JP2002008042A - Action recognizing and processing device, and moving object action analyzing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve such a problem that it is hard to catch a moving object moving in a wide area with the sufficient resolution and it is hard to discriminate the degree of the action of the moving body. SOLUTION: A parameter input unit 2 inputs the control parameter for changing the range of a specified area and a concentration value in an image at a predetermined rate, and a diffusion computing unit 3 separates the moving image of the moving object into an object image and the background image, and assigns a predetermined concentration value to picture elements of an area in which the object image exists at a predetermined time, and changes the range of the described area and the concentration value with the lapse of time on the basis of the control parameter. An action recognizing unit 4 extracts the feature quantity from the distribution of the concentration value in a described area, and recognizes the action of the object on the basis of the collation of the feature quantity with a specified value.

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 a moving object from a captured moving image, and more particularly to a motion recognizing device suitable for use in monitoring or analyzing the behavior of the moving object.

[0002]

2. Description of the Related Art As a conventional technique for recognizing a motion of a person in an image, a method using an optical flow is known. This is, as shown in JP-A-10-334270,
In this method, a change in an image due to movement of an object in an image is regarded as an optical flow, and the flow is analyzed to recognize an action.

[0003] Another prior art is MIT Medi.
a Laboratory Perceptual Computing Section Technica
l As shown in Report No.402, extract the silhouette of the object, give it a density value according to time change,
Motion History Ima by integrating it on the time axis
There is a method of generating an image called ge (MHI) and classifying the motion of the object from the moment feature amount of the MHI.

[0004]

SUMMARY OF THE INVENTION However, Japanese Patent Application Laid-Open No. 10-3342
No. 70 focuses on the density distribution of the image,
It is difficult to perform processing when the feature does not exist as a clear pattern of shading. In particular, in the case of targeting a person moving in a wide area, it is difficult to capture a person image with a sufficient resolution, so that it is difficult to obtain a clear shading pattern and it is also difficult to recognize actions.

[0005] In addition, the method of classifying an action based on a feature of a temporal change in a person's shape using a silhouette image cannot identify the degree of the action. In addition, there is no consideration for a case where a moving object moving in a wide area is targeted.

[0006] The object of the present invention, in view of the state of the prior art,
Recognition of the movement of a moving object moving in a wide area. It is also necessary to recognize the degree of movement of the moving object.

[0007]

In order to achieve the above object, a parameter input unit inputs a control parameter for changing a range of a specific area in an image and a density value at a predetermined rate, and sets a parameter for diffusion. The calculation unit separates the moving image of the moving object into an object image and a background image, assigns a predetermined density value to a pixel in an area where the object image exists at a predetermined time, and calculates a time lapse based on the control parameter. Along with changing the range of the region and the density value, the motion identification unit extracts a feature amount from the distribution of the density value in the region, and performs the motion of the object based on the comparison between the feature amount and a predetermined value. recognize.

Also, a trajectory extraction processing section for obtaining a trajectory of a moving object from moving image data from a video camera photographing the movement of a person or the like and outputting it as trajectory data; An action identification unit that identifies the action of a moving object and outputs it as single action data, and identifies the action as a group from the trajectory data,
A behavior analysis processing unit that outputs as group motion data; and a display processing unit that performs statistical processing and three-dimensional graphics processing from the single operation data and the group motion data to output statistical data and CG video data. Analyze the motion of a moving object.

[0009]

Embodiments of the present invention will be described below. FIG. 1 shows an operation recognition method according to the present invention.
It is explanatory drawing of a cellular diffusion model. In the present invention, by the action of the person, a diffusion phenomenon occurs in the surrounding fluid,
Since the situation is analyzed using a model that is discretized into cells,
This motion recognition technique is called a cellular diffusion model. In addition, although there are various types of motions to be recognized, in this example, identification of a moving motion such as “walking” or “running” is targeted.

FIG. 1A is an image of a moving person photographed by a video camera or the like. Although only 4 × 4 pixels are shown in the figure, this is a part of the vicinity of the person in the image. Here, it is assumed that the person has just been photographed with a size of one pixel, and that there is a person image at a position as shown in the figure at time t0. In the present invention, since the information of the grayscale pattern of the human image is not used, the human image may be a binary image in which the silhouette of the human can be seen.

