Basketball examination system based on AI vision

Abstract

The invention relates to the technical field of AI vision and object recognition, and provides an AI vision-based basketball examination system, which is characterized in that an image acquisition module transmits acquired image information to an information processing module, and the information processing module simultaneously transmits processed information to a basketball recognition module and an action recognition module; the basketball recognition module transmits recognized basketball information to the comprehensive analysis module, the action recognition module transmits recognized examinee action information to the comprehensive analysis module, and the comprehensive analysis module transmits generated score evaluation and training suggestions to the display module for display; a powerful basis is provided for subsequent accurate recognition and analysis through the information processing module, and accurate analysis and scoring of examinee actions are achieved through the basketball recognition module and the action recognition module; the comprehensive analysis module provides instant and comprehensive feedback for examinees and coaches.

Description

Technical Field

[0001] The present invention relates to the technical field of AI vision and object recognition, and in particular to a basketball examination system based on AI vision. Background Art

[0002] As a global sport, basketball is widely loved and participated in. There are a large number of basketball enthusiasts and participants in various fields such as schools, communities, and sports clubs. In order to evaluate and improve the skill level of basketball enthusiasts, basketball exams came into being.

[0003] In school physical education, basketball is one of the important teaching contents. Through the examination, students' learning outcomes in basketball courses can be tested, and students can be encouraged to actively participate in physical exercise and improve their physical fitness and sports skills. At the same time, it also helps teachers understand the teaching effect and adjust teaching methods and strategies. In the physical education examinations for primary and secondary schools, the basketball dribbling and shooting items in the physical education test require students to dribble a certain distance from a fixed starting point according to standard movements, and then return to the starting point after shooting and scoring, and the time ends.

[0004] AI vision, or artificial intelligence vision technology, has gradually emerged based on the continuous development of computer technology, image processing technology, pattern recognition technology, and artificial intelligence algorithms. There are currently no intelligent, automated examination devices and systems in the existing technology, and all judgments are made on-site by examination personnel, or referees.

[0005] The existing background technology still has the following shortcomings:

[0006] 1. The subjective judgment of the examiners may be biased, resulting in inaccurate and unfair test results. Different referees may have different understandings and grasps of the movement norms and technical standards in the basketball test, which means that the same performance may receive different scores, affecting the authority and credibility of the test.

[0007] 2. The manual judging method is inefficient and difficult to cope with large-scale examination needs. When there are a large number of students, manual judging one by one will consume a lot of time and manpower, increasing the difficulty and cost of organizing the examination.

[0008] 3. Manually recording test data is prone to errors and omissions, which is not conducive to comprehensive and accurate analysis and summary of students' learning outcomes. It is impossible to provide effective data support for subsequent teaching improvements and students' personalized training; students can only know their scores and performance after the test, and cannot make timely adjustments and improvements during the test, which is not conducive to the rapid improvement of skills. Summary of the invention

[0009] In order to solve the above technical problems, the present invention provides a basketball examination system based on AI vision to solve the problems in the prior art.

[0010] The technical solution of the present invention is as follows:

[0011] A basketball examination system based on AI vision, comprising:

[0012] An image acquisition module, which is used to collect image information of the examinee during the basketball test;

[0013] An information processing module, the information processing module is used to establish a three-dimensional coordinate system according to the information collected by the image acquisition module, and generate the movement of the basketball and the examinee in the three-dimensional coordinate system during the examinee's basketball test;

[0014] A basketball recognition module, the basketball recognition module is used to recognize basketball information in the three-dimensional coordinate system established by the information processing module and recognize the movement trajectory of the basketball machine;

[0015] A motion recognition module, which is used to recognize the examinee's motion information in the three-dimensional coordinate system established by the information processing module, and to recognize and analyze the examinee's motion information;

[0016] A comprehensive analysis module, which is used to perform performance evaluation and generate training suggestions based on the identified basketball movement information and the examinee's movement information;

[0017] A display module, which is used to display the analysis results in real time;

[0018] The functional interaction relationship between the modules of the basketball examination system based on AI vision is as follows:

[0019] The image acquisition module transmits the collected image information to the information processing module, the information processing module transmits the processed information to the basketball recognition module and the action recognition module at the same time, the basketball recognition module transmits the recognized basketball information to the comprehensive analysis module, the action recognition module transmits the recognized candidate action information to the comprehensive analysis module, and the comprehensive analysis module transmits the generated performance evaluation and training suggestions to the display module for display.

