JP3403363B2 - 3D continuous motion verification system - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a verification device for three-dimensional continuous motion, and more particularly to a verification device for three-dimensional continuous motion which compares the continuous motion of an operator with a reference motion and evaluates the degree of approximation.

[0002]

2. Description of the Related Art An example of a conventional three-dimensional continuous motion test apparatus of this type is disclosed in Japanese Patent Application Laid-Open No. 7-185131 [A63F 9/22] published on July 25, 1995. There is. In this television play system, a player whose markers are attached to each body part is imaged at a predetermined cycle using a video camera, and the movement position of the marker is detected from the imaged image. The display position and display attitude of the game character are calculated from the movement position of the marker, and an image of the game character having a predetermined attitude is displayed on the display screen corresponding to the calculated display position based on this display attitude. Thus, the display control of the game character is executed every time the player is imaged, and the game character is operated in response to the movement of the player.

[0003]

However, in this prior art, since the player is imaged at a predetermined cycle and the motion of the player is detected based on the posture of the player when the image is taken, the player is extremely extreme. When making a motion, it was difficult to estimate the motion. That is, it is difficult to accurately recognize a motion, and when this conventional technique is applied to a device that tests a continuous motion, the motion cannot be accurately tested.

Therefore, a main object of the present invention is to provide a new three-dimensional motion inspection apparatus capable of accurately inspecting motion.

[0005]

According to the present invention, a motion detecting means for detecting a continuous motion of a person in real time based on a plurality of images photographed from a plurality of viewpoints, and a continuous motion detected by the motion detecting means is used as a reference. Comparing means for comparing the motion with each segment, and a verification means for testing the approximation degree of the continuous motion corresponding to the reference motion based on the comparison result of the comparing means,
The ratio of each segment in the three-dimensional continuous motion tester
For comparison, the reference motion acquired in real time is applied to the appropriate boundary.
Further, a dividing means for dividing by
Within the allowable fluctuation range set for each fixed period, the evaluation value
Use the boundary as the point when the highest detection result is obtained.
It is a verification apparatus for three-dimensional continuous motion characterized by the above.

[0006]

In the three-dimensional motion verification apparatus of the present invention, the motion detecting means detects a continuous motion of a person (operator) in real time based on a plurality of images taken from a plurality of viewpoints by using a plurality of cameras, for example. . The continuous operation of the operator is
It is compared with the reference motion segment by segment. The verification means tests the degree of approximation of the continuous motion to the standard motion based on the result of comparison between the continuous motion and the standard motion. Since the reference motion and the continuous motion of the operator are compared for each segment, the motion can be accurately determined. And, in comparison means
A dividing means is provided for each segment comparison.
The dividing means is a model that corresponds to the reference motion acquired in real time.
Partition the segmentation data at appropriate boundaries.
At this time, the dividing means is based on the reference motion acquired in real time.
Corresponding motion data based on a predetermined time interval
Of the most allowable detection results within the allowable variation range set by
It divides when the value is high as a boundary. That is, the dividing means
It is possible to automatically divide the reference motion into segments by
it can.

For example, since the display means visually displays the test result, the operator can know to what degree the motion of the CG character is approximated. You can also see how much you have improved.

[0008]

[0009]

[0010]

[0011]

The comparison means is a DP matching method,
Since the reference motion and the continuous motion are compared, it is possible to convert the time axis and compare only the motions.

Furthermore, the unit test means tests the degree of approximation for each segment, and the comprehensive test means tests the degree of approximation of all the segments. Therefore, it is possible to know the score of each segment and the total score.

[0014]

According to the present invention, since the continuous motion and the reference motion are compared for each segment, the motion can be accurately determined.

The above-mentioned objects, other objects, features and advantages of the present invention will become more apparent from the detailed description of the following embodiments made with reference to the drawings.