FIG. 2B shows a human image at time t1. Compared with FIG. 2A, the human image has moved to the right by one pixel. (C) shows the pressure state of the cellular diffusion model at t1. In this model, the portion of the background image other than the human image in the image is considered to be filled with a certain fluid. Now, from the position of cell A, the person is cell B
In this model, it is assumed that a person has disappeared in cell A and a person has appeared in cell B.
At this time, in the cell A, a negative pressure is generated by the disappearance of the existing person, and the fluid tends to flow in due to the diffusion phenomenon. Similarly, in cell B, the appearance of a person creates a positive pressure, and fluid tends to flow out due to diffusion. The pressure is 0 at the portion where there is no change.

At this time, a pressure difference of the fluid is generated in the cell as shown in (d). In this case, the disappeared portion of the person such as the cell A is the low pressure portion, and the appearance portion of the person such as the cell B is the high pressure portion. Will be called. In this figure, the hatched portion on the lower right indicates the low-pressure portion, and the hatched portion on the upper right indicates the high-pressure portion. The thicker the hatched line, the larger the pressure difference from the surroundings. The pressure is zero in the part without hatching.

(E) shows the diffusion state of the low-pressure section at time t2. In the low-pressure part, the fluid flows in by diffusion, so the pressure difference with the surroundings becomes smaller (the pressure rises)
However, the area of the low-pressure section is expanded.

The pressure of the hatched portion is a value obtained by adding a predetermined value to the pressure value of (d), and the value is stored in the pixel as the density value of the image. The spread of the area is obtained by applying the dilation calculation of the image processing. The dilation operation is a process of thickening a human image by one pixel. It is well known that one pixel of an input image P is Pij, one pixel of an output image Q is Qij, a human image is 1, and a background image is 0. Is defined as in Equation 1.

[0015]

## EQU1 ## When Pij or any of its eight neighbors is 1: Qij
= 1 In other cases: Qij = 0 This expansion operation can be executed by hardware of an image processing board connected to a personal computer, so that high-speed processing is possible.

(F) shows the diffusion state of the high pressure section in the same manner.
As the fluid flows out by diffusion, the pressure drops and the area of the high pressure section widens. Also in this case, the pressure value of the hatched portion is a value obtained by subtracting a predetermined value from the pressure value of (d), and the pressure value is stored in the pixel. The expanded area is similarly obtained by dilation calculation.

(G) is the result of adding the low pressure part of (e) and the high pressure part of (f), and this is the final pressure distribution at time t2. Here, since the absolute value of the pressure value is equal in the low-pressure section and the high-pressure section, the pressure value becomes zero by addition in the region where both are present. The addition of these pressure values is also executed by the image processing board as the addition of the density values storing the pressure values.

Next, referring to FIG. 2, an operation recognition method using a cellular diffusion model will be described. Here, it is assumed that the person image at time t0 is in the state of (a), t1 is in (b), t2 is in (c) and t3 is in (d), and the person is moving to the right by one pixel. .

First, as shown in FIG. 1, the pressure distribution accompanying the state change from (a) to (b) is as shown at (e) at time t1, becomes (f) at t2, and Spread out at t3
(g). Similarly, in the change from (b) to (c), the pressure distribution becomes (h) at time t2 and (i) at t3. Similarly, the change from (c) to (d) is as shown in (j) at time t3.

Here, in order to analyze the operation in the state from time t0 to time t3, the pressure distributions from (e) to (j) are added for each time. As a result, (k) is (e) as is, (l) is the sum of (f) and (h), (m) is (g) and (i)
(j) is added. In the motion recognition of the present invention, (k)
(M) is used in the pressure distribution pattern at each time.

FIG. 3 shows an example of pressure distribution when a person moves. Here, the person image is photographed in a part of the image in a state as shown in FIG. 2A, and the person is moving in the right-hand direction on the screen. This person image is a silhouette of the person in the image.

The pressure distribution is calculated by the method described with reference to FIG. 2, and the result is, for example, a distribution as shown in FIG. Then, a portion where the positive or negative pressure exists is extracted, and its circumscribed rectangle is obtained. Since there may be a case where a plurality of persons exist, in such a case, a label is attached to each circumscribed rectangle to distinguish them. Since the pressure is stored as a density value in the image memory, a circumscribed rectangle or a labeled result is obtained using the labeling calculation of the image processing board.

(C) shows the pressure distribution in the circumscribed rectangle as X
The axial pressure projection value is obtained. The pressure projection value is the value obtained by projecting and adding the pressure value of each cell in a specific axis direction, and specifying the label in the image and performing projection calculation on the image processing board to obtain the value. it can.