[0020] Preferably, the image acquisition module includes a multi-module camera; the multi-module camera is used to capture test images of the examinee during the basketball test, and the test image content includes but is not limited to the running trajectory of the basketball and the examinee's movements.

[0021] Preferably, the information processing module establishes a three-dimensional coordinate system, places the information collected by the image acquisition module in the three-dimensional coordinate system, converts the change information in the image into coordinate information in the three-dimensional coordinate system, and retains the color of the relevant objects.

[0022] Preferably, the basketball recognition module includes a feature extraction unit, a basketball model unit and a recognition unit;

[0023] The feature extraction unit is used to extract feature information from basketball pictures of different angles, types and colors to provide a basis for model establishment;

[0024] The basketball model unit establishes a basketball recognition model by acquiring feature information of basketball pictures of different angles, types and colors extracted by the feature extraction unit;

[0025] The recognition unit inputs the three-dimensional coordinate system image data transmitted in real time by the information processing module into the basketball recognition model for recognition, thereby confirming the movement trajectory of the basketball.

[0026] Preferably, the feature extraction unit collects a specified number of basketball pictures of different angles, types and colors, as well as pictures of non-basketball objects as negative samples; performs operations such as cropping and scaling on the pictures to unify the picture size to M×N pixels;

[0027] Extract color features: Calculate the color mean of each image, and the coordinates of each pixel are (x, y); for color images, convert them to hue, saturation, and brightness color space, and calculate the mean of the three channels of hue H, saturation S, and brightness color V, that is, the color mean, and the formula is:

[0028]

[0029] Extract shape features: calculate the roundness B and area A of the object in the image,

[0030] The edge pixel coordinates of the object are (x i ,y i ), i is a set, meaning the i-th coordinate point, the number is u;

[0031]

[0032]

[0033] Circularity B = 4πA / O 2 .

[0034] Preferably, the basketball model unit establishes a basketball recognition model, firstly, the features extracted by the feature extraction unit and the corresponding labels, basketball or non-basketball, form a training data set;

[0035]

[0036] Among them, 0 represents non-basketball, 1 represents basketball, Where θ is the parameter vector of the model and p is the input feature set;

[0037] The model parameters θ are optimized by minimizing the loss function:

[0038]

[0039] The optimal θ is solved by the gradient descent optimization algorithm, and the formula is:

[0040]

[0041] in, Represents the partial derivative of the loss function L(θ) with respect to the jth parameter of parameter θ, which is used to update parameters in optimization algorithms such as gradient descent;

[0042] p i@ Represents the feature p of the i-th sample i The jth parameter of is taken and the most suitable model parameters are substituted into the model function to obtain the trained optimal basketball recognition model.

[0043] Preferably, the recognition unit extracts color and shape features from the three-dimensional coordinate system image data transmitted in real time in the feature extraction unit;

[0044] Input the extracted features into the trained model to get the output value H / (p i );

[0045] If H / (p i )>0.5, it is judged that there is a basketball in the image; otherwise, it is judged that there is no basketball;

[0046] For a continuous image sequence, the above recognition process is performed sequentially in chronological order;

[0047] In order to determine the trajectory of the basketball, the displacement vector between the basketball coordinates (x1, y1, z1) and (x2, y2, z2) identified in two adjacent frames can be calculated. The motion trajectory of the basketball is approximated by the displacement vectors between multiple consecutive frames.

