JP7231573B2 - VIDEO CONVERSION METHOD, APPARATUS AND PROGRAM - Google Patents

Description

The present invention relates to a video conversion method, apparatus, and program for analyzing a moving image, and in particular, classifies a moving image into an analysis section in which high-quality analysis is desirable and a non-analysis section other than the analysis section, and performs analysis according to the classification result. The present invention relates to a video conversion method, apparatus, and program for converting a bit rate by using

In sports such as baseball, rugby, and golf, where the movement of the skeleton determines progress, not only the movement of the whole body but also the position and movement of the skeleton are important. Therefore, it is often effective to provide specialized instruction for improvement. However, in order to go to a studio or a school and receive guidance from experts, a reasonable cost burden is imposed, and time and place restrictions increase.

Japanese Patent Application Laid-Open No. 2002-200001 discloses a recognition unit that recognizes motion of a person reflected in an input image, and a display control unit that superimposes different virtual objects on the input image according to the effectiveness of the recognized motion. An image processing apparatus has been proposed that calculates a score that indicates the effectiveness of a given exercise and superimposes the calculation result on an input image, thereby presenting feedback regarding the effectiveness of the exercise to the user in a visible form.

However, in Patent Document 1, the user can subjectively evaluate the training by checking the user's training video and comparing it with the movement of the superimposed virtual object. No specific instructions are available. Therefore, effective training cannot be realized unless the user himself/herself has specialized knowledge.

In order to solve such a technical problem, the inventors of the present invention select a training menu from a database storing a plurality of training menus, distribute the selected training menu to the user terminal, and is obtained from the user terminal, the user's skeleton information is extracted from the training video, the user's training is evaluated based on the skeleton information, and the training evaluation is distributed to the user terminal. Invented and filed a patent application (Patent Document 2).

In Patent Document 3, a video recording device mounted on a vehicle generates video data for recording based on a video signal from a camera that captures images of the surroundings of the vehicle while the vehicle is parked, while analyzing the video captured by the camera, A technique is disclosed for controlling the amount of video data stored in a storage unit by controlling the video quality of the video data based on the video change.

Patent Document 4 describes a case where an electronic blackboard connectable to an information processing apparatus receives an input from a user on a screen, acquires a video signal from the information processing apparatus to generate image data, and receives an input from the user. discloses a technique for changing the frame rate of image data to a value lower than that before accepting an input.

Japanese Patent Application Laid-Open No. 2013-103010 Patent application No. 2018-181304 Japanese Patent Application Laid-Open No. 2019-161350 Japanese Patent Application Laid-Open No. 2019-159261

Z. Cao, T. Simon, S. Wei and Y. Sheikh, "Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields," 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, 2017, pp 1302-1310.

According to Patent Literature 2, the user simply shoots and transmits a training video of himself/herself performing the distributed training menu with the user terminal, without going to a facility such as a training gym where a specialist resides, at home or the like. Easy access to professional training.

On the other hand, in Japanese Patent Laid-Open No. 2002-200010, training video must be captured by a user terminal and transmitted, but the amount of data for moving pictures is large, so the amount of traffic increases. Also, on the evaluation side, when videos are distributed from a large number of users at the same time, congestion occurs in the network and the load on the server increases.

According to Patent Literatures 3 and 4, the amount of moving image data can be reduced, so network congestion due to image transmission and an increase in server load can be reduced. However, if the amount of data is reduced, the image quality will be lowered, which may hinder the analysis.

It is an object of the present invention to solve the above technical problems and to provide a video conversion method, apparatus, and program capable of reducing the amount of data by lowering the bit rate so as not to hinder the analysis of moving pictures. be.

In order to achieve the above objects, the present invention is characterized in that it comprises the following configuration in a video conversion apparatus for converting the amount of data of a moving picture in which a person is a subject.

(1) extracting means for extracting skeletal information of a person from each frame, position estimating means for estimating the position of a predetermined evaluation site based on the skeletal information, and determination for determining an analysis section based on the position of the evaluation site. means, and conversion means for reducing the amount of data by converting the bit rate of the non-analysis interval to a bit rate lower than that of the analysis interval.

