[0032]

(Equation 4) Is expressed as However, 0 ≦ α <π / 2, 0 <β <π / 2
It is.

Now, two different points on the intersection line are represented by P₁ (X
₁ , Y₁ , T₁ ), P_Two (X_Two , Y _Two , T_Two ) And this
Two points P₁ , P_Two Equation (2) is simultaneously solved for
Then, by substituting equations (7) to (9), the normal parameter
Data space S_N And the intersection parameter space S_L The relationship

[0034]

(Equation 5) Can be obtained as follows.

The two tangent planes are represented as two points in the normal parameter space S _N , and when these points are converted into the intersection parameter space S _L , respectively, a curve represented by equation (10) is obtained. As a result, the direction of the line of intersection of the tangent plane is obtained.

Here, for all the elements (θ, φ) (referred to as cells) of S _{N in} the parameter space of the normal, the equation (1)
The cell in the intersection parameter space S _L through which the curve 0) passes
Vote on the value of _N (θ, φ). By executing such a Hough transform, the velocity component of the target, which may be included in the target area expressing the velocity component of the target object, is reflected on the voting distribution in the intersection line parameter space. Intersection line parameter space S _L thus obtained is stored in the line of intersection parameter space memory 254.

The velocity component detection unit 205 detects the peak of the distribution of the vote value stored in the line of intersection parameter space memory S _L, the coordinate value of the peak (α _P, β _P) than, in the target area Find the most dominant translation velocity component of the target object. The way of this movement is

[0038]

(Equation 6) And the magnitude V of the velocity is

[0039]

(Equation 7) Is obtained. Voting value S _L indicating the peak (α _P , β _P )
Can be said to be evidence information indicating the likelihood that a translation velocity component in the velocity V and the direction α _P exists in the target area. The speed thus obtained is hereinafter referred to as the speed component

[0040]

[Outside 1] It is written.

The contribution calculation 211 is performed by the speed component detection unit 20.
5. Intersection parameter space memory 254 detected at 5
Voting value S _L (α _P , β _P ) indicating the voting value distribution peak in
, The degree of contribution of each pixel or graphic element in the target area is determined.

Here, the following method will be described as an example. Each pixel in the spatiotemporal difference image D (x, y, t) representing the motion trajectory constructed in the spatiotemporal image memory 251 is:
In the three-dimensional Hough transform unit 202, the vote for each cell of the three-dimensional in voting space memory 252 pointed to voting patterns, such as curved surface S ₁ in FIG. 5 of the formula (2) is performed.

The speed obtained by the speed component detection unit 205
Voting peak S representing degree component_L (Α _P , Β_P ) Vote
2D normal parameter sky that may have contributed to the accumulation of
The set of cells in the inter-memory 253 is
Rewrite equation (10), which is the voting pattern of
Cell coordinates (α_P , Β_P )

[0044]

(Equation 8) Is the cell pointed to by the curve represented by.

Further, the three-dimensional voting space memory which can contribute to the voting value to the cell indicated by the expression (13) in the two-dimensional normal parameter space memory 253 is represented by the following expression (1) as S ₂ in FIG.
The curve of 3) is a set of cells pointed to by the column surface that is the bottom curve. Therefore, each pixel D (x, y,
t) is the velocity component

[0046]

[Outside 2] The set of cells in the three-dimensional voting space memory that can contribute to
A curve that satisfies both Equations (2) and (13) at the same time, that is, a curve where two curved surfaces in FIG.
Is a set of cells indicated by.

Here, each pixel D (x,
y, t) velocity component

[0048]

[Outside 3] The degree of contribution to this (referred to as the degree of contribution) is represented by a curve L
Among the above cells, the maximum value of the accumulated voting value is C
(X, y, t).

It should be noted that other methods of calculating the degree of contribution other than those described above can be used.

The contribution determining section 212 includes the contribution calculating section 21
Based on the degree of contribution of each pixel or graphic element calculated in 1, it is determined whether or not to assign the velocity component detected by the velocity component detection unit 205.

Here, the following method will be described as an example. The voting value in the intersection line parameter space memory 254 includes the influence of noise in an image, variation in speed, and the like. Excluding these effects, the pixel D (x, y, t) in the spatiotemporal difference image which is essential only for the contribution of the velocity component is specified,
It is necessary to add a speed component.