[0016]

EXAMPLES Referring to FIG. 1, a system using the three-dimensional continuous motion test apparatus of this example (hereinafter, simply referred to as "system").
Say. ) 10 includes color cameras (hereinafter, simply referred to as “cameras”) 12 to 22. The cameras 12 to 16 are connected to the first motion capture system 24, and the cameras 18 to 22 are connected to the second motion capture system 3.
Connected to 2. First motion capture system 2
4 is connected to the first segment dividing device 26 and also to the input terminal S1 of the switch SW1. The first segment dividing device 26 is connected to the first motion data storage device 28. The first motion data storage device 28 is connected to the input terminal S2 of the switch SW1 and also to a three-dimensional continuous motion verification device (hereinafter, simply referred to as “motion verification device”) 38. Switch S
The output terminal S3 of W1 is connected to the first image generation device 30.

The second motion capture system 32 is connected to the second segment dividing device 34 and also to the second image generating device 36. The second segment dividing device 34 is connected to the motion verification device 38. The first image generation device 30, the second image generation device 36, and the motion verification device 38 are connected to the monitor 40.

In this system 10, a person other than the operator performing the reference operation, for example, an instructor, is used as the camera 1.
2 to 16 are photographed, and the operator who imitates the reference movement and performs continuous movement is photographed by the cameras 18 to 22. In addition, the cameras 12 to 22 are provided for taking images from the front, side, and top (upright) of a person. The camera 12
And 18 are for the front, cameras 14 and 20 are for the side, and cameras 16 and 22 are for the elevation. Color images corresponding to the images captured by the cameras 12 to 16 on the instructor side are provided to the first motion capture system 24. Also, the camera 1 on the operator side
Color images corresponding to the video images captured in 8 to 22 are provided to the second motion capture system 32.

As shown in FIG. 2, the first capture system 24 and the second motion capture system 32 have the same structure. That is, the image processing devices 50 to 54 corresponding to the cameras 12 to 16 (or 18 to 22).
Is provided. The image processing devices 50 to 54 are connected to the three-dimensional coordinate calculation device 56. Image processing devices 50 to 54
Is a computer having a processing capability of a personal computer, processes color images output from the cameras 12 to 16 (or 18 to 22), and characterizes the body on each image, that is, head vertex, hand, elbow. , Two-dimensional coordinates of characteristic body parts such as knees are detected. It should be noted that the image processing devices 50 to 54 perform the processing in a time-division manner to
Individual computers can be replaced by sharing two more powerful computers.

Information on the two-dimensional coordinates of the characteristic points obtained by the image processing devices 50 to 54 is given to the three-dimensional coordinate calculation device 56. The three-dimensional coordinate calculating device 56 is also a computer having a processing capacity of a personal computer, and the Japanese Patent Application No. 11-11164 previously filed by the applicant of the present application.
As disclosed in No. 5, two-dimensional coordinates of feature points are restored to three-dimensional by the principle of triangulation, and whether the restored data (three-dimensional coordinate data) is valid or not is initially determined, for example. Verification is performed based on the set physical data of the person. In other words, the three-dimensional posture (three-dimensional posture) of the operator or the instructor is estimated based on the color images obtained from a plurality of viewpoints, that is, the color images corresponding to the front face, the side face, and the elevation face.
The three-dimensional posture data (posture data) thus estimated is output from the three-dimensional coordinate calculation device 56, that is, the first motion capture system 24 and the second motion capture system 32.

The posture data output from the first motion capture system 24 is used as the first segment dividing device 2
6 through the switch SW1 to the first
It is given to the image generation device 30. The first segment dividing device 26 is also one computer having a processing capability of a personal computer, and divides the posture data (motion data) continuously given from the first motion capture system 24 into segments.

For example, the motion data corresponding to the previously acquired reference motion is manually divided into segments at arbitrary time intervals T1, T2, ..., T5, as shown in FIG. 3 (A). Further, the motion data corresponding to the reference motion acquired in advance is divided into segments of a predetermined time interval T, as shown in FIG. In this embodiment, since the continuous motion is detected, the time interval T is 2-5.
Set between seconds. In these two division methods, the reference motion is acquired in advance, so the camera 1 shown in FIG.
2-16 and the first motion capture system 24
May be provided separately, and the first segment dividing device 26
Can be omitted. That is, the motion data (divided data) divided into the segments may be given to the first motion data storage device 28.

