Basketball stand based on artificial intelligence

Abstract

The basketball stand comprises a mechanical mechanism and an intelligent data processing mechanism, and the mechanical mechanism comprises a longitudinal basket rod, a transverse basket rod, a backboard and a basket ring; the intelligent data processing mechanism comprises a data collection module, a data preprocessing module, a shooting recognition module, a shooting score prediction model, a shooting score prediction module and a shooting result display module. The invention belongs to the technical field of sports equipment, and provides a basketball stand based on artificial intelligence. The system plays an important role in improving the training efficiency of athletes, enhancing the ornamental value and competitive property of basketball games, providing rich data support for scientific research of basketball sports, promoting deep fusion of sports equipment and artificial intelligence technology, reducing the working intensity of coaches and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of sports equipment, and in particular to a basketball stand based on artificial intelligence. Background Art

[0002] In traditional basketball, the function of basketball stands is limited to being a target device for shooting. In basketball training and competition scenarios, athletes cannot obtain detailed analysis of their own shooting data in real time, such as whether the shooting posture is standard, which is directly related to the stability and efficiency of the shot; the impact of shooting force and angle on scoring, and different combinations of force and angle determine whether the basketball can accurately fall into the basket. As for coaches, due to the lack of accurate data support, it is difficult to accurately guide athletes to improve their shooting techniques based on their personal characteristics.

[0003] As the competitiveness and professionalism of basketball continue to improve, the need for refined training of athletes' technical level is becoming increasingly urgent. At the same time, artificial intelligence technology has made rapid progress in recent years, showing strong capabilities in data processing, pattern recognition, and predictive analysis. Introducing artificial intelligence technology into the design of basketball stands, enabling them to collect, analyze, and feedback shooting data, has become the key to solving the problem of missing data in current basketball training and competitions, and has important practical significance and application value. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a basketball stand based on artificial intelligence, which plays an important role in improving the training efficiency of athletes, enhancing the viewing and competitiveness of basketball games, providing rich data support for scientific research on basketball, promoting the deep integration of sports equipment and artificial intelligence technology, and reducing the workload of coaches.

[0005] The technical solution adopted by the present invention is as follows: a basketball stand based on artificial intelligence, including a mechanical mechanism and an intelligent data processing mechanism, the mechanical mechanism includes a longitudinal basketball pole, a transverse basketball pole, a backboard, and a basketball ring, and the intelligent data processing mechanism includes a data collection module, a data preprocessing module, a shot recognition module, a shot score prediction model, a shot score prediction module, and a shot result display module;

[0006] The longitudinal basket pole is fixed on the ground, the transverse basket pole is horizontally fixed on the longitudinal basket pole, the backboard is arranged on the transverse basket pole, and a basket ring is fixed on the backboard; the longitudinal basket pole is used to support the entire basketball stand structure so that it stands firmly on the ground, the transverse basket pole connects the longitudinal basket pole and the backboard to provide horizontal support for the backboard, and a basket ring is fixed on the backboard as the target area for the basketball to enter;

[0007] The data collection module includes a pressure sensor, an acceleration sensor, a gyroscope sensor, and a camera. The pressure sensor is arranged on the basketball ring, the acceleration sensor and the gyroscope sensor are arranged on the backboard, and the camera is arranged around the basketball stand. The data collection module is composed of multiple sensors distributed at different positions of the basketball stand, including a pressure sensor arranged on the basketball ring, which is used to detect the pressure change when the basketball contacts the basketball ring and obtain the force and position information of the basketball hitting the basketball ring; the acceleration sensor and the gyroscope sensor installed on the backboard are used to collect the vibration data of the backboard when the basketball hits the backboard and the information such as the rebound angle of the basketball on the backboard; and multiple cameras arranged around the basketball stand are used to shoot the athlete's shooting process from different angles and record visual data such as the shooting posture and the flight trajectory of the basketball.

