Multi-angle tennis track analysis method applied to tennis training

Abstract

The invention discloses a multi-angle tennis track analysis method applied to tennis training, and relates to the technical field of tennis training, a training field is divided into a plurality of sub-regions, opposite high-speed cameras are arranged in each sub-region, and image data of each sub-region are acquired through the arranged high-speed cameras; constructing a corresponding training field model, obtaining a tennis flight path and a flight attitude according to the obtained image data of each sub-region, mapping the obtained tennis flight path and flight attitude into the training field model, and obtaining a tennis flight simulation image; according to the obtained tennis ball flight simulation image, the flight path and the flight attitude of the tennis ball are evaluated, and the serving quality of a trainee is judged according to the evaluation result, so that the serving quality of an athlete is effectively evaluated, and the athlete is assisted in correspondingly adjusting the serving mode of the athlete according to the evaluation result.

Description

Technical Field

[0001] The invention relates to the technical field of tennis training, in particular to a multi-angle tennis trajectory analysis method applied to tennis training. Background Art

[0002] Tennis is a highly technical and physically demanding sport that requires athletes to make quick and precise movements in a short period of time. In order to improve the competitive level, tennis training requires not only professional technical guidance, but also scientific training methods and advanced training equipment;

[0003] In the existing technology, automatic serving machines are usually used to train athletes' ball receiving ability and reaction ability. In tennis matches, the athlete's serving level often plays a huge role in the outcome of the game. For serving, it depends entirely on the athlete himself. How to effectively evaluate the quality of the athlete's serve, so as to help the athlete adjust his serving method and improve the quality of the serve, is a problem we need to solve. For this purpose, a multi-angle tennis trajectory analysis method for tennis training is provided. Summary of the invention

[0004] The object of the present invention is to provide a multi-angle tennis trajectory analysis method for tennis training.

[0005] The purpose of the present invention can be achieved by the following technical solution: A multi-angle tennis trajectory analysis method applied to tennis training comprises the following steps:

[0006] Step S1: Divide the training ground into several sub-areas, arrange opposite high-speed cameras in each sub-area, and obtain image data of each sub-area through the arranged high-speed cameras;

[0007] Step S2: construct a corresponding training venue model, obtain the flight trajectory and flight posture of the tennis ball according to the image data of each sub-area obtained, map the obtained flight trajectory and flight posture of the tennis ball into the training venue model, and obtain a simulated image of the tennis ball flight;

[0008] Step S3: Evaluate the flight trajectory and flight posture of the tennis ball based on the obtained tennis ball flight simulation image, and judge the trainee's serve quality based on the evaluation result.

[0009] Furthermore, the training ground includes a serving area, a first half court area, a second half court area, and a receiving area;

[0010] Two high-speed cameras are placed in the serving area, the first half court area, the second half court area and the receiving area;

[0011] The two high-speed cameras arranged in the serving area are recorded as the first camera group;

[0012] The two high-speed cameras arranged in the first half area are recorded as the second camera group;

[0013] The two high-speed cameras arranged in the second half area are recorded as the third camera group;

[0014] The two high-speed cameras arranged in the ball receiving area are recorded as the fourth camera group;

[0015] The image data of the corresponding sub-area is obtained by a high-speed camera arranged in each sub-area.

[0016] Furthermore, the process of constructing the corresponding training site model includes:

[0017] Construct a three-dimensional coordinate system, and generate a corresponding virtual field model in the three-dimensional coordinate system according to the training field;

[0018] The training personnel and tennis balls in the training ground are identified through image data obtained by a high-speed camera arranged in the serving area, and corresponding virtual personnel images and virtual tennis balls are generated based on the identified training personnel and tennis balls, and the virtual personnel images and virtual tennis balls are mapped to corresponding positions of the virtual field model.

