Human body key point analysis and recognition method and system based on video stream

Abstract

The present invention discloses a method and system for analyzing and identifying key points of a human body based on a video stream, and relates to the technical field of image data processing. A target video is obtained based on a video stream, and the corresponding target video is layered to obtain multiple video layers; a human body contour is extracted from the multiple video layers, and multiple video layers involving the human body contour are integrated to obtain a human body video; a human body in the human body video is identified and a human body motion model is constructed, and key points of the human body are marked based on the human body motion model to obtain a key point motion model; the key point motion model is overlapped with the human body contour to obtain a human body key point motion video. The present invention can improve the integrity of obtaining a human body video in a video, has a good human body key point analysis function, can obtain the human body contour in combination with all video information, and simultaneously trace and select the human body contour on the video timeline, and can extract the human body contour under a complex background.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a method and system for analyzing and identifying key points of a human body based on video streams. Background Art

[0002] With the development of society and the improvement of computer hardware performance, the human key point analysis and recognition method based on video stream will develop towards higher accuracy, faster speed and stronger robustness;

[0003] In the comparative patent CN113780253A, a method and system for identifying key points of human joint motion are described as follows: collecting video data of the subject's activities; performing target detection on the video data, obtaining each image frame containing a sub-region where the target is the subject as a key frame set; marking the active components in the sub-region where the target is the subject in each image frame in the key frame set, forming an active component set corresponding to the current image frame by the active components in each image frame, and extracting motion key points from the active component set; determining the coordinates of the human key points according to the extracted motion key points, thereby improving the recognition accuracy of the key points of human joint motion, realizing the intelligent detection of the key points of human joint motion, making the detection simpler and faster, greatly reducing the computational overhead and complexity, and improving the robustness of key point recognition;

[0004] However, in the current process of analyzing and identifying key points of the human body in a video, it is difficult to comprehensively extract the human body contour in the entire video under a complex background environment, and thus the analysis and identification of key points of the human body in the entire video cannot be guaranteed. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for analyzing and identifying key points of a human body based on a video stream, so as to solve the shortcomings in the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for analyzing and identifying key points of a human body based on a video stream, comprising the following steps:

[0007] Obtain a target video based on a video stream, and layer the target video to obtain multiple video layers;

[0008] Extracting a human body contour from multiple video layers, and integrating multiple video layers involving the human body contour to obtain a human body video;

[0009] Identify the human body in the human body video and build a human body motion model, and mark the key points of the human body based on the human body motion model to obtain a key point motion model;

[0010] The key point motion model is overlapped with the human body contour to obtain the human body key point motion video.

[0011] In a preferred embodiment, the step of acquiring a target video based on a video stream and layering the target video to obtain multiple video layers includes:

[0012] Get the target video through the video stream;

[0013] Construct multiple blank framework videos, and layer the target video based on the motion features to obtain multiple layered images;

[0014] Multiple layered images are stored one-to-one in correspondence with the blank frame video to obtain multiple video layers.

[0015] In a preferred embodiment, the step of constructing a plurality of blank framework videos and layering the target video based on the movement characteristics to obtain a plurality of layered images includes:

[0016] Determine a plurality of storage spaces, and set a video playback framework as a blank framework video based on the plurality of storage spaces;

[0017] Extracting and depicting the contours of the object in the target video to obtain multiple object contours, and laying multiple moving tracking points on the multiple object contours;

[0018] Obtain the spatial relationship between multiple objects in the target video, spatially arrange the mobile tracking points based on the spatial relationship, and connect adjacent mobile tracking points to obtain a tracking network;

[0019] Obtaining movement features of a moving tracking point in a tracking network in a target video, wherein the movement features include a moving direction and a moving distance of the moving tracking point;

[0020] The object contours corresponding to the continuous moving tracking points with the same moving direction are extracted and divided to obtain multiple layered images.