Therefore, the contribution C (x, y, t) of each pixel
, And a velocity component detected by the velocity component detection unit 205 is assigned only to pixels having a contribution degree equal to or greater than the threshold T. Here, the threshold value T is from the lowest value of the voting value in the intersection line parameter space memory 254 to the peak value S _L.
Given a ratio ε when (α _P , β _P ) is assumed to be 1.0, the voting value contribution by a pixel having a contribution C (x, y, t) equal to or greater than the threshold T gives The voting value accumulated at the peak point (α _P , β _P ) in the line parameter space is εS _L
Set the value of T such that ε is called a contribution ratio, and when an input pixel contains much additional noise, 1.
Give a value close to 0. If the contribution C (x, y, t) is less than the threshold value, it is determined that the pixel (x, y, t) has no speed. Therefore, the velocity component υ (x, y, t) of each pixel in the spatiotemporal difference pixel D is

[0053]

(Equation 9) Can be requested.

The velocity field construction unit 213 composes a velocity field by putting together the velocity components assigned by the contribution determination unit 212 for each pixel or graphic element included in the target area.

A specific example will be described below. FIG. 6 shows a case where a certain region is extracted from the input time-series image by the target region extracting unit 201, and one frame is displayed. FIG. 7 is an example of extracting edges and contours from the input image of FIG. In this example, the target is moving uniformly from left to right. FIG. 8 is an example in which the velocity field is obtained from the input of FIG. The contribution ratio ε was set to 0.98. It can be seen that edge motion is obtained for each pixel.

FIG. 9 shows that the image shown in FIG.
Is an image to which Gaussian noise is added. FIG.
Is a velocity field obtained from the magnitude of the brightness value of the image of the edge extracted from. FIG. 11 shows the velocity field obtained by the method of the present invention. The contribution ratio ε was set to 0.98. As can be seen from FIG. 10, the influence of noise is significantly reduced.

The memory 25 in the processing section 200 shown in FIG.
Each part except 1-254 can be realized as software,
These processes can be recorded on a recording medium such as an FD or a CD-ROM and read out and executed by a data processing device.

[0058]

As described above, according to the present invention, when a motion field (also referred to as an optical flow or a velocity field) of a time-series image is measured by using the Hough transform or voting, a certain point in the time-series image is measured. Hough transform or voting is performed on each pixel or graphic element included in the area, a peak of a voting distribution is detected in a parameter space (voting space) obtained as a result, and a peak value of each pixel or graphic element is calculated. Is calculated, and the velocity component of each pixel or graphic element is determined according to the magnitude of the contribution, so that a motion field of a time-series image that is robust to noise can be obtained.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a time-series image motion field measurement device according to an embodiment of the present invention.

FIG. 2 is a flowchart of a process of the apparatus of FIG. 1;

FIG. 3 is a diagram illustrating a polar coordinate expression of a plane in a three-dimensional space.

FIG. 4 is a diagram illustrating a method of expressing a straight line in a three-dimensional space.

FIG. 5 is a diagram illustrating a state of a voting pattern in a three-dimensional voting space.

FIG. 6 is a diagram showing one frame of a time-series image used in the apparatus of FIG. 1;

7 is a diagram illustrating one frame of a time-series difference image expressing edges and contours of a target obtained by the apparatus of FIG. 1;

FIG. 8 is a view showing a motion field obtained by the apparatus of FIG. 1;

FIG. 9 is a diagram showing one frame of a time-series image to which noise is added used in the apparatus of FIG. 1;

FIG. 10 is a diagram showing a speed field of a comparison object obtained from FIG. 9;

FIG. 11 is a diagram showing a velocity field obtained from FIG. 9;

[Explanation of symbols]

 Reference Signs List 100 input unit 200 processing unit 201 target area extraction unit 202 three-dimensional Hough transform unit 203 projection unit 204 intersection voting unit 205 speed component detection unit 211 contribution calculation unit 212 contribution determination unit 213 speed field configuration unit 251 spatiotemporal image memory 252 three-dimensional voting space memory 253 two-dimensional normal parameter space memory 254 intersection parameter space memory 300 output unit 401 to 410 step

Claims (15)