Further, as shown in FIG. 3 (C), with reference to the predetermined time interval T, the boundary of the segment within the allowable fluctuation range (T-dt to T + dt) determined by the minute time dt before and after the predetermined time interval T. Can be determined. In other words, the time point at which the detection result with the highest evaluation value is obtained within this variation allowable range (T-dt to T + dt) is determined as the segment boundary. Therefore, it is possible to automatically detect the boundary based on the motion data corresponding to the reference motion and divide it into segments.

Specifically, the processing is performed according to the flow chart shown in FIG. That is, when the first segment dividing device 26 starts the segment dividing process, the time t and the segment number n are initialized in step S1. That is, t = 0
And n = 1. In the following step S3, the maximum evaluation value max and the time index tindex are initialized. That is, max = 0 and tindex = 0 are set. Further, in step S5, the count value i of the loop counter is initialized. Specifically, i = t + T-dt is set. In the following step S7, the maximum value max is the evaluation value E.
(I) Determine if it is greater than. If “YES” in the step S7, that is, the maximum value max is the evaluation value E.
If it is larger than (i), the process directly proceeds to step S11.

On the other hand, if "NO" in the step S7,
That is, if the maximum value max is smaller than the evaluation value E (i), the evaluation value E (i) is substituted for the maximum value max and the count value i is substituted for the time index tindex in step S9, and the process proceeds to step S11. In step S11, it is determined whether the count value i is larger than t + T + dt. Step S
If “NO” in 11, it is determined that all allowable fluctuation ranges have not been searched, and the process returns to step S7. On the other hand, if “YES” in the step S11, it is determined that all the variation ranges have been searched, and the process proceeds to a step S13.

In step S13, the boundary of the motion data, that is, the segment is changed from M (t) to M (tindex).
Up to the period. In the following step S15, the time index tindex is substituted for the time t, the number of segments n is incremented, and the process proceeds to step S17. In step S17, it is determined whether the sum of the time t and the predetermined time T is longer than the time length of the motion data, that is, the length L of the sequence. If "NO" in the step S17, it is determined that all motion data are not divided into segments, and the process returns to the step S3. On the other hand, step S
If “YES” in 17, it is determined that all the motion data are divided into segments, a final segment is created in step S19, and the process ends. That is, since the motion data corresponding to the final segment may be shorter than the length of the final segment, the segment is shortened to the length corresponding to the motion data.

The division data of the reference motion output from the first segment division device 26 is given to the first motion data storage device 28. In the same manner, the motion data of the continuous motion of the operator output from the second motion capture system 32 is divided into segments by the second segment dividing device 34. In other words, divided data corresponding to continuous operation is generated. The divided data corresponding to the standard motion and the continuous motion is the motion verification device 3
It is read according to the instruction of 8. Therefore, the divided data corresponding to the reference motion is supplied to the motion verification device 38 and the first image generation device 30 via the switch SW1. Further, the divided data corresponding to the continuous motion is given to the motion verification device 38.

The motion verification device 38 is also a computer having a processing capacity of a personal computer, and compares the standard motion and the continuous motion. Then, based on the comparison result, the degree to which the continuous motion approximates the reference motion is evaluated. Further, the evaluation result is displayed on the monitor 40. For example, the degree of approximation can be displayed as a ratio, the ratio can be displayed as a score, and further can be displayed at a predetermined level (ranks A, B, C, D and E). It is also possible to display both the approximation ratio or the score and a predetermined level.