[0048] Preferably, the action recognition module includes an action model unit and an action analysis unit;

[0049] The action model unit is used to obtain standard action information in a basketball test from, including but not limited to, basketball test specifications, and establish coordinate points of human joints in the standard action in a three-dimensional coordinate system;

[0050] The motion analysis unit is used to calculate the final test score of the examinee based on the coordinate information of the human joint points of the standard motion in the motion model unit, using human body recognition technology and the positions of the starting point and the end point.

[0051] Preferably, the action analysis unit receives the three-dimensional coordinate system image data transmitted in real time, including the joint point information and the starting point and the basketball hoop coordinate information, and calculates the candidate's violation and action score in combination with the dribbling action standard;

[0052] The coordinates of the human joint points in the kth frame are (x G,i ,y G,i ,z G,i ), i is a set, meaning the i-th coordinate point, the number is u; is the coordinate of the i-th joint point in the k-th frame;

[0053] The starting point of the test is the coordinate f=(x H ,y H ,z H ), the end point coordinate g=(x I ,y I ,z I ); Input the human joint point coordinate sequence of each frame into the action model unit for recognition and comparison to obtain the corresponding accuracy;

[0054] The accuracy is measured by measuring the distance between two point sets. In the action sequence, the set of joint point coordinates is regarded as a point set for calculation;

[0055] The collected action joint point coordinate set is α, and the standard action joint point coordinate set is β. Then the one-way Hausdorff distance r from set α to set β is:

[0056] r(α,β)=max{α i ∈α}min{β i ∈β}w(α i ,β i )

[0057] For each coordinate point α in the set α i , find the point β closest to it in the set β i , calculate the maximum value of all these closest distances; measure the maximum case of the shortest distance from the point of set α to the set β, reflecting the maximum deviation of the points of set α to the set β as a whole;

[0058] where d(a,b) is some distance measure between point a and point b;

[0059] The distance w is

[0060] The similarity is measured based on the distance and the subject's action score is calculated:

[0061] Determine whether the action is illegal: If the distance gap violation judgment condition C(a_i) is met, the score of the action is 0.

[0062] Preferably, the comprehensive analysis module performs real-time scoring according to the action score identified by the action recognition module and the basketball trajectory identified by the basketball recognition module based on a relevant scoring mechanism.

[0063] Preferably, the display module generates a display interface to display the test results in real time, and the content of the display interface includes the test taker's score, action analysis, key points for deduction and improvement suggestions.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. The present invention can convert the basketball test image information of the examinee collected by the image acquisition module into coordinate information in a three-dimensional coordinate system through the information processing module, and fully present the movement of the basketball and the examinee in space, providing a strong foundation for subsequent accurate identification and analysis, which is more comprehensive and accurate than traditional plane image analysis.

[0066] 2. The present invention can accurately identify the basketball and its motion trajectory through the collaborative work of the feature extraction unit, basketball model unit and recognition unit in the basketball recognition module; through the action recognition module, the action model unit establishes standard action coordinate points, and the action analysis unit calculates the scores in combination with the action standards, thereby realizing accurate analysis and scoring of the examinee's actions.

[0067] 3. The present invention can generate performance evaluation and training suggestions based on basketball movement information and examinee action information through a comprehensive analysis module, and can display examinee scores, action analysis, deduction keys and improvement suggestions in real time through a display interface of a display module, providing examinees and coaches with immediate, comprehensive and targeted feedback, effectively improving the efficiency of test evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a work flow chart of the present invention.

[0069] Figure 2 It is a schematic diagram of information admission of examinees during examinations of the present invention.