[0021] In a preferred embodiment, the step of extracting a human body contour from a plurality of video layers and integrating the plurality of video layers related to the human body contour to obtain a human body video comprises:

[0022] The completeness index of the human body contour in multiple video layers is determined based on the human body features, wherein the calculation formula of the completeness index is:

[0023] ,in, is the complete index, is the number of moving tracking points on the human body contour that conforms to human characteristics, is the number of overlapping moving tracking points on the human body contour that conforms to human characteristics, is the weight coefficient that conforms to the characteristics of the human body, and is a constant greater than zero;

[0024] Based on the complete index, multiple video layers with human body contours are determined as video layers to be integrated, and the video layers to be integrated are integrated according to fixed-point sections to obtain human body videos.

[0025] In a preferred embodiment, the step of using multiple video layers in which human body contours are determined based on the complete index as video layers to be integrated, and integrating the video layers to be integrated according to fixed-point sections to obtain human body videos includes:

[0026] Set a blank collection video;

[0027] Position multiple video layers according to the playback time, set fixed-point sections between the multiple video layers, and connect the fixed-point sections with the blank set video;

[0028] Obtain the human body contour with the highest completeness index in multiple video layers to select the video line, and obtain the video layer to be integrated;

[0029] The human body contour in the video layer to be integrated is pasted to the blank set video through a fixed-point section to obtain a human body video.

[0030] In a preferred embodiment, the steps of identifying a human body in a human body video and constructing a human body motion model, and marking key points of the human body based on the human body motion model to obtain a key point motion model include:

[0031] Construct a virtual human model from human videos;

[0032] The human body virtual model follows the human body video, and key points are marked according to the motion bends of the human body virtual model to obtain a key point motion model.

[0033] In a preferred embodiment, the step of overlapping the key point motion model with the human body contour to obtain a human body key point motion video comprises:

[0034] Overlap the key point motion model with the human body contour;

[0035] The human body virtual model of the overlapped key point motion model is hidden to obtain the human body key point motion video.

[0036] The present invention also provides a human body key point analysis and recognition system based on video stream, comprising:

[0037] A layering module is used to obtain a target video based on a video stream, and to layer the target video to obtain multiple video layers;

[0038] An extraction module, connected to the layering module, is used to extract the human body contour from the multiple video layers, and integrate the multiple video layers involving the human body contour to obtain the human body video;

[0039] A construction module, connected to the extraction module, is used to identify the human body in the human body video and construct a human body motion model, and mark the key points of the human body based on the human body motion model to obtain a key point motion model;

[0040] The overlap module is connected to the construction module and is used to overlap the key point motion model with the human body contour to obtain the human body key point motion video.

[0041] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0042] The present invention obtains a human body video by pasting the human body contour in the video layer to be integrated into a blank set video through a fixed-point section. The human body video here is the overall human body contour of all timelines in the target video, and then the human body video corresponding to the human body contour can be obtained, which can improve the integrity of the human body video obtained in the video, has a good human body key point analysis function, can obtain the human body contour in combination with all video information, and trace and select the human body contour on the video timeline at the same time, and can extract the human body contour under complex backgrounds. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0044] Figure 1 The figure is a flow chart of the method of the present invention.

[0045] Figure 2 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Example 1, please refer to Figure 1 As shown, the human body key point analysis and recognition method based on video stream described in this embodiment includes the following steps:

[0048] S1. Obtain a target video based on a video stream, and layer the target video to obtain multiple video layers;

[0049] S2, extracting a human body contour from multiple video layers, and integrating multiple video layers involving the human body contour to obtain a human body video;

[0050] S3, identifying the human body in the human body video and constructing a human body motion model, marking the key points of the human body based on the human body motion model to obtain a key point motion model;

[0051] S4, overlapping the key point motion model with the human body contour to obtain a human body key point motion video;

[0052] As described in the above steps S1-S4, with the development of society and the improvement of computer hardware performance, the human key point analysis and recognition method based on video stream will develop towards higher accuracy, faster speed and stronger robustness;

[0053] At present, in the process of analyzing and identifying the key points of the human body in the video, it is difficult to comprehensively extract the human body contour in the entire video under a complex background environment, and thus the analysis and identification of the key points of the human body in the entire video cannot be guaranteed. In the present application, the human body contour in the video layer to be integrated is pasted to the blank set video through a fixed-point section to obtain a human body video. The human body video here is the overall human body contour of all timelines in the target video, and then the human body video corresponding to the human body contour can be obtained, which can improve the integrity of the human body video obtained in the video, has a good human body key point analysis function, can combine all video information to obtain the human body contour, and at the same time trace and select the human body contour on the video timeline, and can extract the human body contour under a complex background.

