A personalized running training service method, system, and medium

Abstract

This invention belongs to the field of running data processing and discloses a personalized running training service method, system, and medium. The method includes: S1, acquiring a video frame sequence and a speed sequence of an athlete exercising on a treadmill; S2, segmenting the video frame sequence based on the speed sequence to obtain multiple sub-video frame sequences; S3, extracting the foreground from each video frame in each sub-video frame sequence to obtain a foreground image; S4, analyzing each foreground image to obtain the analysis result of the athlete's running posture. This invention can utilize the high foreground similarity in the sub-video frame sequence to acquire the foreground image, avoiding inter-frame differencing of all image frames, effectively improving the efficiency of foreground image acquisition, and also effectively improving the integrity of the foreground image.

Description

Technical Field

[0001] This invention relates to the field of running data processing, and more particularly to a personalized running training service method, system, and medium. Background Technology

[0002] In analyzing the movement posture of a person on a treadmill, the process usually begins by acquiring a sequence of video frames of the person moving on the treadmill using a camera. The video frame sequence is then analyzed to identify areas where the person's running posture needs improvement.

[0003] In existing technologies, to improve analysis efficiency, foreground regions are typically extracted from each video frame before analysis. This reduces the number of pixels that need to be calculated during feature extraction, thus improving efficiency. Current methods for feature extraction from video frame sequences generally use inter-frame differencing, comparing the current frame with the previous frame and identifying areas of significant change as foreground regions. This method has drawbacks: it requires comparing a large number of pixels each time, impacting foreground extraction efficiency; and the obtained foreground regions are prone to holes, affecting the quality of subsequent features and resulting in inaccurate analysis results for running posture. Summary of the Invention

[0004] The purpose of this invention is to disclose a personalized running training service method, system, and medium to solve the technical problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides a personalized running training service method, comprising:

[0007] S1, acquire the video frame sequence and speed sequence of the athlete exercising on the treadmill;

[0008] S2, segment the video frame sequence based on the velocity sequence to obtain multiple sub-video frame sequences, including:

[0009] First sub-video frame sequence It contains N video frames, where N is the preset number;

[0010] For the nth sub-video frame sequence, where n is greater than or equal to 2, the process of determining the number of video frames contained in the nth sub-video frame sequence includes:

[0011] Get the first frame in the (n-1)th sub-video frame sequence and the last frame , This represents the total number of video frames contained in the (n-1)th video frame;

[0012] Obtain each and Foreground area in and ;

[0013] based on and Calculate the first control parameter;

[0014] Obtain the sub-velocity sequence corresponding to the video frame in the (n-1)th sub-video frame from the velocity sequence;

[0015] The second control parameter is calculated based on the sub-velocity sequence;

[0016] The number of video frames contained in the nth sub-video frame sequence is calculated based on the first control parameter and the second control parameter.

[0017] S3, extract the foreground image from the video frames in each sub-video frame sequence;

[0018] S4 analyzes each foreground image to obtain the analysis results of the athlete's running posture.

[0019] Preferably, the speed sequence includes a time series of the rotational speed of the treadmill belt.

[0020] Preferably, the number of video frames in the video frame sequence is the same as the number of rotation speeds in the speed sequence;

[0021] Each video frame in the video frame sequence corresponds to a rotation speed in the velocity sequence. The process of obtaining the rotation speed is as follows:

[0022] For video frames generated at time t ,but The corresponding rotational speed is , Let be the rotational speed of the treadmill belt at time t.

[0023] Preferably, obtain respectively and Foreground area in and ,include:

[0024] For video frames , ∈{ , }, The process of obtaining the foreground region is as follows:

[0025] Use the inter-frame difference algorithm to obtain The first area in;

[0026] The second region is obtained by performing region growing on the edge pixels of the first region;

[0027] The second region is then filtered to obtain the foreground region.

[0028] Preferably, an inter-frame difference algorithm is used to obtain... The first region includes:

[0029] like If the first frame in the video frame sequence is used, then the second frame in the video frame sequence is used as the comparison frame. Otherwise, obtain from the video frame sequence. The previous video frame is used as the comparison frame. ;

[0030] based on Calculate separately The grayscale difference of each pixel in the image;

[0031] Will Pixels with a grayscale difference greater than the adaptive grayscale threshold are selected as pixels in the first region.