[Claims] 1. A time-series image motion field measurement method for inputting a time-series image and measuring a motion field of the image using Hough transform or voting, wherein each of the time-series images includes an arbitrary area included in an arbitrary area in the time-series image. Performing a Hough transform or voting on a pixel or graphic element, detecting a peak of a voting distribution in the resulting parameter space, and determining the degree of contribution to this peak value for each pixel or graphic element. A time-series image motion field measurement method including a calculating step, and determining a speed component of each pixel or graphic element based on the magnitude thereof, and measuring a motion field of the time-series image. 2. A time-series image motion field measurement method for inputting a time-series image and measuring a motion field of the image using Hough transform or voting, wherein the time-series image is obtained by adding a time axis to an image plane. When viewed as a three-dimensional space, an edge and a contour of a target included in the image are detected in a three-dimensional space, and a curved trajectory is detected. A tangent plane tangent to the trajectory is detected. Detecting the direction of the line of intersection by Hough transformation or voting, and measuring the velocity component in the target area, and voting value indicating the direction of the detected velocity in the parameter space obtained by Hough transformation or voting. Detecting the peak of the image, calculating the degree of contribution of each pixel constituting the motion trajectory in the target area to the peak, and calculating the velocity component of each pixel or graphic element according to the magnitude. Series image motion field measurement method when including the step of measuring the motion field of the time-series images Rukoto. 3. A time-series image motion field measurement method for inputting a time-series image and measuring a motion field of the image using Hough transform or voting, wherein the time-series image is obtained by adding a time axis to an image plane. When viewed as a three-dimensional space, a curved motion trajectory in which edges and contours of a target included in the image are drawn in a three-dimensional space is detected, and a tangent plane tangent to the motion trajectory is detected by a three-dimensional Hough transform. Detecting the direction of the line of intersection of the different tangent planes with another Hough transform or voting, measuring the velocity component in the target area, and detecting the velocity component in the parameter space obtained by the Hough transform or voting. Detecting a peak of the voting value indicating the direction of the velocity; calculating the degree of contribution of each pixel constituting the motion trajectory in the target area to the peak; and calculating each pixel or figure according to its magnitude. Series image motion field measurement method when including the step of measuring the motion field of the time-series images by obtaining the velocity component with the element. 4. The method according to claim 1, wherein the step of calculating the degree of contribution of each image to the peak value includes the step of calculating each pixel or graphic element included in the time-series image with respect to the peak value of the voting value in the parameter space used for detecting the velocity component. Contributes to the voting to the detected velocity component among the elements in the parameter space where the 3D Hough voting has been performed for each pixel or graphic element, and the maximum voting value is accumulated. 4. The time-series image motion field measurement method according to claim 3, wherein the voting value of the selected element is set. 5. The step of measuring a motion field of a time-series image,
Only pixels or graphic elements having a certain degree or more of the degree of contribution of each pixel or graphic element included in the time-series image to the peak value of the voting value in the parameter space used for detecting the velocity component are included in the parameter space. 5. The time-series image motion field measurement method according to claim 1, further comprising a step of giving a velocity component obtained from a peak position of the voting value of the time series image. 6. A means for performing Hough transform or voting on each pixel or graphic element included in a certain area in a time-series image, and a means for detecting a peak of a voting distribution in a parameter space obtained as a result. And a means for calculating the degree of contribution of each pixel or graphic element to the peak value, and means for determining the velocity component of each pixel or graphic element according to its magnitude. 7. A means for detecting, when a time-series image is viewed as a three-dimensional space obtained by adding a time axis to an image plane, a motion trajectory in which the edges and contours of a target included in the image are drawn in the space and time. Means for detecting a tangent plane tangent to the motion trajectory, means for detecting the direction of an intersection line intersecting the tangent plane by Hough transformation or voting, and means for voting distribution in a parameter space obtained by Hough transformation or voting. A means for detecting a peak and calculating a target velocity component, a means for calculating a degree of contribution of each pixel constituting a motion trajectory in spatiotemporal to this peak, and a speed of each pixel or graphic element according to its magnitude A time-series image motion field measuring apparatus having means for determining a component. 