Long jump analysis and action correction system and method based on artificial intelligence

Abstract

The invention discloses a long jump analysis and action correction system and method based on artificial intelligence, and belongs to the technical field of artificial intelligence, and the system comprises a background processing module which is used for obtaining a complete long jump video of a user, carrying out the segmentation processing of a target long jump video, and obtaining a processed frame image; the recognition and extraction module is used for recognizing the processed frame images based on a human body bone collateral point recognition model and extracting a plurality of key frame images from the frame images in which bone collateral points are recognized; the analysis and calculation module is used for performing action qualification degree analysis according to the coordinates of the bone collateral points in the plurality of key frame images, and generating action reference scores and prompts according to an analysis result; and the scoring module is used for calculating an actual long jump distance and generating a long jump reference score in combination with the long jump standard score table. According to the invention, the problems of inconvenient operation and low efficiency due to the fact that professional guidance provided in long jump practice can only be guided by a specially-assigned person in a specific field are solved.

Description

Technical Field

[0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a long jump analysis and action correction system and method based on artificial intelligence. Background Art

[0002] In recent years, the Ministry of Education has vigorously promoted comprehensive quality education and attached great importance to the comprehensive quality of users. The status of sports has become increasingly important, which is reflected in the fact that the proportion of sports scores in the total score at each stage has increased significantly. Among them, increasing the proportion of sports scores in the total score is of great significance to physical health, all-round development, competitive awareness, and further study and employment. By taking physical education exams, you can effectively improve your physical fitness and enhance your physique. Moreover, physical education exams can not only test sports skills, but also sports literacy and comprehensive abilities, such as coordination, flexibility, endurance, etc., which are helpful to the comprehensive development of your physical and intellectual abilities. By taking physical education exams, you can cultivate a sense of competition and competitive ability, laying a foundation for future study and life.

[0003] As one of the important sports events, long jump is of self-evident importance. After professional guidance and systematic training, one can significantly improve his or her performance and thus obtain better results in physical examinations. However, currently, professional guidance can only be provided by dedicated personnel in specific venues during long jump training, which is inconvenient and inefficient.

[0004] Therefore, in view of the problems existing in the prior art, it is urgent to provide a long jump analysis and action correction system and method based on artificial intelligence to overcome the above-mentioned shortcomings of the prior art. Summary of the invention

[0005] The purpose of the present invention is to provide a long jump analysis and action correction system and method based on artificial intelligence, so as to solve the problem that professional guidance can only be provided by dedicated personnel in specific venues in the prior art, which is inconvenient and inefficient.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a long jump analysis and action correction system based on artificial intelligence, comprising a background processing module, a recognition and extraction module, an analysis and calculation module, and a scoring module;

[0008] The background processing module is used to receive the user's complete long jump video, detect whether there is a target long jump video at a preset time interval, and if there is, segment and process the target long jump video to obtain a processed frame image;

[0009] The recognition and extraction module is connected to the background processing module, recognizes the processed frame image based on the human body bone point recognition model, and extracts a plurality of key frame images from the frame image in which the bone points are recognized;

[0010] The analysis and calculation module is connected to the recognition and extraction module, and is used to analyze the degree of action qualification according to the coordinates of the bone collateral points in the multiple key frame images, and generate action reference scores and prompts according to the analysis results;

[0011] The scoring module is connected to the analysis and calculation module, and is used to calculate the actual long jump distance and generate a long jump reference score in combination with a long jump standard score table.

[0012] Based on the above public content, the background processing module obtains the user's complete long jump video, and processes the long jump video together with the analysis and calculation module and the scoring module, extracts multiple key frame images from the video, and judges the degree of qualification of the action through coordinate analysis and trigonometric calculation of multiple key frame images, and gives corresponding prompts and action reference scores. The actual long jump distance is calculated according to the long jump distance in the key frame image and combined with the score table to generate a long jump reference score. There is no need for special person to provide guidance in a specific venue. The entire long jump process can be recorded with external photographic equipment and uploaded to the long jump analysis and action correction system to obtain professional guidance and provide corresponding scores.

[0013] As a further technical solution of the present invention, it also includes a video acquisition module, and the video acquisition module is used to acquire a complete long jump video of the user.