[0032] Preferably, calculate separately The grayscale difference of each pixel in the data includes:

[0033] for The coordinates in are pixels The formula for calculating the grayscale difference is:

[0034]

[0035] for The difference in grayscale values, for The coordinates in are pixels, for grayscale value, for grayscale value, for The median coordinate is The pixels.

[0036] Preferably, the pixels at the edges of the first region are subjected to region growing processing to obtain the second region, including:

[0037] Obtain the set A of pixels located at the edge of the first region;

[0038] exist In this process, each pixel in A is used as a seed point for region growing to obtain the second region.

[0039] Preferably, the value of N is 150.

[0040] Secondly, the present invention provides a personalized running training service system, including an acquisition module, a segmentation module, an extraction module, and an analysis module;

[0041] The acquisition module is used to acquire video frame sequences and speed sequences of an athlete exercising on a treadmill;

[0042] The segmentation module is used to segment a video frame sequence based on a velocity sequence to obtain multiple sub-video frame sequences;

[0043] The extraction module is used to extract the foreground from each video frame in each sub-video frame sequence to obtain the foreground image;

[0044] The analysis module is used to analyze each foreground image separately to obtain the analysis results of the athlete's running posture.

[0045] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method of the first aspect of the present invention.

[0046] Beneficial effects:

[0047] Compared with existing technologies, the present invention does not directly perform inter-frame differencing on each frame to obtain the foreground region in the process of acquiring foreground images of video frames containing an athlete's running posture. Instead, it first divides the video frame sequence into multiple sub-video frame sequences, and then extracts the foreground in each sub-video frame sequence. In this way, the high similarity of the foreground in the sub-video frame sequence can be used to acquire the foreground image, avoiding inter-frame differencing on all image frames, effectively improving the efficiency of foreground image acquisition, and also effectively improving the integrity of the foreground image. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of a personalized running training service method according to the present invention.

[0050] Figure 2 This is a schematic diagram of a personalized running training service system according to the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] like Figure 1 As shown, in a first aspect, the present invention provides a personalized running training service method, comprising:

[0053] S1, acquire the video frame sequence and speed sequence of the athlete exercising on the treadmill.

[0054] This step mainly involves acquiring data for analysis, including video frame sequences and speed sequences. Video frame sequences can be acquired using a camera installed on the side of the treadmill's running belt; while speed sequences can be obtained by exporting data from the treadmill's internal storage, or by using a speed sensor to measure the speed of the running belt. It is important to note that the speed measurement frequency must match the frame rate of the video frame sequence to ensure that each image frame corresponds to a rotation speed.

[0055] Preferably, the speed sequence includes a time series of the rotational speed of the treadmill belt.

[0056] Preferably, the number of video frames in the video frame sequence is the same as the number of rotation speeds in the speed sequence;

[0057] Each video frame in the video frame sequence corresponds to a rotation speed in the velocity sequence. The process of obtaining the rotation speed is as follows:

[0058] For video frames generated at time t ,but The corresponding rotational speed is , Let be the rotational speed of the treadmill belt at the time of recording, which is equal to t.

[0059] The above operations mainly involve data alignment, which allows the rotation speed to be used to acquire sub-video frame sequences in subsequent processes.

[0060] S2, the video frame sequence is segmented based on the velocity sequence to obtain multiple sub-video frame sequences.

[0061] This step involves creating a sub-video frame sequence from video frames with high foreground similarity. This avoids the need to use inter-frame differencing to obtain the foreground region for all image frames in subsequent processing.

[0062] Preferably, the video frame sequence is segmented based on the velocity sequence to obtain multiple sub-video frame sequences, including:

[0063] First sub-video frame sequence It contains N video frames, where N is the preset number;

[0064] For the nth sub-video frame sequence, where n is greater than or equal to 2, the process of determining the number of video frames contained in the nth sub-video frame sequence includes:

[0065] Get the first frame in the (n-1)th sub-video frame sequence and the last frame , This represents the total number of video frames contained in the (n-1)th video frame;

[0066] Obtain each and Foreground area in and ;

[0067] based on and Calculate the first control parameter;

[0068] Obtain the sub-velocity sequence corresponding to the video frame in the (n-1)th sub-video frame from the velocity sequence;

[0069] The second control parameter is calculated based on the sub-velocity sequence;

[0070] The number of video frames contained in the nth sub-video frame sequence is calculated based on the first control parameter and the second control parameter.