When the person moves, a convex high pressure projection value is generated on the side of the circumscribed rectangle in the moving direction (in this case, the right side), and a concave low pressure projection value is generated on the opposite moving direction side. That is the usual pattern. The movement of the person is recognized by matching with the pattern of the pressure projection value.

FIG. 4 illustrates a method in which Fourier analysis is applied to pattern matching in order to process movement recognition. (a) is a diagram showing a distribution of pressure projection values. This distribution is considered as a one-dimensional signal, and an amplitude and a phase spectrum are obtained by performing a discrete Fourier transform on the signal.

Assuming that there are N data of i = 0, 1, 2,..., N−1 in the discretized one-dimensional signal f (i), its Fourier coefficient Ck is well known. As shown, it is represented by Equation 2 (j is an imaginary unit).

[0027]

(Equation 2)

An amplitude spectrum and a phase spectrum are obtained from Ck. The measured value of (b) is obtained by Fourier transforming the fluctuation of the pressure projection value of (a) and obtaining a spectrum. At this time, the size of the pressure distribution differs depending on the position where the image is captured in the screen. Here, the size of the pressure distribution is normalized so that the size in the X-axis direction is all N pixels.

(C) is the spectrum of the moving person,
This is used as a reference value. The distribution of this spectrum represents the feature amount for the movement of the person. When it can be determined that the spectrum of the measured value in (b) is similar to the reference value in (c), the person is recognized as moving.

When judging a signal in a frequency domain such as a spectrum, since it is possible to cut and compare a harmonic portion where noise during shooting is likely to occur, there is an advantage that the signal is hardly affected by noise. . In the above example, a spectrum obtained by Fourier transform is used for matching between the measured value and the reference value. However, in addition to this, for example, for a function indicating the distribution of the pressure projection value, the sum of the residual squares of the function of the measured value and the reference value is calculated, and it is determined that the smaller the value is, the more similar the two are. Is possible.

FIG. 5 is an explanatory diagram of the action classification processing. FIG.
Although it can be recognized that the person is moving by the method described above, it is not possible to distinguish how fast the person is moving. Therefore, here, using the information of the circumscribed rectangle, the motion is classified into “standing”, “walking”, or “running”.

Assuming that the person moves in the X-axis direction, the circumscribed rectangle becomes longer in the X-axis direction as the moving speed increases. Since the pressure distribution has a shape corresponding to the amount of movement within a certain period of time, the size of the circumscribed rectangle has movement information. The aspect ratio of the circumscribed rectangle indicates the moving speed.

As shown in the figure, the length of the circumscribed rectangle in the X-axis direction is B
The length in the x- and Y-axis directions is By, and the aspect ratio is Bx / By. If the aspect ratio is large as in (a), the person is in a "running"state; if it is small as in (b), the person is in a "standing" state, and the middle between the two is It can be determined that the user is “walking”. The determination can be made in the same manner even when the movement has occurred in the Y-axis direction.

Next, an example of the operation recognition processing procedure of the present invention will be described with reference to the flowchart of FIG. Step 100
Then, the user uses a keyboard or a mouse to input control parameters for diffusion calculation. Specifically, it is a parameter for controlling the size of the density value assigned to the pixel, the ratio of the density value reduced by diffusion, the ratio of the area expanded by diffusion, and the like. Since these need to be changed depending on the shooting conditions at the time of recognition, the user makes adjustments here.

Step 101 means that the processing of step 102 is repeated for a predetermined number of frames. Step 102 is a step of creating a background image. Generally, a portion that does not change in the image is detected during a predetermined number of frames, and that portion may be used as the background image.

Step 103 means that the processing of steps 104 to 116 is repeated for the number of frames to be measured. In step 104, an image for one frame is read from the video camera into the image memory of the image processing board. In step 105, a person image is extracted. For this purpose, the difference between the background image created in 102 and the current image read in 104 is calculated, and this is further binarized to obtain a silhouette of a person as a person image. If both the background image and the current image are stored in the image memory, these calculations can be executed on the image processing board.

Step 106 means that the processing of steps 107 to 111 is repeated for a predetermined number of times of diffusion. Here, the number of times of diffusion is the number of times of diffusion required to eliminate the generated pressure difference, and is 3, for example, in the example of FIG.
In this case, assuming that the current time is t3, going back three frames, calculating sequentially from the image (a) at t0, (m)
Is calculated.