[0014] As a further technical solution of the present invention, it further comprises a front-end display module, wherein the front-end display module comprises a chart processing submodule and a data display submodule;

[0015] The chart processing submodule is connected to the scoring module and is used to count the actual long jump distance and the reference score to make a chart;

[0016] The data display submodule is connected to the chart processing submodule and is used for front-end display of analysis videos, reference scores and charts.

[0017] As a further technical solution of the present invention, it also includes a database, which is used to store the long jump standard score table, user information, analysis videos and charts, and the data in the database is updated in real time, and the database is connected to the chart processing submodule.

[0018] As a further technical solution of the present invention, it also includes an administrator module for managing user information and user login, long jump standard score table and / or system usage authority.

[0019] As a further technical solution of the present invention, it also includes an external photographic device for capturing a complete long jump video of the user.

[0020] In a second aspect, the present invention provides a long jump analysis and action correction method based on artificial intelligence, comprising:

[0021] Receive the user's complete long jump video;

[0022] Segmenting and processing the long jump video to obtain processed frame images;

[0023] Recognize the processed frame image based on the human body bone point recognition model, and extract a plurality of key frame images from the frame image in which the bone points are recognized;

[0024] Calculating the coordinates of the bone points in the plurality of key frame images, analyzing the degree of action qualification, and generating action reference scores and prompts according to the analysis results;

[0025] Calculate the actual long jump distance and generate a long jump reference score based on the long jump standard score table.

[0026] Based on the above disclosed content, the present invention obtains frame images after long jump video segmentation processing and identifies them based on a human bone collateral point recognition model, extracts multiple key frame images, analyzes the coordinates of key points in different parts of the key frame images, judges the degree of qualification of the action, gives corresponding reference scores and prompts, scores according to the actual long jump distance combined with the long jump standard score table, and provides users with guidance on non-standard actions and the score obtained for the long jump.

[0027] As a further technical solution of the present invention, segmenting and processing the long jump video to obtain processed frame images includes:

[0028] Read the long jump video;

[0029] Segmenting the long jump video into a series of continuous frames;

[0030] Extracting frame images from each of the continuous frames frame by frame;

[0031] A preprocessing operation is performed on the frame image and saved as a processed frame image.

[0032] As a further technical solution of the present invention, the processed frame image is identified based on the human body bone point recognition model, and a plurality of key frame images are extracted from the frame image in which the bone points are identified, including:

[0033] Based on the keypointrcnn_resnet50_fpn() model in Pytorch, the bone points of the processed frame image are identified, wherein the bone points include nose key points, eye key points, ear key points, shoulder key points, elbow key points, wrist key points, hip key points, knee key points and ankle key points;

[0034] When the y-axis coordinate value of the wrist key point is the smallest in the processed frame image before the ankle key point moves, extract the frame image corresponding to the time when the y-axis coordinate value of the wrist key point is the smallest as the first key frame image;

[0035] In the processed frame image before the ankle key point moves, when the difference between the y-axis coordinate values ​​of the wrist key point and the knee key point is the largest for the last time, extract the frame image corresponding to the time when the difference between the y-axis coordinate values ​​of the wrist key point and the knee key point is the largest for the last time as the second key frame image;

[0036] In the processed frame image before the ankle key point moves, when the angle between the line connecting the wrist key point to the ankle key point and the x-axis is the smallest, extract the frame image corresponding to the smallest angle between the line connecting the wrist key point to the ankle key point and the x-axis as the third key frame image;

[0037] When the y-axis coordinate value of the wrist key point is the smallest in the processed frame image after the ankle key point moves, extract the frame image corresponding to the time when the y-axis coordinate value of the wrist key point is the smallest as the fourth key frame image;

[0038] In the processed frame image after the ankle key point moves, when the y-axis coordinate value of the ankle key point is equal to the y-axis coordinate value of the ankle key point before the ankle key point moves, extract the frame image corresponding to the time when the y-axis coordinate value of the ankle key point is equal to the y-axis coordinate value of the ankle key point before the ankle key point moves as the fifth key frame image.