[0071] The above process, by calculating the first control parameter and the second control parameter, can comprehensively calculate the athlete's movement range from two different perspectives, which is beneficial to improving the similarity of the foreground in the image frames of the sub-video frame sequence.

[0072] Preferably, the value of N can be 150. The value of N can be determined based on the frame rate of the video frame sequence; the higher the frame rate, the larger the value of N.

[0073] Preferably, obtain respectively and Foreground area in and ,include:

[0074] For video frames , ∈{ , }, The process of obtaining the foreground region is as follows:

[0075] Use the inter-frame difference algorithm to obtain The first area in;

[0076] The second region is obtained by performing region growing on the edge pixels of the first region;

[0077] The second region is then filtered to obtain the foreground region.

[0078] Traditional inter-frame differencing can easily result in holes in the obtained foreground image and may misidentify areas that do not belong to the foreground as foreground areas. Therefore, this invention performs region growing to fill the holes after inter-frame differencing, and then filters the obtained second region to reduce the number of background pixels contained in the foreground region.

[0079] Preferably, an inter-frame difference algorithm is used to obtain... The first region includes:

[0080] like If the first frame in the video frame sequence is used, then the second frame in the video frame sequence is used as the comparison frame. Otherwise, obtain from the video frame sequence. The previous video frame is used as the comparison frame. .

[0081] This approach is primarily to avoid the inability to acquire the comparison frame used for inter-frame difference during the acquisition of the second sub-video frame sequence, since the first frame in the video frame sequence does not have a preceding frame.

[0082] based on Calculate separately The grayscale difference of each pixel in the image;

[0083] Will Pixels with a grayscale difference greater than the adaptive grayscale threshold are selected as pixels in the first region.

[0084] Preferably, the adaptive grayscale threshold can be determined by the following method:

[0085]

[0086] For adaptive grayscale threshold, for The average grayscale value of the pixels in the image. for The variance of the grayscale values ​​of the pixels in the image. These are control parameters.

[0087] By using an adaptive method to determine the grayscale threshold, the grayscale threshold can be changed as the image content changes, thus enabling more accurate identification of the foreground region in the image.

[0088] In this invention, the control parameter can be 0.2.

[0089] Preferably, calculate separately The grayscale difference of each pixel in the data includes:

[0090] for The coordinates in are pixels The formula for calculating the grayscale difference is:

[0091]

[0092] for The difference in grayscale values, for The coordinates in are pixels, for grayscale value, for grayscale value, for The median coordinate is The pixels.

[0093] When calculating the grayscale difference, the difference in grayscale values ​​is calculated for pixels with the same coordinates in two image frames to obtain the result. x and y are the X-axis coordinates and Y-axis coordinates in a Cartesian coordinate system established with the same origin in the two image frames, respectively.

[0094] Preferably, the pixels at the edges of the first region are subjected to region growing processing to obtain the second region, including:

[0095] Obtain the set A of pixels located at the edge of the first region;

[0096] exist In this process, each pixel in A is used as a seed point for region growing to obtain the second region.

[0097] Since A contains multiple pixels, by performing region growing on each pixel in A, the area of ​​the first region can be expanded, thus achieving the effect of filling holes.

[0098] Preferably, the second region is filtered to obtain the foreground region, including:

[0099] Calculate the filter value for the second region;

[0100] The second region with the largest filter value is selected as the foreground region.

[0101] Preferably, the formula for calculating the screening value is:

[0102]

[0103] This is the filter value for the second region b. M is the total number of pixels contained in the second region b. The total number of pixels, The center of the second region b and The distance between the centers, for The diagonal length of the value, w is the first weight (e.g., 0.7), and Nz indicates normalization.

[0104] The above formula allows for a larger area, with... The closer the second region is to the center, the larger the filter value, thus enabling accurate filtering of the foreground region.