In step 107, a high pressure part is extracted. Figure 1
In the example of the above, the high-pressure part is the part of the background image at time t0 and t1
Is the part that became a human figure. In order to find the high voltage portion, a logical operation of the images (a) and (b) may be performed by the image processing board. Similarly, in step 108, a low pressure portion is extracted. The low-voltage part is a part of the human image at t0 and a part of the background image at t1, and is similarly calculated by a logical operation.

In step 109, the pressure diffusion state is determined. As described with reference to FIG. 1, the manner in which the pressure is diffused is calculated by an expansion operation of the image. The size spread by diffusion is determined in advance, and for example, it is specified that one pixel is expanded by one diffusion.

In step 110, the pressure value changed by the diffusion is corrected. As also described with reference to FIG. 1, the pressure in the low pressure section increases and the pressure in the high pressure section decreases due to diffusion. Here, too, the calculation is performed such that a predetermined pressure value is added or subtracted in one diffusion using a value set in advance.

In step 111, the pressure distribution is calculated. In the example of FIG. 2, the image of (m) is obtained, and (e) to (j)
Is calculated by adding pressure images such as

In step 112, a pressure projection value is calculated.
The pressure projection value is as shown in FIG. 3C, and is calculated from the pressure distribution by the projection calculation of the image processing board. In step 113, a discrete Fourier transform is performed. This obtains the spectral distribution of the pressure projection value by the method described with reference to FIG. Step 114 is a step of matching spectra,
As shown in FIG. 4, the measured value is compared with the reference value. If the similarity is high, it is determined that the person is moving.

Step 115 is a step of calculating the aspect ratio of the circumscribed rectangle, and classifying the motion as "running", "walking" or "standing" from the ratio as shown in FIG. Can be. Step 116 outputs the recognition result. Since the movement of the person in the image has been recognized in the processing so far, the result is output to a file on a disk or the like.

FIG. 7 is a block diagram of a motion recognition processing device according to one embodiment. The video camera 7 outputs a video signal to the image processing board 6. The image processing board 6 has an image memory and an image operation function such as a logical operation and addition / subtraction. When the image processing board 6 is controlled by the personal computer 5, the processing is executed according to the procedure shown in FIG.

The operation recognition processing unit 1 is provided inside the personal computer 5.
Is implemented by software or hardware. Here, the parameter input unit 2 of the motion recognition processing unit 1 executes step 100 of the flow of FIG. Diffusion calculator 3
Executes the processing from step 101 to step 111, processes the enlargement of the area to which the density value is assigned by the dilation calculation of the image with the lapse of time, and changes the absolute value of the density value so as to decrease. . Then, an image created by changing the range of the area and the density value over time is displayed on a display.

The operation discriminating unit 4 performs the steps 1 to
Execute up to 16 processes. A spectrum is calculated by performing a Fourier transform from the distribution of the density values, using a projected density value obtained by projecting the distribution of the density values in a predetermined axial direction as a feature quantity, and comparing the spectrum with a predetermined value based on the spectrum. Further, a circumscribed rectangle circumscribing the area where the density value is distributed is calculated, and the operation is recognized based on the ratio of the length of the vertical and horizontal sides of the circumscribed rectangle. The distribution of the density values is examined for density fluctuations in a predetermined section, and when a fluctuation larger than the predetermined value is detected, it is determined that a portion where an object moves exists in the section.

In this embodiment, the image processing portion is executed by hardware called an image processing board 6.
Without using this, it is also possible to execute all with the software in the personal computer.

FIG. 8 is a block diagram showing a case where a system for analyzing the motion of a person is constructed using the motion recognition processing according to the present invention. The moving image data from the video camera 7 is input to the motion recognition processing 1 and the trajectory extraction processing 20.

In the trajectory extraction process 20, the object extracting unit 21 extracts a moving object such as a person or an object in the image, and
Identifies and tracks each moving object, and the 3D coordinate conversion unit 23 converts the image coordinates into three-dimensional spatial coordinates and outputs the converted coordinates as trajectory data. The content of the trajectory data is, for example, the position coordinates on a plane where the person was present recorded at each time.

Since the operation recognition result output from the personal computer 1 indicates the operation of the moving body alone, it is described as single operation data here. Here, for a person, for example, an identifier indicating an operation type such as “running” or “standing” is recorded at each time.