[0019] Further, according to the image data of each sub-area obtained, the flight trajectory and flight posture of the tennis ball are obtained, and the obtained flight trajectory and flight posture of the tennis ball are mapped to the training venue model. The process of obtaining the simulated image of the tennis ball flight includes:

[0020] After entering the training ground, the training personnel start each high-speed camera, and obtain image data of the serving area through the first camera group arranged in the serving area, and record the image data obtained by the first camera group as the starting image data group;

[0021] Performing human figure recognition in the serving area according to the obtained start-up image data group, and judging whether there is a training person in the serving area according to the human figure recognition result;

[0022] If there is a training person in the serving area, the image data of the corresponding sub-areas are obtained by the first camera group, the second camera group, the third camera group and the fourth camera group, respectively, and are recorded as the first image data group, the second image data group, the third image data group and the fourth image data group;

[0023] Identify the preprocessed tennis ball in the obtained first image data group, and obtain flight parameters of the tennis ball according to the identification result, wherein the flight parameters include a serving point, a serving speed, a rotation angle, a rotation speed, a flight position, and a flight speed;

[0024] generating a tennis ball flight trajectory of the tennis ball according to the obtained initial position and flight position of the tennis ball;

[0025] Generate the flight posture of the tennis ball according to the rotation angle and rotation speed of the tennis ball;

[0026] The generated tennis ball flight trajectory and flight posture are mapped to the training venue model to obtain a tennis ball flight simulation image.

[0027] Furthermore, before using the tennis ball for training, the tennis ball needs to be pre-processed, that is, six circular areas of different colors that are different from the color of the tennis ball are marked on the surface of the tennis ball, and the circular areas are recorded as identification areas.

[0028] Furthermore, the process of identifying the pre-processed tennis ball in the obtained first image data includes:

[0029] Converting the image data in the first image data group into image frames, and respectively summarizing the image frames converted from the image data taken by different high-speed cameras to obtain corresponding image frame groups;

[0030] The obtained image frames are sampled at a fixed frequency to obtain a corresponding sampled image frame group, each image frame in the sampled image frame group is sequentially rasterized at a corresponding shooting time, and the sampled image frames after the rasterization are converted into corresponding grayscale images;

[0031] The obtained grayscale image is input into the trained convolutional neural network model, and features are extracted from the input grayscale image. It is determined whether there is a tennis ball in the grayscale image based on the extracted features. If there is a tennis ball, the area where the corresponding feature is located is calibrated, and the boundary of the calibrated area is obtained;

[0032] Generate tennis ball labels from the obtained boundaries, and associate the tennis ball labels with the calibration areas;

[0033] Taking the obtained calibration area as the reference image, mapping the reference image to the corresponding position in the grayscale image, and intercepting the position corresponding to the reference image in the grayscale image to obtain the image area to be identified;

[0034] Perform secondary feature extraction on the intercepted image area to be identified, mark the identification area in the image area to be identified according to the extracted features, map the identified identification area to the corresponding position in the reference image, obtain the corresponding tennis ball identification image, and obtain the flight posture of the tennis ball at the position according to the tennis ball identification image;

[0035] Acquire all image frames with tennis ball recognition images in the sampled image frame group, and mark the image frame with the earliest shooting time as a reference image frame, and mark the image frame adjacent to the reference image frame and without the tennis ball recognition image as a critical image frame;

[0036] All image frames between the critical image frame and the reference image frame are identified in chronological order, and the image frame in which the tennis ball is first identified is marked as the starting frame;

[0037] The tennis ball recognition image in each image frame is obtained, and the tennis ball recognition image is mapped to the corresponding position in the training venue model, and each tennis ball recognition image is connected in sequence to obtain the flight trajectory of the tennis ball.