[0039] As a further technical solution of the present invention, according to the coordinate calculation of the bone collateral points in the plurality of key frame images, the degree of action qualification is analyzed, an action reference score is generated and a prompt is given, including:

[0040] Calculate the absolute value of the coordinate difference between any two of the wrist key points, shoulder key points, hip key points, knee key points and ankle key points in the first key frame image in the x-axis direction; if the absolute value of all coordinate differences minus the first error value is less than the first preset value, the action is qualified, and a full score is given and a prompt is given; otherwise, a prompt is given, and a corresponding score is deducted according to the absolute value of the coordinate difference between any two of the wrist key points, shoulder key points, hip key points, knee key points and ankle key points in the x-axis direction;

[0041] Determine the x-axis coordinate values ​​of the wrist key point, the knee key point, the ear key point and the ankle key point in the second key frame image. If they are arranged in order according to the size of the x-axis coordinate values, the order is the wrist key point, the ankle key point, the knee key point and the ear key point, then the action is qualified, and a full score is given and a prompt is given; otherwise, a prompt is given, and corresponding scores are deducted according to the different orders between the wrist key point, the ankle key point, the knee key point and the ear key point;

[0042] Calculate the angle between the line connecting the wrist key point to the ankle key point and the x-axis in the third key frame image. If the angle is 45 degrees, the action is qualified and a full score is given while a prompt is given. Otherwise, a prompt is given and a corresponding score is deducted according to the absolute value of the difference between the angle and 45 degrees.

[0043] Calculate the coordinate difference between the knee key point in the fourth key frame image and the knee key point in the second key frame image on the y-axis to obtain a first coordinate difference; if the first coordinate difference is greater than the second preset value, the action is qualified, a full score is given and a prompt is given; if the first coordinate difference is less than or equal to the second preset value, based on the absolute value of the difference between the second preset value and the first coordinate difference, a corresponding score is deducted and a prompt is given;

[0044] Calculate the coordinate difference between the ankle key point in the fifth key frame image and the ankle key point in the first key frame image on the x-axis to obtain a second coordinate difference. If the second coordinate difference is greater than a third preset value, the action is qualified, and a full score is given and a prompt is given; if the second coordinate difference is less than or equal to the third preset value, based on the absolute value of the difference between the third preset value and the second coordinate difference, the corresponding score is deducted and a prompt is given.

[0045] As a further technical solution of the present invention, the actual long jump distance is calculated and scored in combination with the long jump standard score table, including:

[0046] The actual long jump distance S is calculated according to the following formula:

[0047]

[0048] Where S is the actual long jump distance; h is the user's height; y 1 is the value of the y-axis coordinate of the eye key point in the first key frame image; 2 is the y-axis coordinate value of the ankle key point in the first key frame image; 1 is the x-axis coordinate value of the ankle key point in the first key frame image; 2 is the x-axis coordinate value of the ankle key point in the fifth key frame image;

[0049] The long jump reference score is calculated based on the value of the actual long jump distance S in combination with the long jump standard score table.

[0050] As a further technical solution of the present invention, before receiving the complete long jump video of the user, the method further includes:

[0051] Enter the long jump standard score table;

[0052] Enter user information, including name, class, height, student ID and gender.

[0053] Beneficial effects:

[0054] (1) The background processing module obtains the user's complete long jump video, and processes the long jump video together with the analysis and calculation module and the scoring module. It extracts multiple key frame images from the video, and judges the degree of action qualification through coordinate analysis and trigonometric calculation of the multiple key frame images, and gives corresponding prompts and action reference scores. The actual long jump distance is calculated based on the long jump distance in the key frame image and the long jump reference score is generated by combining the score table. There is no need for special personnel to provide guidance in a specific venue. The entire long jump process can be recorded using external photography equipment and uploaded to the long jump analysis and action correction system to obtain professional guidance and provide corresponding scores.

[0055] (2) The present invention obtains frame images after long jump video segmentation and recognizes them based on the human body bone point recognition model, extracts multiple key frame images, analyzes the coordinates of key points in different parts of the key frame images, determines the degree of qualification of the action, gives corresponding reference scores and prompts, scores according to the actual long jump distance combined with the corresponding long jump standard score table, and provides users with guidance on non-standard actions and the score obtained in the long jump.