[0105] Preferably, based on and Calculate the first control parameter, including:

[0106] Step 1, Calculation and The similarity between them is siml;

[0107] The second step is to calculate. The center and The distance dm between the centers;

[0108] The third step is to calculate the first control parameter:

[0109]

[0110] The first control parameter, As the second weight (e.g., 0.5), Nz indicates that normalization is performed.

[0111] The first control parameter is mainly to take into account the degree of change in the content and position of the foreground in the previous sub-video frame sequence. The smaller the degree of change in content (i.e., the greater the similarity) and the smaller the degree of change in position (i.e., the smaller the value of dm), the larger the first control parameter will be. In this way, the next sub-video frame sequence can contain more video frames, thereby improving the overall acquisition efficiency of the foreground image.

[0112] Preferably, obtaining the sub-velocity sequence corresponding to the video frame in the (n-1)th sub-video frame from the velocity sequence includes:

[0113] From the velocity sequence, the rotational speeds whose recording time is the same as the generation time of the video frame in the (n-1)th sub-video frame are stored in the sub-velocity sequence. The earlier the recording time of the rotational speed, the earlier its position in the sub-velocity sequence.

[0114] Preferably, calculating the second control parameter based on the sub-velocity sequence includes:

[0115] The first step is to obtain the weighted rotational velocity based on the sub-velocity sequence, including:

[0116] Calculate the weighted rotational velocity using the following formula:

[0117]

[0118] For weighted rotational speed, Let be the i-th rotational speed in the sub-velocity sequence, and NS be the total number of rotational speeds in the sub-velocity sequence.

[0119] The purpose of calculating the weighted rotational speed is to obtain a speed value that can reflect both the speed change and the overall speed situation when the rotational speed of the treadmill belt changes.

[0120] The second step is to obtain the rotational speed fluctuation values ​​in the sub-velocity sequence, including:

[0121] Obtain the linear regression line corresponding to the rotational speed of the sub-velocity sequence;

[0122] The normalized value corresponding to the absolute value of the slope of the linear regression line is used as the rotational speed fluctuation value k.

[0123] The third step is to calculate the second control parameter based on the weighted rotational speed and the rotational speed fluctuation value, including:

[0124]

[0125] This is the second control parameter. This represents the maximum rotational speed in the sub-velocity sequence. It is the third weight (e.g., 0.6).

[0126] The second control parameter of the present invention is calculated from the characteristics of rotation speed. Therefore, if the weighted rotation speed is greater and the rotation speed fluctuation value is smaller, the second control parameter is larger. Under the same conditions, the number of video frames contained in the next sub-video frame sequence can be increased.

[0127] Preferably, calculating the number of video frames contained in the nth sub-video frame sequence based on the first control parameter and the second control parameter includes:

[0128]

[0129] The number of video frames contained in the nth sub-video frame sequence. It is the fourth weight (e.g., 0.6).

[0130] By combining two control parameters, the number of video frames in the sub-video frame sequence of the present invention can be increased by the smaller the change in the athlete's posture, the greater the speed of the treadmill, and the smaller the fluctuation, thus improving the acquisition efficiency of the sub-video frame sequence; conversely, the number of video frames in the next sub-video frame sequence will be reduced to ensure the similarity of the foreground part in the same sub-video frame sequence.

[0131] After calculating the number of video frames contained in the nth sub-video frame sequence, the nth sub-video frame sequence is then... Zhang video frame to the first The nth video frame is a video frame contained in the nth sub-video frame sequence. The total number of video frames contained in the j-th sub-video frame sequence.

[0132] When it appears When the number of frames exceeds the total number of frames in the video frame sequence, directly select the first frame from the video frame sequence. The sequence of video frames from the first video frame to the last video frame is the nth sub-video frame sequence.

[0133] S3, perform foreground extraction on each video frame in the sub-video frame sequence to obtain foreground images, including:

[0134] The first step is to obtain the foreground regions of the first and last frames in the sub-video frame sequence, respectively;

[0135] The second step is to obtain the union B3 of the set B1 of the coordinates of the pixels in the first frame and the set B2 of the coordinates of the pixels in the foreground area of ​​the last frame.