The first image generation device 30 uses a computer graphics (CG) method based on the divided data of the motion data corresponding to the reference motion output from the first motion capture system 24 or the first motion data storage device 28. The person image of the instructor called "avatar" is reproduced or reproduced in the virtual three-dimensional environment. Therefore, the reference motion of the instructor is
It is displayed (reproduced) on the monitor 40 by the CG character. Further, the switch SW1 can be freely switched,
The output of the first motion capture system 24 or the output of the first motion data is selected. In particular,
When acquiring the reference motion of the instructor in real time, the switch SW1 is connected to the input terminal S1 side,
The output of the motion capture system 24 is selected. On the other hand, when comparing the pre-stored reference motion of the instructor and the continuous motion of the operator in real time, the switch SW1 is connected to the input terminal S2 side and the output of the first motion data storage device 28 is selected.

Further, the second image generating device 36, based on the motion data output from the second motion capture system 32, in the same manner as the first image generating device 30, the operator's human image in the virtual three-dimensional environment. Reproduce or play. Therefore, the continuous operation of the operator is C
It can be displayed on the monitor 40 by the G character.

As described above, the reference motion of the instructor and the continuous motion of the operator are compared, and the comparison result is evaluated. There are two processes for this comparison and evaluation. The first method is a method of using motion data corresponding to a previously acquired reference motion as shown in FIGS. 5 and 6 for comparison. Further, the second method is a method of making a comparison using motion data corresponding to the reference motion earned in real time as shown in FIGS. 8 to 10. That is, as described with reference to FIG. 4, the method of automatically determining the boundaries of the segments and dividing the segments is adopted.

More specifically, first, in the first method, as shown in FIG. 5, when the motion verification device 38 starts the process of comparison and evaluation, in step S31, the linearity of the velocity unevenness path (hereinafter, (It is simply called "linearity".) Initialize Ps. That is, as shown in FIG. 7, it occurs between the path (input data S) corresponding to the continuous motion of the operator indicated by the solid line in the DP matching process and the teacher data T corresponding to the reference motion of the instructor indicated by the dotted line. The initial value of speed unevenness is set. Ps (0,1, ..., N)
Set to = 0. In the following step S33, the number of segments n is initialized. That is, n = 1 is set. And
In step S35, the counter that counts the time t in one segment is reset (t = 0) and started. Next, in step S37, the first motion data storage device 28 is instructed to read the divided data (teacher data T (n, t)) of the reference motion, and the teacher data T (n, t
Monitor the CG character that performs the standard motion according to t) 4
Display at 0. In step S39, the second segment division device 34 is instructed to take in the division data (input data S (t)) of the operator. In the following step S41, time t is calculated from the time interval L (n) corresponding to the segment.
To determine if is large. That is, it is determined whether or not the time corresponding to one segment has elapsed. If "NO" in the step S41, it is determined that the time interval L (n) corresponding to one segment has not elapsed, and the step S41
Return to 37. On the other hand, if “YES” in the step S41, it is determined that the time interval L (n) corresponding to one segment has elapsed, and the distance value D (n) and the path P (t) are determined by the DP matching method in a step S43. To calculate. here,
The distance value D (n) is the teacher data T after the time axis is converted.
It is an error amount between (n, t) and the input data S (t).

As shown in FIG. 6, in a succeeding step S45, the loop counter is reset (t = 0) and started. In step S47, speed unevenness Ps (n) at time t is calculated. Specifically, the speed unevenness is calculated according to the equation 1.

[0034]

## EQU1 ## Ps (n) = Ps (n) + | P (t) -t | where P (t) is the value of the path including the velocity unevenness, and is input as the teacher data T (n, t). When the data S (t) completely matches, P (t) = t. That is, the case where the inclination is t is the ideal value of the path.

In a succeeding step S49, it is determined whether or not the time t is larger than the time interval L (n) corresponding to one segment. That is, it is determined whether the time interval L (n) corresponding to one segment has elapsed. Step S4
If "NO" in 9, it is determined that the time interval L (n) corresponding to one segment has not elapsed, and step S4
Return to 7. On the other hand, if “YES” in the step S49, it is determined that the time L (n) corresponding to one segment has elapsed, and an evaluation value (score) Ts in one segment is calculated in a step S51. Specifically, the calculation is performed according to Equation 2. However, a and b are weighting factors and are set according to the length of the segment and the like.