[0056] (3) Users only need to use an external camera and an AI-based long jump analysis and action correction system to know the shortcomings of their long jump movements, how to improve them, and get professional guidance. By viewing the charts displayed on the front end, users can more clearly and intuitively understand their practice results and improvement range. The operation is easy, saves time and improves efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 A schematic diagram of the structure of a long jump analysis and action correction system based on artificial intelligence provided by the present invention;

[0059] Figure 2 A schematic flow chart of a long jump analysis and action correction method based on artificial intelligence provided by the present invention. DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. It should be noted that although the description of these embodiments is used to help understand the present invention, it does not constitute a limitation of the present invention. The specific structures and functional details disclosed herein are only used to describe the embodiments of the present invention. However, the present invention can be embodied in many alternative forms, and it should not be understood that the present invention is limited to the embodiments set forth herein.

[0061] It should be understood that although the terms first and second, etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another object. For example, a first object can be referred to as a second object, and similarly, a second object can be referred to as a first object without departing from the scope of the exemplary embodiments of the present invention.

[0062] It should be understood that the term "and / or" that may appear in this article is merely a description of the association relationship between associated objects, indicating that there may be three relationships. For example, A and / or B can represent three situations: A exists alone, B exists alone, or A and B exist at the same time; the term " / and" that may appear in this article describes another type of association object relationship, indicating that there may be two relationships. For example, A / and B can represent two situations: A exists alone or A and B exist at the same time; in addition, the character " / " that may appear in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0063] Example

[0064] First, as Figure 1As shown, the present embodiment provides a long jump analysis and action correction system based on artificial intelligence, including but not limited to a background processing module, an identification and extraction module, an analysis and calculation module and a scoring module; the background processing module is used to obtain a complete long jump video of the user, and detect whether there is a target long jump video at a preset time interval. If there is, the target long jump video is segmented and processed to obtain a processed frame image; the identification and extraction module is connected to the background processing module, and the processed frame image is identified based on a human body bone collateral point recognition model, and a plurality of key frame images are extracted from the frame image in which the bone collateral points are identified; the analysis and calculation module is connected to the identification and extraction module, and is used to analyze the degree of action qualification according to the coordinates of the bone collateral points in the plurality of key frame images, and generate action reference scores and prompts according to the analysis results; the scoring module is connected to the analysis and calculation module, and is used to calculate the actual long jump distance and generate a long jump reference score in combination with the long jump standard score table.

[0065] Specifically, the background processing module uses asynchronous operations to process long jump videos. When analyzing long jump videos, users do not need to wait for the upload task to be completed on the page. Other tasks can be performed while the video is being uploaded, which will not block or delay the progress of video upload.

[0066] In one possible design, the prompts are recorded into the target long jump video to generate an analysis video for the user to review.

[0067] In a feasible implementation manner, it also includes a video acquisition module, and the video acquisition module is used to acquire a complete long jump video of the user.

[0068] Furthermore, external devices for capturing long jump videos include but are not limited to mobile phones, camcorders and digital cameras. When using external photographic equipment to capture a user's complete long jump video, there are no site requirements or reference requirements. It only needs to be shot at an angle perpendicular to the long jump direction.

[0069] In a feasible implementation, it further includes a front-end display module, which includes a chart processing submodule and a data display submodule;

[0070] The chart processing submodule is connected to the scoring module and is used to count the actual long jump distance and the reference score to make a chart;

[0071] The data display submodule is connected to the chart processing submodule and is used for front-end display of analysis videos, reference scores and charts.

[0072] The chart processing submodule creates line graphs and bar graphs based on the actual long jump distance calculated in the scoring module and the reference scores generated by gender and the long jump standard score table. The data display submodule displays the analyzed video, reference scores, line graphs and bar graphs on the front-end page to help users understand their own progress trends.

[0073] In a feasible implementation manner, it further includes a database for storing the long jump standard score table, user information, analysis videos and charts, and the data in the database is updated in real time, and the database is connected to the chart processing submodule.

[0074] Specifically, the long jump standard score table includes but is not limited to examination score tables for primary schools, middle schools, universities, and sports schools.