[0038] Furthermore, the process of evaluating the flight trajectory and flight posture of the tennis ball according to the obtained tennis ball flight simulation image includes:

[0039] Before training, the training staff sets the tennis target landing area on the training venue model;

[0040] Obtain the change in the position of the tennis ball between the start frame and the next image frame, obtain the flying speed of the tennis ball, and so on, obtain the flying speed corresponding to the tennis ball in each image frame;

[0041] Constructing a two-dimensional coordinate system of time with respect to flight speed, obtaining the flight speed of the tennis ball corresponding to each position of the flight trajectory of the tennis ball, and generating a corresponding speed change curve in the two-dimensional coordinate system;

[0042] Traversing the speed change curve according to the time axis of the two-dimensional coordinate system, obtaining the tangent of the speed change curve, and obtaining the tangent slope of the speed change curve;

[0043] Set a slope threshold, and compare the obtained tangent slope with the slope threshold;

[0044] The earliest position on the speed change curve where the tangent slope is greater than the slope threshold is marked as the serve hitting point, and the flight speed corresponding to the serve hitting point is recorded as the serve speed;

[0045] Obtaining an image frame corresponding to the serve hitting point and a next image frame, which are recorded as a first image frame and a second image frame respectively;

[0046] Obtaining the position of the center point of the tennis ball in the tennis ball recognition image in the first image frame and the second image frame, connecting the center points to generate a corresponding position change vector, and recording the direction corresponding to the position change vector as the serving direction;

[0047] Obtaining a rotation speed and a rotation angle of the tennis ball at a serving point according to a change in a marked area in a flying posture of the tennis ball in the first image frame and the second image frame;

[0048] Get the location of the tennis ball, and get the shortest distance between the tennis ball's location and the set tennis ball's target location;

[0049] According to the flight trajectory and flight posture of the tennis ball between the hitting point and the landing point, the corresponding serve quality evaluation coefficient Fp is obtained;

[0050] Set the evaluation threshold interval, denoted as F0;

[0051] When Fp≤F0, it means that the serve quality is unqualified, and a training correction plan is generated;

[0052] When Fp>F0, it means the serve quality is qualified.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] By placing high-speed cameras in each area of ​​the training venue to film the trainees' serving process, the trainees' serving speed, serving direction, rotation, flight trajectory, and the deviation between the landing point and the target position can be obtained to evaluate the trainees' serve quality. As a result, after completing a serve, the trainees can quickly obtain the complete flight trajectory and related flight parameters of the tennis ball, thereby effectively evaluating the athletes' serve quality, and then assisting the athletes to make corresponding adjustments to their serving methods based on the evaluation results, thereby improving their serving ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] 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.

[0056] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0057] like Figure 1 As shown, the multi-angle tennis trajectory analysis method applied to tennis training includes the following steps:

[0058] Step S1: Divide the training ground into several sub-areas, arrange opposite high-speed cameras in each sub-area, and obtain image data of each sub-area through the arranged high-speed cameras;

[0059] Step S2: construct a corresponding training venue model, obtain the flight trajectory and flight posture of the tennis ball according to the image data of each sub-area obtained, map the obtained flight trajectory and flight posture of the tennis ball into the training venue model, and obtain a simulated image of the tennis ball flight;

[0060] Step S3: Evaluate the flight trajectory and flight posture of the tennis ball based on the obtained tennis ball flight simulation image, and judge the trainee's serve quality based on the evaluation result.

[0061] It should be further explained that, in the specific implementation process, the training ground is divided into several sub-areas, including a serving area, a first half area, a second half area, and a receiving area;

[0062] Two high-speed cameras are placed in the serving area, the first half court area, the second half court area and the receiving area;

[0063] The two high-speed cameras arranged in the serving area are recorded as the first camera group;

[0064] The two high-speed cameras arranged in the first half area are recorded as the second camera group;

[0065] The two high-speed cameras arranged in the second half area are recorded as the third camera group;

[0066] The two high-speed cameras arranged in the ball receiving area are recorded as the fourth camera group;

[0067] The image data of the corresponding sub-area is obtained by a high-speed camera arranged in each sub-area.

[0068] It should be further explained that, in the specific implementation process, the process of building the corresponding training venue model includes:

[0069] Construct a three-dimensional coordinate system, and generate a corresponding virtual field model in the three-dimensional coordinate system according to the training field;

[0070] The training personnel and tennis balls in the training ground are identified through image data obtained by a high-speed camera arranged in the serving area, and corresponding virtual personnel images and virtual tennis balls are generated based on the identified training personnel and tennis balls, and the virtual personnel images and virtual tennis balls are mapped to corresponding positions of the virtual field model.