(2) The position estimating means further estimates the position of the predetermined reference part based on the skeletal information, and the determining means determines whether the relative positional relationship between the evaluation part and the reference part satisfies a predetermined start requirement. The image from the timing to the end timing that satisfies the predetermined end requirement is determined as the analysis section.

(3) The position estimating means estimates each position of a plurality of evaluation parts and a plurality of corresponding reference parts based on the skeleton information, and the determining means determines relative positions of each evaluation part and each corresponding reference part The image from the start timing when the relative positional relationship satisfies the predetermined start requirements to the end timing when the relative positional relationship between each evaluation site and each corresponding reference site satisfies the predetermined end requirements is determined as the analysis section. I made it

(4) The position estimating means estimates respective positions of a plurality of evaluation parts and a plurality of corresponding reference parts based on skeleton information, and the determining means determines the position of one evaluation part and one corresponding reference part. From the start timing when the relative positional relationship satisfies the predetermined start requirements to the end timing when the relative positional relationship between the other one evaluation part and the other corresponding reference part satisfies the predetermined end requirements video is determined as the analysis section.

(5) The determining means determines, as the analysis section, the image from the start timing at which the movement trajectory of the evaluation site satisfies the predetermined start requirements to the end timing at which the predetermined end requirements are satisfied.

(6) The determining means determines whether one of the relative positional relationship between the evaluation site and the reference site and the movement trajectory of the evaluation site satisfies a predetermined start requirement from a start timing when the other satisfies a predetermined end requirement. The image up to the timing is determined as the analysis section.

(7) There is provided means for receiving designation of an evaluation site, and the position estimation means estimates the position of the designated evaluation site.

(8) Means for determining the type of motion of the person based on the moving image; and means for determining an evaluation site based on the determination result of the motion; position is estimated.

(9) The determining means considers, as the start timing of the analysis interval, a timing that precedes the analysis interval determined based on the position of the evaluation site by a predetermined time.

According to the present invention, the following effects are achieved.

(1) In a video conversion device for converting the amount of data of a moving image of a person as a subject, a position estimating means for estimating the position of a predetermined evaluation site based on skeletal information extracted from each frame; and a conversion means for reducing the amount of data by converting the bit rate of the non-analysis section to a bit rate lower than that of the analysis section. The analysis interval in which the analysis is desired can be precisely determined. Therefore, it is possible to reduce the amount of data and the amount of traffic of the moving image without interfering with the analysis of the moving image.

(2) The position estimating means further estimates the position of the predetermined reference part based on the skeletal information, and the determining means determines the start timing when the relative positional relationship between the evaluation part and the reference part satisfies a predetermined start requirement. to an end timing that satisfies a predetermined end requirement is determined as an analysis section. Therefore, it becomes possible to accurately determine the analysis section when the analysis section depends on the relative positional relationship between the evaluation site and the reference site.

(3) The position estimating means estimates each position of the plurality of evaluation parts and the corresponding plurality of reference parts based on the skeletal information, and the determining means determines the relative positions of each evaluation part and each corresponding reference part. A video from the start timing at which the relationship satisfies a predetermined start requirement to the end timing at which a predetermined end requirement is satisfied is determined as an analysis section. Therefore, even when the analysis interval depends on the relative positional relationship between a plurality of evaluation regions and their corresponding reference regions, it becomes possible to accurately determine the analysis region.

(4) The position estimating means estimates each position of the plurality of evaluation parts and the corresponding plurality of reference parts based on the skeleton information, and the determining means calculates the relative positions of the one evaluation part and the corresponding one reference part. The image from the start timing when the positional relationship satisfies the predetermined start requirements to the end timing when the relative positional relationship between the other one evaluation site and the other corresponding reference site satisfies the predetermined end requirements. is determined as the analysis interval. Therefore, even when the start timing and end timing of the analysis interval depend on the relative positional relationship between a plurality of different evaluation regions and their corresponding reference regions, the analysis interval can be accurately determined.