[0075] In a possible design, the design of the database includes: using MySQL database to build, including 5 tables, namely b_student, b_type, b_video, t_user_audit_hash and user table, among which b_student is a user information table, which includes name, class, student number, height, creation date and gender, etc.; b_video table is a user long jump analysis table, including analysis video, long jump distance and score, etc., which can be queried on the front end, and a linear chart is generated at the same time, so that users can see their own progress and facilitate subsequent queries.

[0076] In a feasible embodiment, an administrator module is also included to manage user information, long jump standard score table, and user login and / or system use rights.

[0077] Among them, it should be noted that users have the authority to view their own information, change passwords, upload long jump video analysis, and view their own analysis videos, reference scores and charts. Administrators can manage user information, reset user passwords, and update the long jump standard score table in the system according to the different long jump standard score tables each year.

[0078] Second, as Figure 2 As shown, this embodiment provides a long jump analysis and action correction method based on artificial intelligence, including but not limited to steps S101 to S105:

[0079] Step S101. Receive a complete long jump video from a user;

[0080] It should be noted that the external devices for capturing long jump videos in this embodiment include but are not limited to mobile phones, video cameras, and digital cameras, and there are no requirements for sites and reference objects during the capturing process.

[0081] In a possible design, before step S101. collecting a complete long jump video of the user, the method further includes:

[0082] Step S1011. Enter the long jump standard score table;

[0083] Step S1012. Enter user information, where the user information includes name, class, height, student ID and gender.

[0084] Based on the above public content, the administrator needs to enter the long jump standard score table for that year and import the user's relevant information, including name, class, height, student number and gender.

[0085] Step S102: Segment and process the long jump video to obtain processed frame images;

[0086] As a possible design of step S102, segmenting and processing the long jump video to obtain processed frame images includes but is not limited to:

[0087] Step S1021. Read the long jump video;

[0088] Step S1022: segmenting the long jump video into a series of continuous frames;

[0089] Step S1023. Extracting frame images from each of the continuous frames frame by frame;

[0090] Step S1024: Perform a preprocessing operation on the frame image and save it as a processed frame image.

[0091] Step S103. Recognize the processed frame image based on the human body bone point recognition model, and extract a plurality of key frame images from the frame image in which the bone points are recognized;

[0092] Wherein, as a possible design of step S103, the processed frame image is identified based on the human body bone point recognition model, and a plurality of key frame images are extracted from the frame image in which the bone points are identified, including but not limited to:

[0093] Step S1031. Based on the keypointrcnn_resnet50_fpn() model in Pytorch, bone points are identified for the processed frame image, wherein the bone points include nose key points, eye key points, ear key points, shoulder key points, elbow key points, wrist key points, hip key points, knee key points and ankle key points;

[0094] Step S1032. The upper left corner of the processed frame image is taken as the coordinate origin, the horizontal direction is the x-axis, the vertical direction is the y-axis, the y-axis is downward as the positive direction, and the unit coordinate length is 1 pixel;

[0095] Step S1033. When the y-axis coordinate value of the wrist key point is the smallest in the processed frame image before the ankle key point moves, the frame image corresponding to the time when the y-axis coordinate value of the wrist key point is the smallest is used as the first key frame image;

[0096] Step S1034. In the processed frame image before the ankle key point moves, when the difference between the y-axis coordinate values ​​of the wrist key point and the knee key point is the largest for the last time, use the frame image corresponding to the time when the difference between the y-axis coordinate values ​​of the wrist key point and the knee key point is the largest for the last time as the second key frame image;

[0097] Step S1035. In the processed frame image before the ankle key point moves, when the angle between the line connecting the wrist key point to the ankle key point and the x-axis is the smallest, the frame image corresponding to the angle between the line connecting the wrist key point to the ankle key point and the x-axis is the smallest is used as the third key frame image;

[0098] Step S1036. When the y-axis coordinate value of the wrist key point is the smallest in the processed frame image after the ankle key point moves, the frame image corresponding to the time when the y-axis coordinate value of the wrist key point is the smallest is used as the fourth key frame image;

[0099] Step S1037. When the y-axis coordinate value of the ankle key point in the processed frame image after the ankle key point moves is equal to the y-axis coordinate value of the ankle key point before the ankle key point moves, the frame image corresponding to the time when the y-axis coordinate value of the ankle key point is equal to the y-axis coordinate value of the ankle key point before the ankle key point moves is used as the fifth key frame image.