[0070] Figure 3 It is a working flow chart of the basketball recognition module and the action recognition module of the present invention. DETAILED DESCRIPTION

[0071] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0072] like Figure 1-3 As shown, the present invention provides a basketball examination system based on AI vision, comprising:

[0073] An image acquisition module, which is used to collect image information of the examinee during the basketball test;

[0074] An information processing module, the information processing module is used to establish a three-dimensional coordinate system according to the information collected by the image acquisition module, and generate the movement of the basketball and the examinee in the three-dimensional coordinate system during the examinee's basketball test;

[0075] A basketball recognition module, the basketball recognition module is used to recognize basketball information in the three-dimensional coordinate system established by the information processing module and recognize the basketball motion trajectory;

[0076] A motion recognition module, which is used to recognize the examinee's motion information in the three-dimensional coordinate system established by the information processing module, and to recognize and analyze the examinee's motion information;

[0077] A comprehensive analysis module, which is used to perform performance evaluation and generate training suggestions based on the identified basketball movement information and the examinee's movement information;

[0078] A display module, which is used to display the analysis results in real time;

[0079] The functional interaction relationship between the system modules is as follows: the image acquisition module transmits the collected image information to the information processing module, the information processing module transmits the processed information to the basketball recognition module and the action recognition module at the same time, the basketball recognition module transmits the recognized basketball information to the comprehensive analysis module, the action recognition module transmits the recognized candidate action information to the comprehensive analysis module, and the comprehensive analysis module transmits the generated performance evaluation and training suggestions to the display module for display.

[0080] Example: Figure 1-3 As shown, in this embodiment, when the examinee is taking a basketball test, the multi-module camera in the image acquisition module captures the test image of the examinee during the basketball test, and the test image content includes the trajectory of the basketball and the examinee's movements.

[0081] The information processing module establishes a three-dimensional coordinate system, places the information collected by the image acquisition module in the three-dimensional coordinate system, converts the change information in the image into coordinate information in the three-dimensional coordinate system, and retains the color of the relevant objects.

[0082] Through the information processing module, the basketball test image information of the examinee collected by the image acquisition module can be converted into coordinate information in a three-dimensional coordinate system, fully presenting the movement of the basketball and the examinee in space, providing a strong foundation for subsequent accurate identification and analysis, which is more comprehensive and accurate than traditional plane image analysis.

[0083] The information processing module transmits the processed information to the basketball recognition module at the same time, and the basketball recognition module includes a feature extraction unit, a basketball model unit and a recognition unit;

[0084] The feature extraction unit is used to extract feature information from basketball pictures of different angles, types and colors to provide a basis for model establishment;

[0085] The basketball model unit establishes a basketball recognition model by acquiring feature information of basketball pictures of different angles, types and colors extracted by the feature extraction unit;

[0086] The recognition unit inputs the three-dimensional coordinate system image data transmitted in real time by the information processing module into the basketball recognition model for recognition and confirms the movement trajectory of the basketball.

[0087] The feature extraction unit collects a specified number of basketball pictures of different angles, types and colors, as well as some pictures of non-basketball objects as negative samples; performs operations such as cropping and scaling on the pictures to unify the picture size to M×N pixels;

[0088] Extract color features: Calculate the color mean of each image, and the coordinates of each pixel are (x, y); for color images, convert them to hue, saturation, and brightness color space, and calculate the mean of the three channels of hue H, saturation S, and brightness color V, that is, the color mean, and the formula is:

[0089]

[0090] Extract shape features: calculate the roundness B and area A of the object in the image,

[0091] The edge pixel coordinates of the object are (x i ,y i ), i is a set, meaning the i-th coordinate point, the number is u;

[0092]

[0093] Circularity B = 4πA / O 2 .

[0094] The basketball model unit establishes a basketball recognition model. First, the features extracted by the feature extraction unit and the corresponding labels, basketball or non-basketball, form a training data set;

[0095]

[0096] Among them, 0 represents non-basketball, 1 represents basketball, Where θ is the parameter vector of the model and p is the input feature set;

[0097] The model parameters θ are optimized by minimizing the loss function:

[0098]

[0099] The optimal θ is solved by the gradient descent optimization algorithm, and the formula is:

[0100]

[0101] in, Represents the partial derivative of the loss function L(θ) with respect to the jth parameter of parameter θ, which is used to update parameters in optimization algorithms such as gradient descent;

[0102] p i@ Represents the feature p of the i-th sample i The jth parameter of is taken and the most suitable model parameters are substituted into the model function to obtain the trained optimal basketball recognition model.