[0071] It should be further explained that, in the specific implementation process, the flight trajectory and flight posture of the tennis ball are obtained according to the image data of each sub-area, and the obtained flight trajectory and flight posture of the tennis ball are mapped to the training venue model. The process of obtaining the simulated image of the tennis ball flight includes:

[0072] After the training personnel enter the training ground, they start each high-speed camera, and obtain image data of the serving area through the first camera group arranged in the serving area, and record the image data obtained by the first camera group as the start-up image data group; it should be further explained that in the specific implementation process, after each high-speed camera is started, only the two high-speed cameras of the first camera group arranged in the serving area will shoot the serving area, and other high-speed cameras will not shoot. Through the recognition result of the image data shot by the high-speed camera in the serving area, the shooting amount of the high-speed cameras in other areas can be reduced, and the use efficiency of the high-speed cameras can be indirectly improved;

[0073] According to the obtained start-up image data group, human figure recognition is performed in the serving area, and whether there is a training person in the serving area is determined according to the human figure recognition result; it should be further explained that, in the specific implementation process, the human figure recognition is to complete the recognition process using the trained convolutional neural network model, which is a commonly used existing technical means by those skilled in the art and will not be described in detail here;

[0074] If there is a training person in the serving area, the image data of the corresponding sub-areas are obtained by the first camera group, the second camera group, the third camera group and the fourth camera group, respectively, and are recorded as the first image data group, the second image data group, the third image data group and the fourth image data group;

[0075] Identify the preprocessed tennis ball in the obtained first image data group, and obtain flight parameters of the tennis ball according to the identification result, wherein the flight parameters include a serving point, a serving speed, a rotation angle, a rotation speed, a flight position, and a flight speed;

[0076] generating a tennis ball flight trajectory of the tennis ball according to the obtained initial position and flight position of the tennis ball;

[0077] Generate the flight posture of the tennis ball according to the rotation angle and rotation speed of the tennis ball;

[0078] The generated tennis ball flight trajectory and flight posture are mapped to the training venue model to obtain a tennis ball flight simulation image.

[0079] It should be further explained that, in the specific implementation process, the tennis ball used for training needs to be pre-processed before use, that is, six circular areas of different colors, which are also different from the color of the tennis ball, are marked on the surface of the tennis ball, and the circular areas are recorded as identification areas.

[0080] It should be further explained that, in a specific implementation process, the process of identifying the pre-processed tennis ball in the obtained first image data includes:

[0081] Converting the image data in the first image data group into image frames, and respectively summarizing the image frames converted from the image data taken by different high-speed cameras to obtain corresponding image frame groups;

[0082] The obtained image frames are sampled at a fixed frequency to obtain a corresponding sampled image frame group, each image frame in the sampled image frame group is sequentially rasterized at a corresponding shooting time, and the sampled image frames after the rasterization are converted into corresponding grayscale images;

[0083] The obtained grayscale image is input into the trained convolutional neural network model, and features are extracted from the input grayscale image. It is determined whether there is a tennis ball in the grayscale image based on the extracted features. If there is a tennis ball, the area where the corresponding feature is located is calibrated, and the boundary of the calibrated area is obtained. It should be further explained that, in the specific implementation process, the process of determining whether there is a tennis ball in the grayscale image is to use the trained convolutional neural network model to complete the tennis ball recognition process. This process is a conventional technical means commonly used by those skilled in the art and will not be described in detail here.