[0100] Step S104. Calculating the coordinates of the bone points in the plurality of key frame images, analyzing the degree of action qualification, and generating action reference scores and prompts according to the analysis results;

[0101] Among them, as a possible design of step S104, according to the coordinate calculation of the bone collateral points in the multiple key frame images, the degree of action qualification is analyzed, the action reference score is generated and prompts are given, including but not limited to:

[0102] Step S1041. Calculate the absolute value of the coordinate difference between any two of the wrist key points, shoulder key points, hip key points, knee key points and ankle key points in the first key frame image in the x-axis direction. If the absolute value of all coordinate differences minus the first error value is less than the first preset value, the action is qualified, and a full score is given and a prompt is given; otherwise, a prompt is given, and the corresponding score is deducted according to the absolute value of the coordinate difference between any two of the wrist key points, shoulder key points, hip key points, knee key points and ankle key points in the x-axis direction.

[0103] For example, when the absolute value of the difference between the x-coordinate values ​​of the wrist key point and the shoulder key point minus the error value of 5 pixels is greater than the first preset value, a prompt is given: "Stretch your hands upward, stand up straight, and try to coincide with the red line from your hands to your feet"; if the absolute value of the difference minus the first error value is less than the first preset value, the action is qualified, and a prompt "You did it very standardly" is given, where the first preset value is entered manually.

[0104] Step S1042. Determine the x-axis coordinate values ​​of the wrist key point, the knee key point, the ear key point and the ankle key point in the second key frame image. If they are arranged in order according to the size of the x-axis coordinate values, the order is the wrist key point, the ankle key point, the knee key point and the ear key point, then the action is qualified, and a full score is given and a prompt is given; otherwise, a prompt is given, and the corresponding score is deducted according to the different orders among the wrist key point, the ankle key point, the knee key point and the ear key point;

[0105] Step S1043. Calculate the angle between the line connecting the wrist key point to the ankle key point and the x-axis in the third key frame image. If the angle is 45 degrees, the action is qualified and a full score is given while a prompt is given. Otherwise, a prompt is given and a corresponding score is deducted according to the absolute value of the difference between the angle and 45 degrees.

[0106] For example, when a move is unsatisfactory, the prompt given may be "Your current angle is 69 degrees, try to align your body with the red auxiliary line." Since the difference between 69 degrees and 45 degrees is 24 degrees, the score obtained is 100 points minus 24 points, and the final score is 76 points.

[0107] Step S1044. Calculate the coordinate difference between the knee key point in the fourth key frame image and the knee key point in the second key frame image on the y-axis to obtain a first coordinate difference. If the first coordinate difference is greater than the second preset value, the action is qualified, and a full score is given and a prompt is given; if the first coordinate difference is less than or equal to the second preset value, then based on the absolute value of the difference between the second preset value and the first coordinate difference, the corresponding score is deducted and a prompt is given;

[0108] Specifically, the second preset value is entered manually.

[0109] Step S1045. Calculate the coordinate difference between the ankle key point in the fifth key frame image and the ankle key point in the first key frame image on the x-axis to obtain a second coordinate difference. If the second coordinate difference is greater than a third preset value, the action is qualified, and a full score is given and a prompt is given; if the second coordinate difference is less than or equal to the third preset value, then based on the absolute value of the difference between the third preset value and the second coordinate difference, the corresponding score is deducted and a prompt is given.

[0110] Step S105. Calculate the actual long jump distance and score it in combination with the long jump standard score table;

[0111] Among them, as a possible design of step S105, calculating the actual long jump distance and scoring in combination with the long jump standard score table includes:

[0112] The actual long jump distance S is calculated according to the following formula:

[0113]

[0114] Where S is the actual long jump distance; h is the user's height; y 1 is the y-axis coordinate value of the eye key point in the first key frame image; 2 is the y-axis coordinate value of the ankle key point in the first key frame image; 1 is the x-axis coordinate value of the ankle key point in the first key frame image; 2 is the x-axis coordinate value of the ankle key point in the fifth key frame image;

[0115] The long jump reference score is calculated based on the value of the actual long jump distance S in combination with the long jump standard score table.