[0103] The recognition unit extracts color and shape features from the three-dimensional coordinate system image data transmitted in real time in the feature extraction unit;

[0104] Input the extracted features into the trained model to get the output value H / (p i );

[0105] If H / (p i )>0.5, it is judged that there is a basketball in the image; otherwise, it is judged that there is no basketball;

[0106] For a continuous image sequence, the above recognition process is performed sequentially in chronological order;

[0107] In order to determine the trajectory of the basketball, the displacement vector between the basketball coordinates (x1, y1, z1) and (x2, y2, z2) identified in two adjacent frames can be calculated. The motion trajectory of the basketball is approximated by the displacement vectors between multiple consecutive frames.

[0108] Through the coordinated work of the feature extraction unit, basketball model unit and recognition unit in the basketball recognition module, the basketball and its movement trajectory can be accurately identified; accurate analysis and scoring of the examinee's movements are achieved, reducing the inaccuracy of test results caused by deviations and improving the authority and credibility of the test.

[0109] The information processing module transmits the processed information to the action recognition module at the same time, and the action recognition module includes an action model unit and an action analysis unit;

[0110] The action model unit is used to obtain standard action information in basketball examinations from, including but not limited to, basketball examination specifications, and establish coordinate points of human joints in the standard actions in a three-dimensional coordinate system;

[0111] The motion analysis unit is used to calculate the final test scores of candidates based on the coordinate information of the human joints in the standard motions in the motion model unit, using human body recognition technology and the positions of the starting and ending points.

[0112] The motion analysis unit receives the real-time transmitted three-dimensional coordinate system image data, including joint point information, starting point and basketball hoop coordinate information, and calculates the candidate's violation and motion score in combination with the dribbling motion standard;

[0113] The coordinates of the human joint points in the kth frame are (x G,i ,y G,i ,z G,i ), i is a set, meaning the i-th coordinate point, the number is u; is the coordinate of the i-th joint point in the k-th frame;

[0114] The starting point of the test is the coordinate f=(x H ,y H ,z H ), the end point coordinate g=(x I ,y I ,z I ); Input the human joint point coordinate sequence of each frame into the action model unit for recognition and comparison to obtain the corresponding accuracy;

[0115] The accuracy is measured by measuring the distance between two point sets. In the action sequence, the set of joint point coordinates is regarded as a point set for calculation;

[0116] The collected action joint point coordinate set is α, and the standard action joint point coordinate set is β. Then the one-way Hausdorff distance r from set α to set β is:

[0117] r(α,β)=max{α i ∈α}min{β i ∈β}w(α i ,β i )

[0118] For each coordinate point α in the set α i , find the point β closest to it in the set β i, calculate the maximum value of all these closest distances; measure the maximum case of the shortest distance from the point of set α to the set β, reflecting the maximum deviation of the points of set α to the set β as a whole;

[0119] where d(a,b) is some distance measure between point a and point b;

[0120] The distance w is

[0121] The similarity is measured based on the distance and the subject's action score is calculated:

[0122] Through the motion recognition module, the motion model unit establishes standard motion coordinate points, and the motion analysis unit calculates the scores based on the motion standards, thereby achieving the scoring of the candidates' movements and improving the accuracy of the test calculations.

[0123] Determine whether the action is illegal: If the distance difference violation judgment condition C(a_i) is met, the score of the action is 0. The comprehensive analysis module performs real-time scoring based on the action score identified by the action recognition module and the basketball trajectory identified by the basketball recognition module according to the relevant scoring mechanism.