8. A three-dimensional Hough transform means for detecting a tangent plane in which an edge and a contour of an object are in contact with a curved motion trajectory drawn in time and space, and a Hough transform means for detecting a direction of an intersection line where the tangent planes intersect. The time-series image motion field measurement device according to claim 7, comprising: 9. A means for detecting the degree of contribution of each pixel to a peak, comprising: detecting a velocity component detected in a parameter space in which a three-dimensional Hough transform has been voted for each pixel or graphic element; Find the set of elements that contributed to the voting, search for the element with the largest voting value among them, and determine the maximum voting value as the detected velocity component of each pixel or graphic element included in the time-series image 8. The time-series image motion field measuring apparatus according to claim 7, wherein the time-series image motion field measuring device is means for calculating the degree of contribution to voting. 10. A method for determining a speed component of each pixel or graphic element, comprising the step of: contributing each pixel or graphic element included in the time-series image to a peak value of a voting value in a parameter space used for detecting the speed component. 7. A means for comparing a degree of the voting value with a certain threshold value and providing a velocity component obtained from the peak position of the voting value in the parameter space only to pixels or graphic elements having a degree of contribution equal to or greater than the threshold value. 10. The time-series image motion field measurement device according to any one of items 9 to 9. 11. A recording medium for recording a time-series image motion field measurement program for inputting a time-series image and measuring a motion field of the image using Hough transform or voting, wherein the arbitrary area in the time-series image is recorded. A process of performing Hough transform or voting on each pixel or graphic element included in, a process of detecting a peak of a voting distribution in a resulting parameter space, and a degree of contribution to the peak value. A recording medium storing a time-series image motion field measurement program having a process of calculating a pixel or a graphic element and a process of determining a speed component of each pixel or a graphic element based on the magnitude thereof and measuring a motion field of a time-series image . 12. A recording medium for recording a time-series image motion field measurement program for inputting a time-series image and measuring a motion field of the image using Hough transform or voting, wherein the time-series image is expressed in time on an image plane. When viewed as a three-dimensional space to which an axis is added, an edge and a contour of a target included in the image detect a curved motion trajectory drawn in the three-dimensional space, and a tangent plane tangent to the motion trajectory is detected. The process of detecting the direction of the intersection line where the different tangent planes intersect by Hough transformation or voting, and measuring the velocity component in the target area, and calculating the detected velocity in the parameter space obtained by Hough transformation or voting. A process of detecting a peak in the voting value indicating the direction, a process of calculating the degree of contribution of each pixel constituting the motion trajectory in the target region to the peak, and a process of calculating each pixel or Recording medium for recording a series images motion field measurement program when it has a process of measuring the motion field of the time-series images by obtaining the velocity component having the shape element. 13. A recording medium for recording a time-series image motion field measurement program for inputting a time-series image and measuring a motion field of the image using Hough transform or voting, wherein the time-series image is stored on an image plane in time. When viewed as a three-dimensional space to which an axis is added, a curved motion trajectory in which an edge and a contour of a target included in the image are drawn in the three-dimensional space is detected, and a tangent plane tangent to the motion trajectory is detected as a three-dimensional Hough. The process of detecting by the transformation and detecting the direction of the intersection line where the different tangent planes intersect by another Hough transformation or voting, and measuring the velocity component in the target area, and the parameter space obtained by the Hough transformation or voting A process of detecting a peak of a voting value representing a direction of the detected speed, a process of calculating a contribution of each pixel constituting a motion trajectory in the target region to the peak, Series image motion field measuring method when it has a process of measuring the motion field of the time-series images by obtaining the velocity component possessed by each pixel or graphic element by magnitude. 14. A process for calculating a degree of contribution of each pixel to a peak value, wherein each pixel or graphic element included in a time-series image is compared with a peak value of a voting value in a parameter space used for detecting a velocity component. Contributes to the voting to the detected velocity component among the elements in the parameter space where the 3D Hough voting has been performed for each pixel or graphic element, and the maximum voting value is accumulated. The recording medium according to claim 13, further comprising a process of setting a voting value of the selected element. 15. A process for detecting a time-series motion case,
Only pixels or graphic elements having a certain degree or more of the degree of contribution of each pixel or graphic element included in the time-series image to the peak value of the voting value in the parameter space used for detecting the velocity component are included in the parameter space. 2. A step of providing a velocity component obtained from the peak position of the voting value.
15. The recording medium according to any one of 1 to 14.