[0084] Generate tennis ball labels from the obtained boundaries, and associate the tennis ball labels with the calibration areas;

[0085] Taking the obtained calibration area as the reference image, mapping the reference image to the corresponding position in the grayscale image, and intercepting the position corresponding to the reference image in the grayscale image to obtain the image area to be identified;

[0086] Perform secondary feature extraction on the intercepted image area to be identified, mark the identification area in the image area to be identified according to the extracted features, map the identified identification area to the corresponding position in the reference image, obtain the corresponding tennis ball identification image, and obtain the flight posture of the tennis ball at the position according to the tennis ball identification image;

[0087] Acquire all image frames with tennis ball recognition images in the sampled image frame group, and mark the image frame with the earliest shooting time as a reference image frame, and mark the image frame adjacent to the reference image frame and without the tennis ball recognition image as a critical image frame;

[0088] All image frames between the critical image frame and the reference image frame are identified in chronological order, and the image frame in which the tennis ball is first identified is marked as the starting frame;

[0089] The image data captured by the second camera group, the third camera group and the fourth camera group are intercepted according to the shooting time corresponding to the starting frame, that is, only the image data after the shooting time corresponding to the starting frame in the second camera group, the third camera group and the fourth camera group are intercepted, thereby reducing the processing amount of image data, optimizing the processing speed of image data, so that the evaluation result of the serving process can be obtained more quickly, and the evaluation efficiency is improved;

[0090] The tennis ball recognition image in each image frame is obtained, and the tennis ball recognition image is mapped to the corresponding position in the training venue model, and each tennis ball recognition image is connected in sequence to obtain the flight trajectory of the tennis ball.

[0091] It should be further explained that, in the specific implementation process, the process of evaluating the flight trajectory and flight posture of the tennis ball according to the obtained tennis ball flight simulation image includes:

[0092] Before training, the training staff sets the tennis target landing area on the training venue model;

[0093] Obtain the change in the position of the tennis ball between the start frame and the next image frame, obtain the flying speed of the tennis ball, and so on, obtain the flying speed corresponding to the tennis ball in each image frame;

[0094] Constructing a two-dimensional coordinate system of time with respect to flight speed, obtaining the flight speed of the tennis ball corresponding to each position of the flight trajectory of the tennis ball, and generating a corresponding speed change curve in the two-dimensional coordinate system;

[0095] Traversing the speed change curve according to the time axis of the two-dimensional coordinate system, obtaining the tangent of the speed change curve, and obtaining the tangent slope of the speed change curve;

[0096] Set a slope threshold, and compare the obtained tangent slope with the slope threshold;

[0097] The earliest position on the speed change curve where the tangent slope is greater than the slope threshold is marked as the serve hitting point, and the flight speed corresponding to the serve hitting point is recorded as the serve speed;

[0098] Obtaining an image frame corresponding to the serve hitting point and a next image frame, which are recorded as a first image frame and a second image frame respectively;

[0099] Obtaining the position of the center point of the tennis ball in the tennis ball recognition image in the first image frame and the second image frame, connecting the center points to generate a corresponding position change vector, and recording the direction corresponding to the position change vector as the serving direction;

[0100] Obtaining a rotation speed and a rotation angle of the tennis ball at a serving point according to a change in a marked area in a flying posture of the tennis ball in the first image frame and the second image frame;

[0101] Get the location of the tennis ball, and get the shortest distance between the tennis ball's location and the set tennis ball's target location;

[0102] According to the flight trajectory and flight posture of the tennis ball between the hitting point and the landing point, the corresponding serve quality evaluation coefficient is obtained, which is recorded as Fp;

[0103] in,

[0104] Among them, V c is the speed of the tennis ball at the serving point, V L is the flying speed of the tennis ball at the landing point, K x represents the correction factor, which depends on the rotation angle at the point of impact of the serve, V xc V is the rotation speed of the tennis ball at the point of service. xL represents the rotation speed of the tennis ball at the landing point, L represents the total length of the flight trajectory between the serving point and the landing point, and P p Indicates the landing point deviation distance, that is, the shortest distance between the tennis ball landing point and the set tennis ball target landing area, α is the proportional coefficient, and α>1;

[0105] Set the evaluation threshold interval, denoted as F0;

[0106] When Fp≤F0, it means that the serve quality is unqualified, and a training correction plan is generated;

[0107] When Fp>F0, it means the serve quality is qualified.