[0054] In one embodiment, the step S1 of acquiring a target video based on a video stream and layering the target video to obtain multiple video layers includes:

[0055] S11, obtaining a target video through a video stream;

[0056] S12, constructing multiple blank framework videos, and layering the target video based on the movement characteristics to obtain multiple layered images;

[0057] S13, synchronously storing the multiple layered images in one-to-one correspondence with the blank frame video to obtain multiple video layers;

[0058] As described in the above steps S11-S13, video streaming is a transmission method of video data information, which involves dividing the video content into multiple small data packets and transmitting them to users in real time through the network. The video content is divided into multiple small data packets, and these data packets are transmitted to users in real time through the network; real-time playback: users do not need to wait for the entire video file to be downloaded, and can start playing the video after receiving part of the data packet; streaming transmission: this is a special transmission method, which allows users to play the video while receiving data, rather than downloading the entire file first and then playing it, so it can improve the efficiency of video acquisition, and then construct multiple blank structure videos. The blank structure video is a video storage with blank content. The target video is layered based on the movement characteristics to obtain multiple layered pictures. Here, the objects in the target video can be divided, and then the divided layered pictures are one-to-one matched on the blank structure video, and the multiple layered pictures are stored in the blank structure video synchronously according to time to obtain multiple video layers;

[0059] In one embodiment, the step S12 of constructing a plurality of blank framework videos and layering the target video based on the movement characteristics to obtain a plurality of layered images includes:

[0060] S121, determining a plurality of storage spaces, and setting a video playback framework as a blank framework video based on the plurality of storage spaces;

[0061] S122, extracting and depicting the object contours in the target video to obtain multiple object contours, and laying multiple moving tracking points on the multiple object contours;

[0062] S123, obtaining the spatial relationship between multiple objects in the target video, spatially arranging the mobile tracking points based on the spatial relationship, and connecting adjacent mobile tracking points to obtain a tracking network;

[0063] S124, obtaining movement characteristics of a mobile tracking point in a tracking network in a target video, wherein the movement characteristics include a movement direction and a movement distance of the mobile tracking point;

[0064] S125, extracting and dividing the object contours corresponding to the continuous moving tracking points with the same moving direction to obtain a plurality of layered images;

[0065] As described in the above steps S121-S125, multiple storage spaces are determined for storing subsequent layered images to obtain the required video layers, and a video architecture is set in the storage space. The video architecture here is a video playback storage in a blank state, which is used as a blank architecture video here. In other words, multiple storage media are determined to store the subsequently generated layered image frames to construct the required video layers. In these storage media, a video frame structure is configured. The structure is an empty video playback container in the initial state, which can be called a "blank video frame". Then, the object contours in the target video are extracted. The extraction method here can be based on color and spatial continuity. After the object contours are extracted, the spatial relationship between multiple objects in the target video is obtained. For example, an object is in front of or behind another object. After determining the spatial relationship, the mobile tracking points are spatially arranged according to the space where the corresponding object contours are located. The mobile tracking points can be distributed front and back according to the spatial relationship. Then, adjacent mobile tracking points are connected to obtain a tracking network. Then, the moving direction and moving distance of the mobile tracking points are obtained. If it is an object, it will basically move in the same direction during the movement. The moving distance here is uncertain but needs to be collected through mobile tracking points. The purpose here is to obtain the movement of the human body in the video. Therefore, when the human body moves, although the individual movements of the overall limbs are not synchronized, if the human body is viewed as a point, the entire human body has a tendency to move in the same direction. Here, it moves in the general direction of the human body. Therefore, multiple object contours in the target video can be extracted based on the movement features. The extraction here may not be comprehensive, but the partial or full contours of the object can be obtained. Then, the target video will be divided into multiple layered pictures according to the object contours divided by the movement features, which has a better object extraction effect.