[0036]

[Formula 2] Ts = a * D (n) + b * Ps (n) However, a + b = 1.

As can be seen from Equation 2, the score Ts is represented by the distance value D (n) and the path Ps (n),
The smaller the value of Ts shown in Expression 2, the closer the continuous motion of the operator is to the reference motion of the instructor.

When the score Ts for one segment is calculated,
Increment the number of segments (n =
n + 1), and it is determined in step S55 whether the segment number n is larger than the total segment number N. That is,
Determine whether comparisons and assessments have been performed for all segments. If "NO" in the step S55, it is determined that the comparison and evaluation have not been completed for all the segments, and the process directly returns to the step S35. That is, the comparison and evaluation for the next segment is performed. On the other hand, if “YES” in the step S55, it is determined that the comparison and evaluation processes for all the segments are completed, and a total score (total score) in a step S57.
To calculate. Specifically, the total score (TotalScore) is calculated according to Equation 3.

[0039]

## EQU00003 ## TotalScore = .SIGMA. {Ts (n) * L
(N) / ΣL (n)} In this way, the weight is weighted to calculate the score for each segment, and these are averaged to calculate the total score.
Appropriate evaluation can be performed as compared with the case of averaging scores for each segment without weighting. The score and level are determined according to the calculated total score. As described above, as Ts (n) is smaller, the continuous motion of the operator is closer to the standard motion. Therefore, the smaller the total score, the higher the score and level.
Further, although the score is calculated for the entire movement of the body, the score for each part of the body may be displayed.

Further, in the second method, as shown in FIG. 8, when the comparison and evaluation processes are started, step S7 is performed.
1 initializes each variable. Specifically, the linearity Ps
(0, 1, ..., N) = 0, start position S of input data
p = 0, number of segments n = 1, start position Tp = 0 of teacher data T (t), and time t = 0 required for one sequence
Set to. In a succeeding step S73, the boundary Out of the segment is initialized. That is, Out = Tp + L
Set to (n). In step S75, the time interval L corresponding to the segment of the teacher data T (t) at time t
It is determined whether it is larger than the sum of (n). That is,
It is determined whether all the teacher data T (t) have been presented. If “NO” in the step S75, it is determined that there is remaining teacher data T (t), the teacher data T (t) is presented in a step S77, and the process proceeds to a step S79. On the other hand, if “YES” in the step S75, it is determined that all the teacher data T (t) have been presented, and the process directly proceeds to the step S79.

In step S79, the input data S (t) is fetched, and in step S81, it is determined whether or not the time t is larger than Out + dt. That is, it is determined whether the sequence range is exceeded. Step S81
If “NO” in S, that is, if the sequence range is not exceeded, the process returns to step S75. On the other hand, step S
If "YES" in 81, that is, if it exceeds the range of the sequence, the variable min and the time index index are initialized in step S83. That is, min = maximum value and index = 0.

As shown in FIG. 9, in a succeeding step S85, the count value e of the loop counter is set (reset) to Out-dt and the operation is started. Next, in step S87, it is determined whether or not the variable min is larger than the absolute value of the difference between the input data S (e) and the teacher data T (e) with respect to time e. If "NO" in the step S87, that is, if the absolute value of the difference is smaller than the variable min, the process directly proceeds to the step S91. On the other hand, if “YES” in the step S87, the absolute value of the difference is substituted in the variable min in the step S89, the count value e at that time is substituted in the time index index, and the process proceeds to the step S91. In this way, when the matching is best (minimum difference) between the teacher data and the input data, that is, when the evaluation value is high, the segment boundary is set.

In step S91, it is determined whether the time T exceeds the sequence range. Step S9
If "NO" in 1, that is, if the sequence range is not exceeded, the process returns to step S87. On the other hand, step S
If “YES” in 91, that is, if it exceeds the range of the sequence, in step S93, the distance value D (n) and the path P (x) are set to T (Tp to Out) and S (Sp to index).
) Is calculated by the DP matching method. In a succeeding step S95, a value obtained by subtracting the start time Sp of the input data S (t) from the time index index is substituted for the time Sq, and the time difference R (n) between the teacher data T (t) and the input data S (t) is substituted. The absolute value of the difference between the time interval L (n) and the time Sq is substituted into. That is, the time axis is converted.