[0108] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments with equivalent changes without departing from the scope of the technical solution of the present invention. However, any modification or equivalent replacement of the above embodiments made according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A multi-angle tennis trajectory analysis method for tennis training, characterized in that: The following steps are involved: Step S1: Divide the training ground into several sub-areas, arrange opposite high-speed cameras in each sub-area, and obtain image data of each sub-area through the arranged high-speed cameras; Step S2: construct a corresponding training venue model, obtain the flight trajectory and flight posture of the tennis ball according to the image data of each sub-area obtained, map the obtained flight trajectory and flight posture of the tennis ball into the training venue model, and obtain a simulated image of the tennis ball flight; Step S3: Evaluate the flight trajectory and flight posture of the tennis ball based on the obtained tennis ball flight simulation image, and judge the trainee's serve quality based on the evaluation result.

2. The multi-angle tennis trajectory analysis method for tennis training according to claim 1, characterized in that: The training venue includes a serving area, a first half court area, a second half court area and a receiving area; Two high-speed cameras are placed in the serving area, the first half court area, the second half court area and the receiving area; The two high-speed cameras arranged in the serving area are recorded as the first camera group; The two high-speed cameras arranged in the first half area are recorded as the second camera group; The two high-speed cameras arranged in the second half area are recorded as the third camera group; The two high-speed cameras arranged in the ball receiving area are recorded as the fourth camera group; The image data of the corresponding sub-area is obtained by a high-speed camera arranged in each sub-area.

3. The multi-angle tennis trajectory analysis method for tennis training according to claim 2, characterized in that: The process of building the corresponding training site model includes: Construct a three-dimensional coordinate system, and generate a corresponding virtual field model in the three-dimensional coordinate system according to the training field; The training personnel and tennis balls in the training ground are identified through image data obtained by a high-speed camera arranged in the serving area, and corresponding virtual personnel images and virtual tennis balls are generated based on the identified training personnel and tennis balls, and the virtual personnel images and virtual tennis balls are mapped to corresponding positions of the virtual field model.

4. The multi-angle tennis trajectory analysis method for tennis training according to claim 3, characterized in that: The process of obtaining the flight trajectory and flight posture of the tennis ball according to the image data of each sub-area obtained, and mapping the obtained flight trajectory and flight posture of the tennis ball into the training field model to obtain the simulated image of the tennis ball flight includes: After entering the training ground, the training personnel start each high-speed camera, and obtain image data of the serving area through the first camera group arranged in the serving area, and record the image data obtained by the first camera group as the starting image data group; Performing human figure recognition in the serving area according to the obtained start-up image data group, and judging whether there is a training person in the serving area according to the human figure recognition result; If there is a training person in the serving area, the image data of the corresponding sub-areas are obtained by the first camera group, the second camera group, the third camera group and the fourth camera group, respectively, and are recorded as the first image data group, the second image data group, the third image data group and the fourth image data group; Identify the preprocessed tennis ball in the obtained first image data group, and obtain flight parameters of the tennis ball according to the identification result, wherein the flight parameters include a serving point, a serving speed, a rotation angle, a rotation speed, a flight position, and a flight speed; generating a tennis ball flight trajectory of the tennis ball according to the obtained initial position and flight position of the tennis ball; Generate the flight posture of the tennis ball according to the rotation angle and rotation speed of the tennis ball; The generated tennis ball flight trajectory and flight posture are mapped to the training venue model to obtain a tennis ball flight simulation image.

5. The multi-angle tennis trajectory analysis method for tennis training according to claim 4, characterized in that: Before using the tennis ball for training, it is necessary to pre-process the tennis ball, that is, mark six circular areas of different colors on the surface of the tennis ball, which are also different from the color of the tennis ball, and record the circular areas as identification areas.