The behavior analysis processing 30 is, for example, a part for analyzing a motion as a group of persons, and includes a feature amount estimating unit 31 for obtaining a feature amount of a motion and a matching unit 32 for collating the feature amount. The behavior analysis processing 30 receives the trajectory data as input, and outputs group movement data describing the meaning of the movement of the group. As an example of this data, in situations where many people gather, such as commercial facilities and public facilities, as an example of security application of person monitoring, identifiers that represent the meaning of collective actions such as "avoidance" and "assembly" are used. Are recorded at each time.

The display processing 40 receives the single operation data and the group operation data as input, and first outputs the statistical data by the statistical analysis unit 41. The contents of the statistical data are information such as the moving distance of each person and the number of classified operations, and are displayed together with the CG image on a display or the like. Next model operation generator
Reference numeral 42 denotes a part for determining the joint angle of each joint in order to move the three-dimensional CG human body model. Then, the rendering unit 43 creates a CG image from the human body model and outputs this as CG image data.

The present invention can be used not only for security but also for game analysis in sports such as soccer. In this case, the single motion data indicates the recognition result of the motion of each player, and the collective motion data indicates the recognition result of the team play.

According to this embodiment, not only the position of the person is displayed, but also the identifier of the operation independently performed can be provided for each person, so that information effective for analyzing the operation of the person can be provided. There are effects that can be provided.

In the embodiment described above, in the movement recognition processing, the spectrum is obtained from the pressure projection value as shown in FIG. 4 and used as a feature for recognition. However, other feature can be used. .

In the embodiment shown in FIG. 9, the recognition is made by checking the average value of the pressure projection values. When the person moves from left to right as shown in FIG. 9A, the peak of the distribution of the pressure projection values always occurs on the right side, and the valley of the distribution is on the left side. As shown in FIG. 4B, X = Xmin to X
Assuming that the distribution of the pressure projection values is obtained in the section of = Xmax, the center of gravity X = G for this distribution shape is obtained. The average value of the pressure projection values on the left side of G is Px0, and the average value on the right side of G is Px1. Here, if the absolute values of Px0 and Px1 are equal to or larger than a predetermined threshold value and Px1> Px0, the person is recognized as moving to the right. The same applies to the movement in the Y-axis direction.

According to this embodiment, the average value of the pressure projection values before and after the center of gravity can be checked, so that the calculation is simple and the processing can be performed at high speed.

In the above embodiment, the motion of the person in response to the movement of the entire body is recognized. However, it is also possible to recognize the part of the body that has moved.

FIG. 10 shows an example of recognition of a moving part.
As shown in the figure, the pressure value due to the movement of the person is changed from X = Xmin to X
= Xmax and Y = Ymin to Y = Ymax, and there is a pressure projection value obtained by projecting in the X-axis and Y-axis directions. Here, the midpoint in the X-axis and Y-axis directions is Xmid and Ymid, and the moving part of the human image is
It is assumed that the image is divided into regions such as four circled numbers shown in FIG. It is assumed that thresholds for determination are Cx and Cy, and a moving part exists in a region where the absolute value of the pressure projection value exceeds these values.

In the example shown in the drawing, only the region at the lower left portion moves largely, and the other regions hardly move. In such a case, it can be recognized that only the right leg of the person has largely moved. According to the present embodiment,
The advantage is that you can know the part of the body that has moved significantly, and from this, you can recognize the movement more finely, for example, "Kick with your right foot."

[0061]

According to the present invention, there is an effect that a silhouette of a moving object moving in a wide area can be stably obtained and action recognition becomes possible. Further, since the variation amount of the silhouette shape is converted into the density and stored, it is possible to recognize not only the motion classification but also the degree of the motion by examining the density distribution.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a cellular diffusion model.

FIG. 2 is an explanatory diagram of motion recognition using a cellular diffusion model.

FIG. 3 is a view showing an example of a pressure distribution of a human image.

FIG. 4 is an explanatory diagram of a movement recognition process.

FIG. 5 is an explanatory diagram of a motion classification process.

FIG. 6 is a flowchart of a motion recognition process according to an embodiment of the present invention.

FIG. 7 is a configuration diagram of a motion recognition processing device according to an embodiment of the present invention.

FIG. 8 is a configuration diagram of a person motion analysis system according to one embodiment of the present invention.

FIG. 9 is an explanatory diagram of motion recognition based on an average value.