As shown in FIG. 10, the following step S97
Then, reset the count value x of the loop counter (x =
0) and start, and speed unevenness is calculated in step S99. Specifically, the speed unevenness is calculated according to the equation (4).

[0045]

## EQU00004 ## Ps (n) = Ps (n) + | P (x)-(x *
L) / Sq | In this case, unlike the first method using the previously acquired reference motion, the gradient of the teacher data T (t) does not become 1. Therefore, as can be seen from FIG. The ideal value is shown by the slope ((x * L) / Sq). However, the teacher data T (t) is a function of y = αx (α ≠ 0, 1).

Subsequently, in step S101, it is determined whether the count value x of the loop counter is larger than the time Sq. If “NO” in the step 101, that is, if the count value x is smaller than the time Sq, the step S99.
Return to. On the other hand, if “YES” in the step S101, that is, if the count value x is larger than the time Sq, the score Ts (n) of one segment is calculated in a step S103. Specifically, it is calculated according to Equation 5. Also,
Similar to the above, a, b and c are weighting factors.

[0047]

## EQU00005 ## Ts (n) = a * D (n) + b * Ps (n) +
c * R (n) where a + b + c = 1.

When the speed unevenness is calculated, step S10
At 5, the start position Tp of the teacher data is changed. Specifically, the time interval L (n) of the evaluated segment is added to the start position Tp. That is, Tp = Tp + L (n). Succeedingly, in a step S107, the number of segments is incremented (n = n + 1). And step S1
At 09, it is determined whether the segment number n is larger than the total segment number N. That is, it is determined whether the comparison and evaluation processes have been executed for all the segments. If “NO” in the step S109, it is determined that there are remaining segments, and the process returns to the step S73. On the other hand, if “YES” in the step S109, it is determined that the comparison and evaluation processes have been executed for all the segments, and the total score is calculated in a step S111,
The process ends. Specifically, the total score is calculated according to equation 6.

[0049]

[Equation 6] TotalScore = Σ {Ts (n) * L
(N) / ΣL (n)} As in the first method, the score and level are determined according to the total score. Also, the smaller the total score, the higher the score and level.

Such a system 10 can be applied to a practicing device such as a kempo or tai chi, a dance lesson device, a dance game machine, or the like. For example, as shown in FIG. 11, when applied to a dance game machine, the operator goes up to, for example, the stage of the game machine in which the cameras 18 to 22 are installed. Then, after charging a predetermined fee, according to the menu displayed on the monitor 40, the degree of difficulty (easy, standard, har
d) and the song to dance etc. When the game starts, as shown in FIG. 12 (A), a CG character is displayed on the monitor 40 according to a reference motion stored in advance, and the operator imitates this and dances. Then, the operator performing the dance (continuous motion) moves to the camera 18
Taken at ~ 22. That is, the input data is acquired. As shown in FIG. 12B, the standard motion and the continuous motion of the operator may be displayed on the monitor 40 in two screens.

Further, when presenting a dance or inputting a continuous motion of an operator, a system may be added in which a specific rhythm or music selected is simultaneously presented and the motion is synchronized.

Then, the motion verification device 38 compares the teacher data and the input data for each segment, and calculates the score for each segment. In the comparison and evaluation process described above, only the score of each segment is calculated,
The score for each segment may be displayed using the monitor 40. Such an operation is repeatedly executed, and when the music ends, that is, when the comparison and evaluation processing of the entire sequence ends, the total score is calculated. The score is determined according to the total score and is displayed on a part of the monitor 40 as can be seen from FIGS. 12 (A) and 12 (B). As described above, the level (rank A, B, C, D and E) may be displayed in addition to the score.

When the system 10 is applied to the dance game machine as described above, the CG character is displayed by using the teacher data corresponding to the reference motion acquired in advance. When applied to a device or the like, the reference motion of the instructor may be acquired in real time and the teacher data may be presented.