[0124] The display module generates a display interface to display the test results in real time. The display interface content includes the candidate's score, action analysis, key points for deduction and improvement suggestions.

[0125] The comprehensive analysis module can be used to generate performance evaluation and training suggestions based on basketball movement information and candidate movement information. The display module's display interface can display candidate scores, movement analysis, key points for deductions and improvement suggestions in real time, providing candidates and coaches with immediate, comprehensive and targeted feedback, improving the efficiency of test evaluation, reducing labor costs and providing guiding value.

[0126] The embodiments of the present invention are provided for the purpose of illustration and description. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A basketball examination system based on AI vision, characterized in that: include: An image acquisition module, which is used to collect image information of the examinee during the basketball test; An information processing module, the information processing module is used to establish a three-dimensional coordinate system according to the information collected by the image acquisition module, and generate the movement of the basketball and the examinee in the three-dimensional coordinate system during the examinee's basketball test; A basketball recognition module, the basketball recognition module is used to recognize basketball information in the three-dimensional coordinate system established by the information processing module and recognize the basketball motion trajectory; A motion recognition module, which is used to recognize the examinee's motion information in the three-dimensional coordinate system established by the information processing module, and to recognize and analyze the examinee's motion information; A comprehensive analysis module, which is used to perform performance evaluation and generate training suggestions based on the identified basketball motion trajectory information and the examinee's movement information; A display module, which is used to display the analysis results of the comprehensive analysis module in real time; The functional interaction relationship between the modules of the basketball examination system based on AI vision is as follows: The image acquisition module transmits the collected image information to the information processing module, the information processing module transmits the processed information to the basketball recognition module and the action recognition module at the same time, the basketball recognition module transmits the recognized basketball information to the comprehensive analysis module, the action recognition module transmits the recognized candidate action information to the comprehensive analysis module, and the comprehensive analysis module transmits the generated performance evaluation and training suggestions to the display module for display.

2. The AI ​​vision-based basketball examination system as claimed in claim 1, characterized in that: The image acquisition module includes a multi-module camera; The multi-module camera is used to collect test images of the examinee during the basketball test, and the test image content includes but is not limited to the running trajectory of the basketball and the examinee's movements.

3. The basketball examination system based on AI vision as claimed in claim 1, characterized in that: The information processing module establishes a three-dimensional coordinate system, places the information collected by the image acquisition module in the three-dimensional coordinate system, converts the change information in the image into coordinate information in the three-dimensional coordinate system, and retains the color of the relevant objects.

4. The AI ​​vision-based basketball examination system as claimed in claim 1, characterized in that: The basketball recognition module includes a feature extraction unit, a basketball model unit and a recognition unit; The feature extraction unit is used to extract feature information from basketball pictures of different angles, types and colors to provide a basis for model establishment; The basketball model unit establishes a basketball recognition model by acquiring feature information of basketball pictures of different angles, types and colors extracted by the feature extraction unit; The recognition unit inputs the three-dimensional coordinate system image data transmitted in real time by the information processing module into the basketball recognition model for recognition, thereby confirming the movement trajectory of the basketball.

5. The AI ​​vision-based basketball examination system as claimed in claim 4, characterized in that: The feature extraction unit collects a specified number of basketball pictures of different angles, types and colors, as well as some pictures of non-basketball objects as negative samples; crops and scales the pictures to unify the picture size to M×N pixels; Extract color features: Calculate the color mean of each image, and the coordinates of each pixel are (x, y); for color images, convert them to hue, saturation, and brightness color space, and calculate the mean of the three channels of hue H, saturation S, and brightness color V, that is, the color mean, and the formula is: Extract shape features: calculate the roundness B and area A of the object in the image, The edge pixel coordinates of the object are (x i ,y i ), i is a set, meaning the i-th coordinate point, the number is u; perimeter area Circularity B = 4πA / O 2 .