[0066] In one embodiment, the step S2 of extracting a human body contour from a plurality of video layers and integrating the plurality of video layers related to the human body contour to obtain a human body video includes:

[0067] S21, determining the completeness index of the human body contour in multiple video layers based on the human body features, wherein the calculation formula of the completeness index is:

[0068] ,in, is the complete index, is the number of moving tracking points on the human body contour that conforms to human characteristics, is the number of overlapping moving tracking points on the human body contour that conforms to human characteristics, is the weight coefficient that conforms to the characteristics of the human body, and is a constant greater than zero. It should be noted that The larger the value of The smaller the value of The larger the value, the higher the integrity of the human body contour;

[0069] S22, determining multiple video layers with human body contours as video layers to be integrated based on the complete index, and integrating the video layers to be integrated according to fixed-point sections to obtain human body videos;

[0070] In one embodiment, the step S22 of using multiple video layers in which human body contours are determined based on the complete index as video layers to be integrated, and integrating the video layers to be integrated according to fixed-point sections to obtain human body videos includes:

[0071] S221, setting a blank set video;

[0072] S222, positioning the multiple video layers according to the playback time, setting a fixed-point section between the multiple video layers, and performing data connection between the fixed-point section and the blank set video;

[0073] S223, obtaining the human body contour with the highest integrity index in multiple video layers to select video lines, and obtaining a video layer to be integrated;

[0074] S224, pasting the human body contour in the video layer to be integrated to the blank set video through a fixed-point section to obtain a human body video;

[0075] As described in the above steps S21-S22, the human body feature here is the shape of the human body. Since multiple video layers are divided by movement features, the human body can be distinguished from the background during the movement of the human body. The video layer with the human body contour can be obtained through the human body features (the shape of the human body trunk and limbs and the connection features between the human body trunk and limbs. Here, the human body contour can be obtained by using a learning model after training the human body model). Then, the integrity of the human body contour in the video layer is calculated, which is reflected here by the completeness index. The number of moving tracking points on the human body contour that conforms to the human body characteristics and the number of human body wheels that conform to the human body characteristics are collected. The number of overlapping moving tracking points on the contour and the weight coefficient that conforms to the human body characteristics can obtain a relatively complete human body contour in the video layer. The weight coefficient that conforms to the human body characteristics here refers to the human body contour with more human body characteristics in the video layer, for example, the head, limbs and torso all have them. The more human body features it has, the higher the weight value is. Then a blank collection video is set. The blank collection video is also a player with a video length and a space for storing videos. Multiple video layers are positioned according to the playback time. A fixed-point section is set between multiple video layers. The fixed-point section can be used to determine the playback time of multiple video layers, and it can also It is possible to connect multiple video layers with blank set videos respectively. The positions where the fixed-point sections are connected to the multiple video layers are ports. There is a data connection relationship between the multiple ports and the storage devices corresponding to the blank set videos. The multiple video layers are storage devices for storing multiple divided videos. Then, the human body contour with the highest completeness index in the multiple video layers is selected, and the missing human body parts of the selected human body contour are recommended in the multiple video layers. The recommendation here is based on the playback time of the video layer corresponding to the selected human body contour. The human body parts are searched forward and backward under the playback time to find the missing human body parts, and the corresponding missing human body parts can also be obtained. The video layer is used as the video layer to be integrated, then the video layer of the overall human body contour distribution of all timelines in the target video can be obtained, and then the human body contour in the video layer to be integrated is pasted to the blank set video through a fixed-point section to obtain a human body video. The human body video here is the overall human body contour of all timelines in the target video, and then the human body video corresponding to the human body contour can be obtained, which can improve the integrity of the human body video obtained in the video, has a good human body key point analysis function, can combine all video information to obtain the human body contour, and at the same time, the human body contour is traced and selected on the video timeline, and the human body contour can be extracted under complex backgrounds;

[0076] In one embodiment, the step S3 of identifying a human body in a human body video and constructing a human body motion model, and marking key points of the human body based on the human body motion model to obtain a key point motion model includes:

[0077] S31, constructing a human body virtual model in the human body video;