[0136] The third step involves forming the foreground image from the pixels in the foreground region for the first and last frames in the sub-video frame sequence; and forming the foreground image corresponding to the video frame from the pixels whose coordinates belong to set B3 for the remaining video frames.

[0137] This step takes advantage of the high similarity of the foreground regions in the sub-video frame sequence. By first calculating the foreground regions of the first and last frames, the union of the coordinates of the pixels in these two foreground regions can be used to determine the pixels belonging to the foreground image in the remaining video frames. This eliminates the need for frame-by-frame difference calculation, effectively improving the efficiency of foreground image acquisition while ensuring the integrity of the foreground image.

[0138] S4 analyzes each foreground image to obtain the analysis results of the athlete's running posture, including:

[0139] Each foreground image is identified separately, and the analysis results are obtained.

[0140] Preferably, the process of recognizing the foreground image includes:

[0141] The first step is to calculate biomechanical parameters such as joint angles, angular velocity, linear velocity, and changes in the center of gravity.

[0142] The second step is temporal pattern analysis: using models such as LSTM or Transformer to analyze the coordination and rhythm of action sequences.

[0143] The third step is to compare with the standard model: compare the athlete's movements with the champion model or ideal model to find the differences.

[0144] The fourth step is visualization and feedback: The analysis results are fed back to coaches and athletes in an intuitive form (such as overlaying videos of key points and angles, generating training reports) to point out specific areas that need improvement.

[0145] The analysis results can focus on the differences between the gait, ground contact angle, and body lean angle and the standard model.

[0146] like Figure 2 Secondly, the present invention provides a personalized running training service system, including an acquisition module, a segmentation module, an extraction module, and an analysis module;

[0147] The acquisition module is used to acquire video frame sequences and speed sequences of an athlete exercising on a treadmill;

[0148] The segmentation module is used to segment a video frame sequence based on a velocity sequence to obtain multiple sub-video frame sequences;

[0149] The extraction module is used to extract the foreground from each video frame in each sub-video frame sequence to obtain the foreground image;

[0150] The analysis module is used to analyze each foreground image separately to obtain the analysis results of the athlete's running posture.

[0151] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method of the first aspect of the present invention.

[0152] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A personalized running training service method, characterized in that, include: S1, acquire the video frame sequence and speed sequence of the athlete exercising on the treadmill; S2, segment the video frame sequence based on the velocity sequence to obtain multiple sub-video frame sequences, including: First sub-video frame sequence It contains N video frames, where N is the preset number; For the nth sub-video frame sequence, where n is greater than or equal to 2, the process of determining the number of video frames contained in the nth sub-video frame sequence includes: Get the first frame in the (n-1)th sub-video frame sequence and the last frame , This represents the total number of video frames contained in the (n-1)th video frame; Obtain each and Foreground area in and ; based on and Calculate the first control parameter, including: Step 1, Calculation and The similarity between them is siml; The second step is to calculate. The center and The distance dm between the centers; The third step is to calculate the first control parameter: The first control parameter, As the second weight, Nz indicates that normalization is performed; Obtain the sub-velocity sequence corresponding to the video frame in the (n-1)th sub-video frame from the velocity sequence; The second control parameter is calculated based on the sub-velocity sequence, including: The first step is to obtain the weighted rotational velocity based on the sub-velocity sequence, including: Calculate the weighted rotational velocity using the following formula: For weighted rotational speed, Let be the i-th rotational velocity in the sub-velocity sequence, and NS be the total number of rotational velocities in the sub-velocity sequence; The second step is to obtain the rotational speed fluctuation values ​​in the sub-velocity sequence, including: Obtain the linear regression line corresponding to the rotational speed of the sub-velocity sequence; The normalized value corresponding to the absolute value of the slope of the linear regression line is used as the rotational speed fluctuation value k. The third step is to calculate the second control parameter based on the weighted rotational speed and the rotational speed fluctuation value, including: This is the second control parameter. This represents the maximum rotational speed in the sub-velocity sequence. It is the third weight; The number of video frames contained in the nth sub-video frame sequence is calculated based on the first control parameter and the second control parameter. S3, extract the foreground image from the video frames in each sub-video frame sequence; S4 analyzes each foreground image to obtain the analysis results of the athlete's running posture.