[0008] The data preprocessing module processes the raw data collected by the data collection module, including denoising the sensor data using a wavelet transform algorithm, and performing image enhancement and grayscale processing on the image data captured by the camera;

[0009] The shot recognition module uses human posture recognition based on deep learning based on the pre-processed data to recognize the athlete's body posture, arm movement, position and angle information of the moment the basketball is shot, and combines with the data collection module to determine whether the basketball is shot, the direction of the shot, and whether it hits the basket;

[0010] The shooting score prediction model adopts a model based on long short-term memory network. The shooting score prediction model uses the shooting action feature data and basketball flight trajectory data output by the shooting recognition module as input, and after training with training data, learns the intrinsic relationship between the shooting action and the score, and predicts whether the shot can score and the probability of scoring.

[0011] The shooting score prediction module calls the shooting score prediction model, uses parallel computing technology to quickly calculate the shooting score prediction result based on the real-time shooting data provided by the shooting recognition module, and transmits the prediction result to the shooting result display module;

[0012] The shooting result display module is a display screen, and the shooting result display module is installed in a position convenient for viewing on or around a basketball stand.

[0013] Furthermore, the sensor data is subjected to denoising processing using a wavelet transform algorithm, and the formula is:

[0014]

[0015] Where D(s, a) represents the wavelet transform coefficient, s is the translation parameter, a is the scale parameter, d(t) is the original sensor data, and ψ is the wavelet basis function.

[0016] Furthermore, the image data captured by the camera is subjected to image enhancement processing, and the formula is:

[0017]

[0018] In the formula, s k is the gray value after the k-th grayscale transformation, n j is the number of pixels with gray level j in the original image, and N is the total number of pixels in the image;

[0019] At the same time, grayscale processing is performed using the formula: Gray = 0.299R + 0.587G + 0.114B; where Gray is the grayscale value, and R, G, and B are the red, green, and blue channel values ​​of the image respectively;

[0020] And perform time synchronization processing based on timestamps for different types of data.

[0021] Furthermore, the determination of whether the basketball has been shot is to determine the coordinates of the arm starting point and the coordinates of the basketball leaving the hand from the coordinates of the key points of the image identified by the camera, and then calculate the arm extension vector, which reflects the extension direction and distance of the arm during the shooting process; combine the vertical acceleration in the acceleration sensor data and the angular velocity in the gyroscope sensor data to make a comprehensive judgment; when the modulus of the arm extension vector is greater than the set arm extension distance threshold, and the vertical acceleration is greater than the vertical acceleration threshold, and the angular velocity is greater than the angular velocity threshold, it is determined that the basketball has been shot. This is because the arm is extended to a certain distance, and the acceleration and angular velocity reach specific values, which comprehensively reflects that the athlete has completed the action of shooting.

[0022] Furthermore, the direction of the shot is determined by taking the center coordinates of the basket ring as a reference, and calculating the shooting direction vector according to the position of the basketball at the moment of shooting. The vector intuitively represents the direction from the shooting point of the basketball to the center of the basket ring; in order to more accurately describe the shooting direction, the direction vector is normalized to obtain a unit direction vector; then, the angle of the unit direction vector is calculated by the inverse tangent function, and this angle determines the direction of the shot. In this way, the shooting direction can be accurately quantified, providing a basis for the subsequent analysis of the accuracy and strategy of the shot.

[0023] Furthermore, in the determination of whether the basketball hoop is hit, the basketball hoop is evenly divided into K areas, and an independent pressure sensor is installed in each area; suppose the data collected by the pressure sensor of the kth (k=1, 2, ..., K) area at a certain moment is When the basketball comes into contact with the hoop, due to the different contact positions and forces, the pressure sensors in each area will generate corresponding electrical signal output according to the change in pressure. The output value is the pressure sensor data.

[0024] In order to accurately determine whether the basketball hits the hoop, two key threshold conditions are set; one is the total pressure threshold P total-thres , this threshold is an empirical value obtained based on a large number of experiments and data analysis, representing the minimum value of the sum of all pressure sensor data when the basketball and the hoop are in effective contact; when at a certain moment, the sum of all regional pressure sensor data Greater than P total-thres This indicates that there is a certain degree of physical contact between the basketball and the hoop, and it is preliminarily judged that the basketball may hit the hoop;

[0025] The second is the single area pressure threshold P single-thres , this threshold is also determined based on a large amount of experimental data and the physical properties of the hoop; it represents the minimum effective pressure value that a single area pressure sensor should detect when the basketball hits the hoop; when the total pressure threshold condition is met, there is at least one area k such that Greater than P single-thres This further confirmed that the basketball hit the rim;

[0026] Only when the above two conditions are met at the same time, that is, and This multi-condition judgment method can effectively avoid misjudgments caused by the basketball just lightly brushing the edge of the hoop or the pressure sensor producing slight fluctuations due to external interference, greatly improving the accuracy and reliability of hit judgment.