(5) Since the determination means determines the image from the start timing at which the movement trajectory of the evaluation site satisfies the predetermined start requirements to the end timing at which the predetermined end requirements are satisfied as the analysis section, the start of the analysis section is determined. It becomes possible to accurately determine the analysis interval when the timing and end timing depend on the movement trajectory of the evaluation site.

(6) The determination means determines the relative positional relationship between the evaluation site and the reference site and the movement trajectory of the evaluation site from the start timing at which one satisfies a predetermined start requirement to the end timing at which the other satisfies a predetermined end requirement. Even if one of the start timing and end timing of the analysis interval depends on the relative positional relationship between the evaluation site and the reference site, and the other depends on the movement trajectory of the evaluation site. It becomes possible to accurately determine the analysis interval.

(7) Since means for receiving the designation of the evaluation site is provided and the position of the designated evaluation site is estimated, the evaluation site can be set at the user's will.

(8) Means for determining the type of motion of a person based on a moving image; Since the estimation is performed, the optimum evaluation site can be automatically set without forcing the user to have specialized knowledge or burden, and the optimum analysis interval can be determined.

(9) Since the analysis interval determined based on the position of the evaluation site is considered to start at a predetermined time before the analysis interval, the analysis interval start timing, which is important for video analysis, is not extracted. can be prevented.

1 is a block diagram showing the configuration of a video analysis system to which the present invention is applied; FIG. 1 is a functional block diagram showing the configuration of a local terminal according to a first embodiment; FIG. FIG. 4 is a diagram showing a skeleton to be extracted; FIG. 10 is a diagram (part 1) showing a method of determining an analysis interval according to the first embodiment; FIG. 10 is a diagram (part 2) showing a method of determining an analysis interval according to the first embodiment; FIG. 8 is a functional block diagram showing the configuration of a local terminal according to a second embodiment; FIG. 11 is a functional block diagram showing the configuration of a local terminal according to a third embodiment; FIG. 11 is a diagram showing a method of determining analysis intervals according to the third embodiment;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a video analysis system to which the present invention is applied. A local terminal 1 equipped with a video conversion function according to the present invention in addition to a camera and communication functions, and a The main configuration is a video analysis server 2 that acquires and analyzes the captured camera video via, for example, Wi-Fi, wireless base station BS, and network NW. However, if the processing capability of the local terminal 1 is sufficiently high, the video analysis system can also be configured with only the local terminal 1 .

The local terminal 1 analyzes the camera image of the user exercising, dancing, training, etc. at the skeleton level, and classifies the user's movement into analysis intervals in which detailed analysis is to be performed and other non-analysis intervals. The local terminal 1 further converts the camera video so that the bit rate of the non-analysis interval is lower than the bit rate of the analysis interval, accumulates the converted camera video, and further distributes it to the video analysis server 2 .

In this embodiment, as a result of reducing the bit rate of the non-analysis section, the amount of data of the camera image after conversion is less than the amount of data of the camera image before conversion. In addition, the amount of traffic related to video distribution from the local terminal 1 to the video analysis server 2 can be reduced.

In the video analysis server 2, an expert or the like confirms the received converted camera video on the monitor. At this time, since the video in the analysis section does not have a reduced bit rate, it maintains sufficient video quality. Therefore, an expert or the like can analyze the movement of the user in detail while referring to the high-definition video and provide accurate advice and coaching to the user.

FIG. 2 is a functional block diagram of the local terminal 1 according to the first embodiment. The section 106 is the main configuration.

Such a local terminal 1 can be configured by installing an application (program) that implements each function described later in a general-purpose computer or mobile terminal. Alternatively, a part of the application can be configured as a dedicated machine or a single-function machine that is hardware or programmed. In this embodiment, a case where a smart phone or a tablet terminal is substituted for the local terminal 1 will be described as an example.