[0116] In this embodiment, multiple computers with graphics cards are collected, and the graphics cards are converted into GPU computing power through GUDA (General-Purpose GPU Development Accelerator), which provides a programming model and interface that enables developers to use the parallel computing capabilities of graphics cards to perform general computing tasks, not just graphics rendering. GUDA is usually used in conjunction with programming languages ​​​​(such as CUDA), which provide tools and libraries for writing parallel computing codes, significantly speeding up the speed of video analysis and providing higher computing throughput.

[0117] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A long jump analysis and action correction system based on artificial intelligence, characterized in that: It includes background processing module, recognition and extraction module, analysis and calculation module and scoring module; The background processing module is used to receive the user's complete long jump video, detect whether there is a target long jump video at a preset time interval, and if there is, segment and process the target long jump video to obtain a processed frame image; The recognition and extraction module is connected to the background processing module, recognizes the processed frame image based on the human body bone point recognition model, and extracts a plurality of key frame images from the frame image in which the bone points are recognized; The analysis and calculation module is connected to the recognition and extraction module, and is used to analyze the degree of action qualification according to the coordinates of the bone collateral points in the multiple key frame images, and generate action reference scores and prompts according to the analysis results; The scoring module is connected to the analysis and calculation module, and is used to calculate the actual long jump distance and generate a long jump reference score in combination with a long jump standard score table.

2. The artificial intelligence-based long jump analysis and action correction system according to claim 1, characterized in that: It also includes a video acquisition module, which is used to acquire a complete long jump video of the user.

3. The artificial intelligence-based long jump analysis and action correction system according to claim 1, characterized in that: It also includes a front-end display module, which includes a chart processing submodule and a data display submodule; The chart processing submodule is connected to the scoring module and is used to count the actual long jump distance and the reference score to make a chart; The data display submodule is connected to the chart processing submodule and is used for front-end display of analysis videos, reference scores and charts.

4. The artificial intelligence-based long jump analysis and action correction system according to claim 3, characterized in that: It also includes a database for storing the long jump standard score table, user information, analysis videos and charts, and the data in the database is updated in real time. The database is connected to the chart processing submodule.

5. The artificial intelligence-based long jump analysis and action correction system according to claim 1, characterized in that: It also includes an administrator module for managing user information, long jump standard score tables, and user rights to log in and / or use the system.

6. A long jump analysis and action correction method based on artificial intelligence, characterized in that: include: Receive the user's complete long jump video; Segmenting and processing the long jump video to obtain processed frame images; Recognize the processed frame image based on the human body bone point recognition model, and extract a plurality of key frame images from the frame image in which the bone points are recognized; Calculating the coordinates of the bone points in the plurality of key frame images, analyzing the degree of action qualification, and generating action reference scores and prompts according to the analysis results; Calculate the actual long jump distance and generate a long jump reference score based on the long jump standard score table.

7. The method for long jump analysis and action correction based on artificial intelligence according to claim 6, characterized in that: Segmenting and processing the long jump video to obtain processed frame images includes: Read the long jump video; Segmenting the long jump video into a series of continuous frames; Extracting frame images from each of the continuous frames frame by frame; A preprocessing operation is performed on the frame image and saved as a processed frame image.