[0078] S32, following the human body video with the human body virtual model, and marking key points according to the motion bends of the human body virtual model to obtain a key point motion model;

[0079] As described in the above steps S31 and S32, a detailed human virtual model is constructed based on the human body contour information captured in the human body video; the two-dimensional human body contour data is converted into a three-dimensional virtual representation; first, by analyzing the human body contour in the video frame, we can obtain the basic shape and proportion of the human body, and generate a preliminary human body plane model based on this; this plane model retains the key features of the human body in two-dimensional space, such as height, limb length ratio, etc.; then, using advanced three-dimensional modeling technology, according to the characteristics of each part in the human body plane model, three-dimensional processing is performed one by one; taking the head as an example, we will carefully construct a head with a three-dimensional shape through three-dimensional modeling based on the head contour in the plane model. The three-dimensional and realistic three-dimensional head model can make the shape of the model close to that of a real human head; similar three-dimensional construction methods are used for other parts of the human body, such as the torso, limbs, etc.; based on standard human three-dimensional contour data (these data may come from medical scans, anthropometry research or three-dimensional scanning technology), three-dimensional models that conform to human proportions and forms are created for these parts; these models are not only similar to the real human body in appearance, but can also simulate the movement and posture changes of the human body to a certain extent. The entire construction process is a conversion and refinement process from two-dimensional to three-dimensional, which relies on accurate human contour recognition, advanced three-dimensional modeling technology and a deep understanding of human structure and proportions. Through this process, a human virtual model that conforms to human characteristics and has a high sense of reality can be created; the human virtual model is constructed according to the human body contour in the human body video, and the human virtual model is three-dimensionalized according to the human body contour and the characteristics of human body parts. For example, a flat human body planar model is obtained according to the proportion of the human body contour, and the corresponding head is three-dimensionally constructed into a head three-dimensional model. Other parts are three-dimensionally constructed according to the standard human body three-dimensional contour. Then, the human virtual model follows the movement of the human body video. During the movement, the limbs can be moved according to the movement in the human body video, so that the limb joints will be bent. Then, key points are marked according to the movement bending points of the human virtual model to obtain a key point motion model.

[0080] In one embodiment, the step S4 of overlapping the key point motion model with the human body contour to obtain the human body key point motion video comprises:

[0081] S41, overlapping the key point motion model with the human body contour;

[0082] S42, hiding the human body virtual model of the overlapped key point motion model to obtain a human body key point motion video;

[0083] As described in the above steps S11-S42, the construction of the human virtual model of the human key point motion model starts with the accurate recognition and extraction of the human body contour in the human body video; the human body contour is accurately outlined from the video sequence using image processing techniques, such as edge detection, contour tracking, etc.; since the human body contour is directly extracted based on the actual human body shape in the video, the constructed human virtual model can achieve a high degree of overlap with the human body in the original video in terms of proportion and shape, and then the key points can be attached to the human body contour, and then the human virtual model of the key point motion model after the overlap is hidden to obtain the human body key point motion video, then the human body movement can be displayed through the key points in the target video, and the movement of the key points can be understood through the human body key point motion video.

[0084] Example 2, please refer to Figure 2 As shown, the human body key point analysis and recognition system based on video stream described in this embodiment includes:

[0085] A layering module is used to obtain a target video based on a video stream, and to layer the target video to obtain multiple video layers;

[0086] An extraction module, connected to the layering module, is used to extract the human body contour from the multiple video layers, and integrate the multiple video layers involving the human body contour to obtain the human body video;

[0087] A construction module, connected to the extraction module, is used to identify the human body in the human body video and construct a human body motion model, and mark the key points of the human body based on the human body motion model to obtain a key point motion model;

[0088] The overlap module is connected to the construction module and is used to overlap the key point motion model with the human body contour to obtain the human body key point motion video.