6. The multi-angle tennis trajectory analysis method for tennis training according to claim 5, characterized in that: The process of identifying the pre-processed tennis ball in the obtained first image data includes: Converting the image data in the first image data group into image frames, and respectively summarizing the image frames converted from the image data taken by different high-speed cameras to obtain corresponding image frame groups; The obtained image frames are sampled at a fixed frequency to obtain a corresponding sampled image frame group, each image frame in the sampled image frame group is sequentially rasterized at a corresponding shooting time, and the sampled image frames after the rasterization are converted into corresponding grayscale images; The obtained grayscale image is input into the trained convolutional neural network model, and features are extracted from the input grayscale image. It is determined whether there is a tennis ball in the grayscale image based on the extracted features. If there is a tennis ball, the area where the corresponding feature is located is calibrated, and the boundary of the calibrated area is obtained; Generate tennis ball labels from the obtained boundaries, and associate the tennis ball labels with the calibration areas; Taking the obtained calibration area as the reference image, mapping the reference image to the corresponding position in the grayscale image, and intercepting the position corresponding to the reference image in the grayscale image to obtain the image area to be identified; Perform secondary feature extraction on the intercepted image area to be identified, mark the identification area in the image area to be identified according to the extracted features, map the identified identification area to the corresponding position in the reference image, obtain the corresponding tennis ball identification image, and obtain the flight posture of the tennis ball at the position according to the tennis ball identification image; Acquire all image frames with tennis ball recognition images in the sampled image frame group, and mark the image frame with the earliest shooting time as a reference image frame, and mark the image frame adjacent to the reference image frame and without the tennis ball recognition image as a critical image frame; All image frames between the critical image frame and the reference image frame are identified in chronological order, and the image frame in which the tennis ball is first identified is marked as the starting frame; The tennis ball recognition image in each image frame is obtained, and the tennis ball recognition image is mapped to the corresponding position in the training venue model, and each tennis ball recognition image is connected in sequence to obtain the flight trajectory of the tennis ball.

7. The multi-angle tennis trajectory analysis method for tennis training according to claim 6, characterized in that: Based on the obtained tennis ball flight simulation image, the process of evaluating the flight trajectory and flight posture of the tennis ball includes: Before training, the training staff sets the tennis target landing area on the training venue model; Obtain the change in the position of the tennis ball between the start frame and the next image frame, obtain the flying speed of the tennis ball, and so on, obtain the flying speed corresponding to the tennis ball in each image frame; Constructing a two-dimensional coordinate system of time with respect to flight speed, obtaining the flight speed of the tennis ball corresponding to each position of the flight trajectory of the tennis ball, and generating a corresponding speed change curve in the two-dimensional coordinate system; Traversing the speed change curve according to the time axis of the two-dimensional coordinate system, obtaining the tangent of the speed change curve, and obtaining the tangent slope of the speed change curve; Set a slope threshold, and compare the obtained tangent slope with the slope threshold; The earliest position on the speed change curve where the tangent slope is greater than the slope threshold is marked as the serving point, and the flight speed corresponding to the serving point is recorded as the serving speed; Obtaining an image frame corresponding to the serve hitting point and a next image frame, which are recorded as a first image frame and a second image frame respectively; Obtaining the position of the center point of the tennis ball in the tennis ball recognition image in the first image frame and the second image frame, connecting the center points to generate a corresponding position change vector, and recording the direction corresponding to the position change vector as the serving direction; Obtaining a rotation speed and a rotation angle of the tennis ball at a serving point according to a change in a marked area in a flight posture of the tennis ball in the first image frame and the second image frame; Get the location of the tennis ball, and get the shortest distance between the tennis ball's location and the set tennis ball's target location; According to the flight trajectory and flight posture of the tennis ball between the hitting point and the landing point, the corresponding serve quality evaluation coefficient Fp is obtained; Set the evaluation threshold interval, denoted as F0; When Fp≤F0, it means that the serve quality is unqualified, and a training correction plan is generated; When Fp>F0, it means the serve quality is qualified.

Images