2. The personalized running training service method according to claim 1, characterized in that, The velocity sequence includes a time series of the rotational speed of the treadmill belt.

3. The personalized running training service method according to claim 1, characterized in that, The number of video frames in a video frame sequence is the same as the number of rotational speeds in a velocity sequence; Each video frame in the video frame sequence corresponds to a rotation speed in the velocity sequence. The process of obtaining the rotation speed is as follows: For video frames generated at time t ,but The corresponding rotational speed is , Let be the rotational speed of the treadmill belt at time t.

4. The personalized running training service method according to claim 1, characterized in that, Obtain each and Foreground area in and ,include: For video frames , ∈{ , }, The process of obtaining the foreground region is as follows: Use the inter-frame difference algorithm to obtain The first area in; The second region is obtained by performing region growing on the edge pixels of the first region; The second region is then filtered to obtain the foreground region.

5. The personalized running training service method according to claim 4, characterized in that, Use the inter-frame difference algorithm to obtain The first region includes: like If the first frame in the video frame sequence is used, then the second frame in the video frame sequence is used as the comparison frame. Otherwise, obtain from the video frame sequence. The previous video frame is used as the comparison frame. ; based on Calculate separately The grayscale difference of each pixel in the image; Will Pixels with a grayscale difference greater than the adaptive grayscale threshold are selected as pixels in the first region.

6. The personalized running training service method according to claim 5, characterized in that, Calculate separately The grayscale difference of each pixel in the data includes: for The coordinates in are pixels The formula for calculating the grayscale difference is: for The difference in grayscale values, for grayscale value, for grayscale value, for The median coordinate is The pixels.

7. The personalized running training service method according to claim 4, characterized in that, The second region is obtained by performing region growing on the edge pixels of the first region, including: Obtain the set A of pixels located at the edge of the first region; exist In this process, each pixel in A is used as a seed point for region growing to obtain the second region.

8. The personalized running training service method according to claim 1, characterized in that, The value of N is 150.

9. A personalized running training service system, characterized in that, It includes an acquisition module, a segmentation module, an extraction module, and an analysis module; The acquisition module is used to acquire video frame sequences and speed sequences of an athlete exercising on a treadmill; The segmentation module is used to segment the video frame sequence based on the velocity sequence to obtain multiple sub-video frame sequences, including: First sub-video frame sequence It contains N video frames, where N is the preset number; For the nth sub-video frame sequence, where n is greater than or equal to 2, the process of determining the number of video frames contained in the nth sub-video frame sequence includes: Get the first frame in the (n-1)th sub-video frame sequence and the last frame , This represents the total number of video frames contained in the (n-1)th video frame; Obtain each and Foreground area in and ; based on and Calculate the first control parameter, including: Step 1, Calculation and The similarity between them is siml; The second step is to calculate. The center and The distance dm between the centers; The third step is to calculate the first control parameter: The first control parameter, As the second weight, Nz indicates that normalization is performed; Obtain the sub-velocity sequence corresponding to the video frame in the (n-1)th sub-video frame from the velocity sequence; The second control parameter is calculated based on the sub-velocity sequence, including: The first step is to obtain the weighted rotational velocity based on the sub-velocity sequence, including: Calculate the weighted rotational velocity using the following formula: For weighted rotational speed, Let be the i-th rotational velocity in the sub-velocity sequence, and NS be the total number of rotational velocities in the sub-velocity sequence; The second step is to obtain the rotational speed fluctuation values ​​in the sub-velocity sequence, including: Obtain the linear regression line corresponding to the rotational speed of the sub-velocity sequence; The normalized value corresponding to the absolute value of the slope of the linear regression line is used as the rotational speed fluctuation value k. The third step is to calculate the second control parameter based on the weighted rotational speed and the rotational speed fluctuation value, including: This is the second control parameter. This represents the maximum rotational speed in the sub-velocity sequence. It is the third weight; The number of video frames contained in the nth sub-video frame sequence is calculated based on the first control parameter and the second control parameter. The extraction module is used to extract the foreground from each video frame in each sub-video frame sequence to obtain the foreground image; The analysis module is used to analyze each foreground image separately to obtain the analysis results of the athlete's running posture.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-8.

Images