[0089] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A human body key point analysis and recognition method based on video stream, characterized in that: The following steps are involved: Obtain a target video based on a video stream, and layer the target video to obtain multiple video layers; Extracting a human body contour from multiple video layers, and integrating multiple video layers involving the human body contour to obtain a human body video; Identify the human body in the human body video and build a human body motion model, and mark the key points of the human body based on the human body motion model to obtain a key point motion model; Overlap the key point motion model with the human body contour to obtain a human body key point motion video; The step of extracting a human body contour from a plurality of video layers and integrating the plurality of video layers related to the human body contour to obtain a human body video comprises: The completeness index of the human body contour in multiple video layers is determined based on the human body features, wherein the calculation formula of the completeness index is: ,in, is the complete index, is the number of moving tracking points on the human body contour that conforms to human characteristics, is the number of overlapping moving tracking points on the human body contour that conforms to human characteristics, is the weight coefficient that conforms to the characteristics of the human body, and is a constant greater than zero; Based on the complete index, multiple video layers with human body contours are determined as video layers to be integrated, and the video layers to be integrated are integrated according to fixed-point sections to obtain human body videos.

2. The human body key point analysis and recognition method based on video stream according to claim 1, characterized in that: The step of obtaining a target video based on a video stream and layering the target video to obtain multiple video layers includes: Get the target video through the video stream; Construct multiple blank framework videos, and layer the target video based on the motion features to obtain multiple layered images; Multiple layered images are stored one-to-one in correspondence with the blank frame video to obtain multiple video layers.

3. The human body key point analysis and recognition method based on video stream according to claim 2 is characterized in that: The step of constructing a plurality of blank framework videos and layering the target video based on the movement characteristics to obtain a plurality of layered images includes: Determine a plurality of storage spaces, and set a video playback framework as a blank framework video based on the plurality of storage spaces; Extracting and depicting the contours of the object in the target video to obtain multiple object contours, and laying multiple moving tracking points on the multiple object contours; Obtain the spatial relationship between multiple objects in the target video, spatially arrange the mobile tracking points based on the spatial relationship, and connect adjacent mobile tracking points to obtain a tracking network; Obtaining movement features of a moving tracking point in a tracking network in a target video, wherein the movement features include a moving direction and a moving distance of the moving tracking point; The object contours corresponding to the continuous moving tracking points with the same moving direction are extracted and divided to obtain multiple layered images.

4. The human body key point analysis and recognition method based on video stream according to claim 3 is characterized in that: The step of using multiple video layers in which human body contours are determined based on the complete index as video layers to be integrated, and integrating the video layers to be integrated according to fixed-point sections to obtain human body videos includes: Set a blank collection video; Position multiple video layers according to the playback time, set fixed-point sections between the multiple video layers, and connect the fixed-point sections with the blank set video; Obtain the human body contour with the highest completeness index in multiple video layers to select the video line, and obtain the video layer to be integrated; The human body contour in the video layer to be integrated is pasted to the blank set video through a fixed-point section to obtain a human body video.

5. The human body key point analysis and recognition method based on video stream according to claim 1, characterized in that: The steps of identifying a human body in a human body video and constructing a human body motion model, and marking key points of the human body based on the human body motion model to obtain a key point motion model include: Construct a virtual human model from human videos; The human body virtual model follows the human body video, and key points are marked according to the motion bends of the human body virtual model to obtain a key point motion model.

6. The human body key point analysis and recognition method based on video stream according to claim 1, characterized in that: The step of overlapping the key point motion model with the human body contour to obtain a human body key point motion video comprises: Overlap the key point motion model with the human body contour; The human body virtual model of the overlapped key point motion model is hidden to obtain the human body key point motion video.

7. A human body key point analysis and recognition system based on video stream, used to implement the human body key point analysis and recognition method based on video stream according to any one of claims 1 to 6, characterized in that: include: A layering module is used to obtain a target video based on a video stream, and to layer the target video to obtain multiple video layers; An extraction module, connected to the layering module, is used to extract the human body contour from the multiple video layers, and integrate the multiple video layers involving the human body contour to obtain the human body video; A construction module, connected to the extraction module, is used to identify the human body in the human body video and construct a human body motion model, and mark the key points of the human body based on the human body motion model to obtain a key point motion model; The overlap module is connected to the construction module and is used to overlap the key point motion model with the human body contour to obtain the human body key point motion video.

Images