6. The AI ​​vision-based basketball examination system as claimed in claim 4, characterized in that: The basketball model unit establishes a basketball recognition model. First, the features extracted by the feature extraction unit and the corresponding labels, basketball or non-basketball, form a training data set; Among them, 0 represents non-basketball, 1 represents basketball, Where θ is the parameter vector of the model and p is the input feature set; The model parameters θ are optimized by minimizing the loss function: The optimal θ is solved by the gradient descent optimization algorithm, and the formula is: in, Represents the partial derivative of the loss function L(θ) with respect to the jth parameter of parameter θ, which is used to update the parameters in the gradient descent optimization algorithm; p i@ Represents the feature p of the i-th sample i The jth parameter of is taken and the most suitable model parameters are substituted into the model function to obtain the trained optimal basketball recognition model.

7. The AI ​​vision-based basketball examination system as claimed in claim 4, characterized in that: The recognition unit extracts color and shape features from the three-dimensional coordinate system image data transmitted in real time in the feature extraction unit; Input the extracted features into the trained model to get the output value H / (p i ); If H / (p i )>0.5, it is judged that there is a basketball in the image; otherwise, it is judged that there is no basketball; For a continuous image sequence, the above recognition process is performed sequentially in chronological order; In order to determine the trajectory of the basketball, the displacement vector between the basketball coordinates (x1, y1, z1) and (x2, y2, z2) identified in two adjacent frames can be calculated. The trajectory of the basketball is specified by the displacement vectors between multiple consecutive frames.

8. The AI ​​vision-based basketball examination system as claimed in claim 1, characterized in that: The action recognition module includes an action model unit and an action analysis unit; The action model unit is used to obtain standard action information in a basketball test from, including but not limited to, basketball test specifications, and establish coordinate points of human joints in the standard action in a three-dimensional coordinate system; The motion analysis unit is used to calculate the final test score of the examinee based on the coordinate information of the human joint points of the standard motion in the motion model unit, using human body recognition technology and the positions of the starting point and the end point.

9. The AI ​​vision-based basketball examination system as claimed in claim 8, characterized in that: The action analysis unit receives the three-dimensional coordinate system image data transmitted in real time, including the joint point information and the starting point and the basketball hoop coordinate information, and calculates the candidate's violation and action score in combination with the dribbling action standard; The coordinates of the human joint points in the kth frame are (x G,i ,y G,i ,z G,i ), i is a set, meaning the i-th coordinate point, the number is u; is the coordinate of the i-th joint point in the k-th frame; The starting point of the test is the coordinate f=(x H ,y H ,z H ), the end point coordinate g=(x I ,y I ,z I ); Input the human joint point coordinate sequence of each frame into the action model unit for recognition and comparison to obtain the corresponding accuracy; The accuracy is measured by measuring the distance between two point sets. In the action sequence, the set of joint point coordinates is regarded as a point set for calculation; The collected action joint point coordinate set is α, and the standard action joint point coordinate set is β. Then the one-way Hausdorff distance r from set α to set β is: r(α,β)=max{α i ∈α}min{β i ∈β}w(α i ,b i ) For each coordinate point α in the set α i , find the point β closest to it in the set β i , calculate the maximum value of all these closest distances; reflect the overall deviation of the points of set α from set β; where d(a,b) is some distance measure between point a and point b; The distance w is The standardization of the action is measured according to the length of the calculation distance, and the action score of the subject is calculated: Determine whether the action is illegal: If the distance gap violation judgment condition C(a_i) is met, the score of the action is 0.

10. The AI ​​vision-based basketball examination system as claimed in claim 1, characterized in that: The comprehensive analysis module performs real-time scoring according to the action score identified by the action recognition module and the basketball trajectory identified by the basketball recognition module, based on the relevant scoring mechanism; The display module generates a display interface to display the test results in real time. The display interface content includes the test taker's score, action analysis, key points for deduction and improvement suggestions.

Images