FIG. 10 is an explanatory diagram of recognition of a moving part.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Motion recognition processing part, 2 ... Parameter input part, 3 ... Diffusion calculation part, 4 ... Motion identification part, 5 ... Personal computer, 6 ... Image processing board, 7 ... Video camera, 20 ... Locus extraction processing part, 30 ...
Behavior analysis processing unit, 40 ... display processing unit.

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Kazuya Takahashi 7-1-1, Omika-cho, Hitachi City, Ibaraki Prefecture F-term in Hitachi Research Laboratory, Hitachi Ltd. 5C023 AA04 AA38 BA04 DA01 DA08 5C054 AA01 EA01 FC03 FC12 FC13 FC16 FD03 FE13 GA04 GB15 HA18 5L096 CA22 EA02 EA43 FA18 FA23 FA35 FA70 FA81 GA21 GA34 HA05 MA07

Claims (12)

[Claims] 1. A parameter input unit for inputting a control parameter for changing a range and a density value of a specific area in an image at a predetermined rate, and separating a moving image of a moving object into an object image and a background image. A diffusion calculation unit that assigns a predetermined density value to a pixel in an area where an object image exists at a predetermined time, and changes the range of the area and the density value over time based on the control parameter; A motion identification unit that extracts a feature amount from the distribution of the density values in (1) and recognizes a motion of the object based on a comparison between the feature amount and a predetermined value. 2. The method according to claim 1, wherein the diffusion calculation unit assigns a density value representing a negative value to the pixel when the pixel which is currently a background image was previously an object image, When the pixel which is the object image was previously a background image, a density value expressing a positive value is assigned to the pixel, and a density value expressing a value of 0 is assigned to a portion which does not change. Motion recognition processor. 3. The diffusion calculation unit according to claim 1, wherein the diffusion calculation unit enlarges the area to which the density value is assigned and changes the absolute value of the density value so as to decrease with time. Characteristic motion recognition processing device. 4. The motion recognition processing device according to claim 1, wherein the diffusion calculation unit processes the expansion of the area to which the density value is assigned by an expansion operation of an image. 5. The diffusion calculation unit according to claim 1, wherein the diffusion calculation unit outputs and displays an image created by changing the range of the area and the density value over time on a display. A motion recognition processing device characterized by the following. 6. The operation according to claim 1, wherein the operation identification unit uses a projection density value obtained by projecting the distribution of the density values in a predetermined axial direction as the feature amount. Recognition processing device. 7. The operation according to claim 1, wherein the operation identification unit calculates a spectrum by performing a Fourier transform from the distribution of the density values, and compares the spectrum with a predetermined value based on the spectrum. Recognition processing device. 8. The operation identification unit according to claim 1, wherein the motion identification unit calculates a circumscribed rectangle circumscribing the area where the density value is distributed, and calculates a ratio of lengths of the vertical and horizontal sides of the circumscribed rectangle. A motion recognition processing device for recognizing a motion based on the motion recognition. 9. The motion identification unit according to claim 1, wherein the motion identification unit calculates an average value of densities in a predetermined section with respect to the distribution of the density values, based on a magnitude relationship between the average values. An operation recognition processing device for recognizing an operation. 10. The motion identification unit according to claim 1, wherein the action discriminating unit investigates a density variation in a predetermined section of the density value distribution, and detects a large variation larger than a predetermined value. In such a case, it is determined that there is a moving part of the object in the section. 11. A trajectory extraction processing unit that obtains a trajectory of a moving object from moving image data from a video camera that captures a moving motion of a person or the like, and outputs the trajectory as trajectory data. An action identification unit that identifies the action of the moving object and outputs the action data as independent action data, an action analysis processing unit that identifies the action as a group from the trajectory data and outputs the action as group action data, A motion analysis system for a moving object, comprising: a display processing unit that performs statistical processing and three-dimensional graphics processing from the collective motion data and outputs statistical data and CG video data. 12. The apparatus according to claim 11, wherein the operation identification unit includes a parameter input unit for inputting a control parameter for changing a range and a density value of a specific area in the image at a predetermined ratio. The moving image is separated into an object image and a background image, a predetermined density value is assigned to a pixel in an area where the object image exists at a predetermined time, and the range of the area and the density are determined with the lapse of time based on the control parameters. A diffusion calculation unit that changes a value; and an operation identification unit that extracts a feature amount from the distribution of the density values in the region and that recognizes an operation of the object based on a comparison between the feature amount and a predetermined value. A motion analysis system for a moving object.