The video acquisition unit 101 acquires a user's camera video (moving image) captured by the camera function of the local terminal 1 in units of frames. A person area extraction unit 102 extracts a person area from each frame image of a camera video. An SSD (Single Shot Multibox Detector), for example, can be used to extract the person area.

The skeleton information extraction unit 103 extracts a skeleton registered in advance as an extraction target from the person region of the frame image, and registers the position information and the connection state with other skeletons as skeleton information. Existing skeleton extraction technology (Cascaded Pyramid Network) can be used to extract skeleton information.

FIG. 3 is a diagram showing a skeleton to be extracted by the skeleton information extraction unit 103, and includes left and right elbow joints P3 and P6, left and right wrist joints P4 and P7, left and right knee joints P9 and P12, and left and right ankles. Joints P10 and P13 and bones connecting these joints are extracted.

Note that the skeleton extraction method is not limited to the method targeting the human region extracted in advance as described above. For example, as disclosed in Non-Patent Document 1, a Confidence Map that encodes the positions of body parts and Part Affinity Fields (PAFs) that encode connectivity between body parts are added to a feature map extracted from a frame image. A skeleton model is constructed by sequentially applying the two sequential prediction processes used to estimate the positions and connectivity of body parts of human objects (users) extracted from frame images in a single inference using a bottom-up approach. You may

At this time, by implementing a process that excludes the connectivity of body parts extracted from different partial regions from the target of estimation, the position and connectivity of body parts are estimated for each partial region, that is, for each user, and the skeleton model of the object is estimated. become able to.

The position estimating unit 104 receives the evaluation site designated by the user at the evaluation site designation receiving unit 104a, and estimates the position P of the skeleton corresponding to the designated evaluation site.

The analysis interval determination unit 105 calculates the relative positional relationship between the position P of the evaluation site estimated by the position estimation unit 104 and a predetermined reference site, and analyzes the timing when this positional relationship satisfies a predetermined start requirement. The start point of the section is registered, and the timing at which a predetermined end requirement is satisfied is registered as the end point of the analysis section. The reference site can be set at any skeletal position other than the evaluation site.

FIG. 4 is a diagram showing a method of determining an analysis interval for a user's golf swing video. Since the movement of the left hand is important in golf swing evaluation, the left wrist joint P7 is designated as the evaluation site, and the head H, for example, is set as the reference site. If the head cannot be directly extracted as skeleton information as shown in FIG. 3, the position of the head H can be represented by the position of the nose skeleton P0 or the central position of both eyes P14 and P15.

When evaluating a golf swing, it is desirable that the analysis interval is from the backswing to the downswing and then to the follow swing. In general, the left wrist joint P7 is positioned on the right side of the head H at the timing when the swing of a right-handed user switches from the backswing to the downswing, and the left wrist joint P7 is located on the right side of the head H at the timing of the end of the follow swing. It is empirically recognized to be located in the lower left.

Therefore, in the present embodiment, the left wrist joint P7 is moved to the right side of the head H so that the analysis interval is determined from the time t1 when the swing switches from the backswing to the downswing to the time t2 when the follow swing ends. After that, the positional relationship in which the left wrist joint P7 has moved to the lower left position of the head H is defined as the end requirement of the analysis section. As a result, the video section from the backswing to the downswing to the follow swing can be used as the analysis section.

According to this embodiment, it is possible to accurately determine an analysis section when the analysis section depends on the relative positional relationship between the evaluation site and the reference site.

In addition, it is desirable that the start timing of the extraction of the analysis section is not delayed. Therefore, as shown in FIG. 5, when the start timing t1 is detected based on the relative positional relationship between the evaluation site and the reference site, the analysis interval is set to go back from the timing t1 by a predetermined time α. Also good.

By doing so, it is possible to prevent omission of extraction of the start timing of the analysis section, which is important for video analysis.