8. The method for long jump analysis and action correction based on artificial intelligence according to claim 6, characterized in that: The processed frame image is identified based on the human body bone point recognition model, and a plurality of key frame images are extracted from the frame image in which the bone points are identified, including: Based on the keypointrcnn_resnet50_fpn() model in Pytorch, the bone points of the processed frame image are identified, wherein the bone points include nose key points, eye key points, ear key points, shoulder key points, elbow key points, wrist key points, hip key points, knee key points and ankle key points; In the processed frame image before the ankle key point moves, when the y-axis coordinate value of the wrist key point is the smallest, the frame image corresponding to the time when the y-axis coordinate value of the wrist key point is the smallest is used as the first key frame image; In the processed frame image before the ankle key point moves, when the difference between the y-axis coordinate values ​​of the wrist key point and the knee key point is the largest for the last time, the frame image corresponding to the time when the difference between the y-axis coordinate values ​​of the wrist key point and the knee key point is the largest for the last time is used as the second key frame image; In the processed frame image before the ankle key point moves, when the angle between the line connecting the wrist key point to the ankle key point and the x-axis is the smallest, the frame image corresponding to the angle between the line connecting the wrist key point to the ankle key point and the x-axis is the smallest is taken as the third key frame image; When the y-axis coordinate value of the wrist key point is the smallest in the processed frame image after the ankle key point moves, the frame image corresponding to the time when the y-axis coordinate value of the wrist key point is the smallest is used as the fourth key frame image; In the processed frame image after the ankle key point moves, when the y-axis coordinate value of the ankle key point is equal to the y-axis coordinate value of the ankle key point before the ankle key point moves, the frame image corresponding to the time when the y-axis coordinate value of the ankle key point is equal to the y-axis coordinate value of the ankle key point before the ankle key point moves is taken as the fifth key frame image.

9. The method for long jump analysis and action correction based on artificial intelligence according to claim 8, characterized in that: Calculating the coordinates of the bone points in the multiple key frame images, analyzing the degree of action qualification, generating action reference scores and giving prompts, including: Calculate the absolute value of the coordinate difference between any two of the wrist key points, shoulder key points, hip key points, knee key points and ankle key points in the first key frame image in the x-axis direction; if the absolute value of all coordinate differences minus the first error value is less than the first preset value, the action is qualified, and a full score is given and a prompt is given; otherwise, a prompt is given, and a corresponding score is deducted according to the absolute value of the coordinate difference between any two of the wrist key points, shoulder key points, hip key points, knee key points and ankle key points in the x-axis direction; Determine the x-axis coordinate values ​​of the wrist key point, the knee key point, the ear key point and the ankle key point in the second key frame image. If they are arranged in order according to the size of the x-axis coordinate values, the order is the wrist key point, the ankle key point, the knee key point and the ear key point, then the action is qualified, and a full score is given and a prompt is given; otherwise, a prompt is given, and corresponding scores are deducted according to the different orders between the wrist key point, the ankle key point, the knee key point and the ear key point; Calculate the angle between the line connecting the wrist key point to the ankle key point and the x-axis in the third key frame image. If the angle is 45 degrees, the action is qualified and a full score is given while a prompt is given. Otherwise, a prompt is given and a corresponding score is deducted according to the absolute value of the difference between the angle and 45 degrees. Calculate the coordinate difference between the knee key point in the fourth key frame image and the knee key point in the second key frame image on the y-axis to obtain a first coordinate difference; if the first coordinate difference is greater than the second preset value, the action is qualified, a full score is given and a prompt is given; if the first coordinate difference is less than or equal to the second preset value, based on the absolute value of the difference between the second preset value and the first coordinate difference, a corresponding score is deducted and a prompt is given; Calculate the coordinate difference between the ankle key point in the fifth key frame image and the ankle key point in the first key frame image on the x-axis to obtain a second coordinate difference. If the second coordinate difference is greater than a third preset value, the action is qualified, and a full score is given and a prompt is given; if the second coordinate difference is less than or equal to the third preset value, based on the absolute value of the difference between the third preset value and the second coordinate difference, the corresponding score is deducted and a prompt is given.

10. The method for long jump analysis and action correction based on artificial intelligence according to claim 8, characterized in that: Calculate the actual long jump distance and score it based on the long jump standard score table, including: The actual long jump distance S is calculated according to the following formula: Wherein, S is the actual long jump distance; h is the height of the user; y1 is the y-axis coordinate value of the eye key point in the first key frame image; y2 is the y-axis coordinate value of the ankle key point in the first key frame image; x1 is the x-axis coordinate value of the ankle key point in the first key frame image; x2 is the x-axis coordinate value of the ankle key point in the fifth key frame image; The long jump reference score is calculated based on the value of the actual long jump distance S in combination with the long jump standard score table.

11. The method for long jump analysis and action correction based on artificial intelligence according to claim 5, characterized in that: Before receiving the user's complete long jump video, it also includes: Enter the long jump standard score table; Enter user information, including name, class, height, student ID and gender.

Images