According to this embodiment, the teacher data corresponding to the reference motion is divided into segments, and the teacher data is compared with the input data corresponding to the continuous image motion of the operator for each segment. Can be tested. Moreover, since different weights are assigned to the respective segments, a more appropriate total score can be calculated.

In this embodiment, the DP matching method is used to compare the reference motion and the continuous image motion.
The comparison may be performed using an M (Hidden Markov Model) method or a recurrent neural network.

[Brief description of drawings]

FIG. 1 is an illustrative view showing one embodiment of the present invention.

2 is an illustrative view showing a motion capture system shown in FIG. 1 embodiment. FIG.

FIG. 3 is an illustrative view showing a segment division method.

FIG. 4 is an illustrative view showing a process of a first segment dividing device shown in FIG. 1 embodiment.

FIG. 5 is a flowchart showing a part of the processing of the three-dimensional motion verification apparatus shown in FIG. 1 embodiment.

FIG. 6 is a flowchart showing another part of the processing of the three-dimensional motion inspection apparatus shown in FIG. 1 embodiment.

FIG. 7 is an illustrative view for explaining a process of the DP matching method.

FIG. 8 is a flowchart showing a part of other processing of the three-dimensional motion verification apparatus shown in FIG. 1 embodiment.

FIG. 9 is a flowchart showing another part of the other processing of the three-dimensional motion inspection apparatus shown in the embodiment in FIG. 1;

FIG. 10 is a flowchart showing another part of the other processing of the three-dimensional movement inspection device shown in the embodiment in FIG. 1;

FIG. 11 is an illustrative view showing a specific example of the embodiment in FIG. 1;

12 is an illustrative view showing a display screen of a monitor of the concrete example of FIG. 11. FIG.

[Explanation of symbols]

10 ... System 12, 14, 16, 18, 20, 22, ... Camera 24 ... 1st motion capture system 26 ... First segment dividing device 28 ... First motion data storage device 30 ... First image generating device 32 ... Second motion capture system 34 ... Second segment dividing device 36 ... Second image generating device 38. Motion verification device 40 ... Monitor 50, 52, 54 ... Image processing device 56 ... Three-dimensional coordinate calculation device

Front page continuation (72) Inventor Atsushi Otani 5 Seiji-cho, Seika-cho, Soraku-gun, Kyoto Prefecture Mihiratani No. 5 ATR Co., Ltd. Intelligent Video Communications Laboratory (56) Reference JP-A-9-245178 (JP , A) JP 10-334270 (JP, A) JP 7-302341 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T ⁷ /00-7/60 A63B 69/00 G06T 1/00 H04N 7/18 JISST file (JOIS)

Claims (4)

(57) [Claims] 1. A motion detecting means for detecting a continuous motion of a person in real time based on a plurality of images taken from a plurality of viewpoints, and comparing the continuous motion detected by the motion detecting means with a reference motion for each segment. In a three-dimensional continuous motion test apparatus , which comprises a comparison unit and a verification unit that tests the degree of approximation of the continuous motion corresponding to the reference motion based on the comparison result of the comparison unit, for the comparison for each segment, Get in real time
A dividing means for dividing the reference motion at an appropriate boundary is further provided.
The dividing means is a variable permission set every predetermined period.
The detection result with the highest evaluation value is obtained in the range
3 is characterized in that the time point is set as the boundary.
Dimensional continuous motion tester. 2. The inspection apparatus for three-dimensional continuous motion according to claim 1, further comprising display means for visually displaying an inspection result of said inspection means. 3. The comparison means is configured to perform the continuous operation and the reference.
The three-dimensional continuous motion test apparatus according to claim 1 or 2 , which compares the motion with a motion by a DP matching method . 4. The test means is configured to perform the test for each segment.
Unit test means for testing the degree of approximation, and the continuous motion
Includes comprehensive verification means for testing the above-mentioned approximations for all
A test apparatus for three-dimensional continuous motion according to claim 1 .