The video conversion unit 106 reduces the bit rate of the camera video so that the bit rate of the non-analysis interval is relatively lower than the bit rate of the analysis interval. At this time, the video conversion unit 106 can selectively lower only the bit rate of the non-analysis section while maintaining the bit rate of the analysis section at the bit rate at the time of shooting. Alternatively, both the bit rates of the analysis interval and the non-analysis interval may be reduced, and at that time, the bit rate of the non-analysis interval may be made relatively lower than the bit rate of the analysis interval.

When converting the bit rate, the video conversion unit 106 can reduce the frame rate of the non-analysis section to be lower than the frame rate of the analysis section. For example, if the camera video is captured at a frame rate of 30 fps, the frame rate of the analysis section may be maintained at 30 fps, and only the frame rate of the non-analysis section may be reduced to, for example, 10 fps.

Alternatively, the rate of frame rate reduction may be varied such that the frame rate in the analysis section is reduced to 20 fps and the frame rate in the non-analysis section is reduced to 10 fps.

Furthermore, if the camera video is a color image, the bit rate may be converted or lowered by converting only the non-analysis section to monochrome video, or the resolution of the non-analysis section may be set to a lower resolution than the analysis section. You can lower the bitrate by converting.

According to the present embodiment, it is possible not only to accurately determine an analysis section for which analysis using high-quality video is desirable, but also to request analysis using high-quality video for the bit rate of non-analysis sections for which analysis using high-quality video is not requested. Since the bit rate is lower than the bit rate of the analysis interval that is used, it is possible to reduce the transfer data amount and storage capacity of the camera video without affecting the accuracy and accuracy of video analysis.

FIG. 6 is a functional block diagram of the second embodiment of the local terminal 1. Since the same reference numerals as in FIG. 2 denote the same or equivalent parts, the description thereof will be omitted. This embodiment is characterized in that the position estimation unit 104 includes a motion type determination unit 104b and an evaluation part determination unit 104c, and the evaluation part is automatically set based on the camera image.

The motion type determination unit 104b determines the motion type of the user by applying the camera image to an estimation model that has been learned in advance. For example, when the camera image of the golf swing described with reference to FIGS. 4 and 5 is input, the motion type of the user is determined as "golf swing".

Based on the determination result of the motion type of the user, the evaluation region determination unit 104c registers in advance an evaluation region, a reference region, and a start requirement and an end requirement of the analysis region, which serve as indicators for determining the analysis region for the motion type. It is determined based on the correspondence relationship with the type of motion that is present.

For example, if the motion type of the user is determined to be "golf swing", the evaluation site is "left wrist joint", the reference site is "head", the start requirement is "left wrist joint is on the right side of the head", and the end requirement is is determined as "the left wrist joint is on the lower left side of the head".

According to the present embodiment, the evaluation site, the start requirement, and the end requirement, which serve as determination indicators for the analysis section, are automatically determined based on the camera image. It becomes possible to automatically determine the motion type and the optimum analysis interval.

FIG. 7 is a functional block diagram showing the configuration of the main parts of the local terminal 1 having a video conversion function according to the third embodiment of the present invention. Therefore, the description is omitted. This embodiment is characterized in that the analysis section determination unit 105 includes a trajectory registration unit 105a in order to determine the analysis section based on the trajectory of the movement of the evaluation site.

FIG. 8 is a diagram for explaining the method of determining the analysis interval in the third embodiment. Here, the position of the left wrist joint P7 during a golf swing is taken as an example, and the analysis interval is determined based on the trajectory TR. explain how to

In a scene such as a golf swing that requires regular and uniform movements of hands and feet, whether or not each joint moves according to a predetermined rule can be an analysis index. Therefore, in the present embodiment, the trajectory TR of the left wrist joint P7 estimated by the position estimation unit 104 is registered in the trajectory registration unit 105a, and the position of the left wrist joint P7 on the trajectory ends from the start timing that satisfies a predetermined start requirement. Determine the analysis interval up to the end timing that satisfies the requirements.

In this embodiment, the position of the left wrist joint P7 makes a U-turn when switching from the backswing to the downswing in order to determine the interval from the backswing to the downswing to the end of the follow swing as the analysis interval. Focusing on this, the fact that the trajectory TR of the left wrist joint P7 has made a U-turn is registered as a start requirement.

Furthermore, when switching from the downswing to the follow swing, the trajectory TR descends along the downward arc and turns to rise after impact. Elapsed time is registered as a termination requirement. As a result, the analysis interval can be from the time t1 when the backswing switches to the downswing to the time t2 before the follow swing.

According to this embodiment, when the start timing and end timing of an analysis section depend on the movement trajectory of the evaluation region, the analysis section can be determined accurately.

Note that in the first and third embodiments described above, one evaluation site is focused on, and based on the relative positional relationship with the corresponding reference site (first embodiment) or movement trajectory (third embodiment) It was described as evaluating the user's movement. However, the present invention is not limited only to this, and a combination of the first and third embodiments satisfies the starting requirements corresponding to the relative positional relationship between the target evaluation site and the reference site, and The start timing may be determined when the movement trajectory of the evaluation site satisfies the corresponding start requirement. The same applies to the end timing.

Furthermore, in the above-described first and third embodiments, only one evaluation site is focused, but the present invention is not limited to this. The start timing may be determined when predetermined start requirements regarding the relative position and movement trajectory are satisfied simultaneously or in a predetermined order.

For example, in the golf swing of the first embodiment, one of the left and right elbow joints P3 and P6 is defined as an evaluation site and the other is defined as a reference site, and the left wrist joint P7 is positioned on the right side of the head H, and at the same time or Back and forth, it may be defined as a start requirement that the distance between the left and right elbow joints P3 and P6 is less than the reference distance. The same applies to the end timing.

In the third embodiment, the start timing is when it is recognized that the trajectory TR of the left wrist joint P7 has made a U-turn, and at the same time or before or after this, the movement of the right knee joint in the front right direction is recognized. You may make it judge. The same applies to the end timing.

In this way, when the analysis interval depends on the relative positional relationship between a plurality of evaluation parts and their reference parts, when the analysis interval depends on the movement trajectories of the plurality of evaluation parts, or when the analysis interval is a single Even when it depends on both the relative positional relationship between the evaluation site and its reference site and the movement trajectory of the one or other evaluation site, the analysis section can be accurately determined.

Furthermore, in each of the above-described embodiments, it has been described that whether or not the start requirements and end requirements are satisfied is determined by focusing on the same single evaluation site, but the present invention is not limited to this. , the judgment may be made by paying attention to different evaluation sites.

For example, in the golf swing of the first embodiment, one of the left and right elbow joints P3 and P6 is defined as an evaluation site and the other is defined as a reference site. After that, the movement of the left wrist joint P7 to the lower left position of the head H may be set as the termination requirement.

In the third embodiment, it is assumed that the movement of the right knee joint in the front right direction is recognized as a starting condition, and then the trajectory of the left wrist joint P7 changes from falling to rising along a downward arc, and further The end requirement may be that a predetermined time has passed.

Furthermore, by combining the first and third embodiments, when one of the relative positional relationship between the evaluation site and the reference site and the movement trajectory of the evaluation site satisfies a predetermined start requirement, the start timing is determined, The termination timing may be determined when the other satisfies a predetermined termination requirement.

In this way, even when the start timing and end timing of an analysis interval depend on the positions and trajectories of a plurality of different evaluation regions, the analysis interval can be determined accurately.

Furthermore, the camera image to be converted into data is not limited to the moving image output from the camera in real time, but may be the moving image captured by the camera in advance and temporarily stored in memory disassembly or the like. Also good.

Reference Signs List 1 Local terminal 2 Image analysis server 101 Image acquisition unit 102 Person region extraction unit 103 Skeleton information extraction unit 104 Position estimation unit 104a Evaluation part designation reception unit 104b Movement type discrimination Part 104c...Evaluation part determination part 105...Analysis section determination part 105a...Trajectory registration part 106...Video conversion part

Claims (14)

In a video conversion device that converts the amount of data of a moving image with a person as a subject,
extracting means for extracting human skeleton information from each frame;
position estimating means for estimating the position of a predetermined evaluation site based on skeletal information;
a determining means for determining an analysis interval based on the position of the evaluation site;
and conversion means for reducing the amount of data by converting the bit rate of the non-analysis interval to a bit rate lower than that of the analysis interval. The position estimating means further estimates the position of the predetermined reference part based on the skeletal information,
The determination means determines, as the analysis section, a video from a start timing at which a relative positional relationship between the evaluation site and the reference site satisfies a predetermined start requirement to an end timing at which a predetermined end requirement is satisfied. 2. The video conversion device according to claim 1. the position estimating means estimates the positions of the plurality of evaluation parts and the corresponding plurality of reference parts based on the skeleton information;
The determining means determines the relative positional relationship between each evaluation site and each corresponding reference site from a start timing at which the relative positional relationship between each evaluation site and each corresponding reference site satisfies a predetermined start requirement. 2. The video converter according to claim 1, wherein the video up to the end timing that satisfies a predetermined end requirement is determined as the analysis section. the position estimating means estimates the positions of the plurality of evaluation parts and the corresponding plurality of reference parts based on the skeleton information;
The determining means selects another reference portion corresponding to another evaluation portion from a start timing at which a relative positional relationship between one evaluation portion and one corresponding reference portion satisfies a predetermined start requirement. 2. The image conversion apparatus according to claim 1, wherein the image up to an end timing satisfying a predetermined end requirement is determined as the analysis section. 2. The determining means determines, as the analysis section, the image from the start timing at which the movement trajectory of the evaluation site satisfies a predetermined start requirement to the end timing at which the predetermined end requirement is satisfied, as the analysis section. Video converter. The position estimating means further estimates the position of the predetermined reference part based on the skeletal information,
The determination means determines the relative positional relationship between the evaluation site and the reference site and the movement trajectory of the evaluation site from a start timing when one satisfies a predetermined start requirement to an end timing when the other satisfies a predetermined end requirement. 2. The image conversion apparatus according to claim 1, wherein the image is determined as an analysis interval. 7. The video conversion apparatus according to claim 1, wherein said conversion means lowers the bit rate of the non-analysis section while maintaining the bit rate of the analysis section. 2. The converting means reduces the bit rates of the analysis section and the non-analysis section such that the bit rate of the non-analysis section is lower than the bit rate of the analysis section. 7. The video conversion device according to any one of 6. 9. The video conversion apparatus according to claim 1, wherein said conversion means converts the frame rate of the non-analysis section into a frame rate lower than the frame rate of the analysis section. Equipped with means for accepting designation of an evaluation site,
10. The image conversion apparatus according to any one of claims 1 to 9, wherein said position estimating means estimates the position of the designated evaluation site. means for determining the type of movement of a person based on the moving image;
and means for determining an evaluation part based on the determination result of the movement type,
11. The image conversion apparatus according to any one of claims 1 to 10, wherein said position estimating means estimates the position of the determined evaluation site. 12. The determining unit according to any one of claims 1 to 11, wherein said determining means regards a timing a predetermined time before the analysis interval determined based on the position of the evaluation site as the start timing of the analysis interval. Video converter. In a video conversion method in which a computer converts the amount of data of a moving video image in which a person is a subject,
extract the skeleton information of the person from each frame,
estimating the position of a predetermined evaluation site based on skeletal information;
determining an analysis interval based on the position of the evaluation site;
A video conversion method, comprising converting a bit rate of a non-analysis interval into a bit rate lower than that of an analysis interval. In a video conversion program that converts the data volume of a moving picture with a person as a subject,
a procedure for extracting skeletal information of a person from each frame;
a procedure for estimating the position of a predetermined evaluation site based on skeletal information;
a procedure for determining an analysis interval based on the position of the evaluation site;
A video conversion program for causing a computer to execute a procedure for converting a bit rate of a non-analysis section into a bit rate lower than that of an analysis section.

Images