[0027] Furthermore, the shooting score prediction model is based on a long short-term memory network, and uses the data output by the shooting recognition module as input to predict the shooting score;

[0028] In data processing, the shooting action feature data obtains the coordinates of the key parts of the athlete through the human posture recognition algorithm, calculates the joint angle and speed derivative features, and forms the feature vector A; the basketball flight trajectory data is measured by multiple cameras and auxiliary sensors, and the speed, acceleration and curvature are calculated after smoothing to form the feature vector T; A and T are spliced ​​and fused into a comprehensive feature vector X, and standardized preprocessing is performed to ensure that the data is at the same scale;

[0029] In the model training phase, the LSTM model structure is adopted, and the core units include input gate, forget gate, output gate and cell state. Through the gating mechanism of each layer, the input data is screened, integrated and output, and multiple layers are stacked to learn complex relationships. The training data covers shooting samples under various shooting methods, positions and physical fitness differences of athletes, and is divided into training set and validation set. The cross entropy loss function is used to measure the prediction error, and the model parameters are updated with the help of Adam optimizer. At the same time, regularization and early stopping methods are used to prevent overfitting.

[0030] When predicting, the shooting action and basketball flight trajectory data are obtained in real time, processed according to the training process and input into the trained model; the model outputs a scoring probability between 0 and 1; a threshold of 0.5 is set to determine whether the shot is successful. The probability can also be displayed directly, or the prediction stability can be improved through sliding average statistical analysis. Combined with the shooter's historical hit rate and game pressure information, comprehensive analysis and suggestions are provided to athletes and coaches.

[0031] Furthermore, the shooting score prediction module obtains real-time data from the shooting recognition module, including the athlete's body posture, force shooting action characteristics, and speed and acceleration information in the basketball flight trajectory; these data are integrated into a feature vector X, and standardized preprocessing is performed to eliminate data scale differences;

[0032] Enter the parallel computing stage; divide the feature vector X into multiple partitions according to the time step and feature dimension; use multi-threading combined with GPU acceleration technology, and each thread is responsible for the operation of one partition; under the powerful parallel processing capability of GPU, the computing speed is greatly improved;

[0033] During the calculation process, for each partition, a multi-head self-attention mechanism is introduced to mine the data value, and a series of matrix operations are used to generate query, key, and value matrices to calculate the attention output;

[0034] The input parallel LSTM units are processed and combined with adaptive weight adjustment items to dynamically adjust the information input, allowing the model to learn data features more accurately;

[0035] After the calculation is completed, the dynamic weight is determined according to the importance of each partition, and the final hidden state of each partition is obtained by fusing;

[0036] After the fully connected layer, the weight matrix and bias vector will be mapped into the predicted scoring probability and quickly transmitted to the shooting result display module to intuitively present the possibility of shooting scoring.

[0037] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0038] The present invention can collect, analyze and feedback the athlete's shooting data in real time, helping the athlete to timely understand his or her shooting performance and adjust the shooting strategy. The present invention improves the training efficiency of the athlete, and the athlete can adjust the shooting action in time according to the real-time feedback data, and quickly improve the shooting technique level.

[0039] The present invention enhances the viewing and competitiveness of basketball games. Audiences can learn more detailed information about shots through the display screen, and athletes can also perform better in the game based on data feedback. In the game, athletes can better perform their level and improve their competitiveness based on real-time shot score prediction and action analysis.

[0040] The present invention provides accurate data support for coaches, enabling them to provide targeted guidance based on the characteristics of each athlete and improve training results. The present invention provides rich data support for scientific research on basketball, and researchers can explore the best strategies and technical points for shooting by analyzing a large amount of shooting data.

[0041] The large amount of shooting data collected by the present invention can be used for scientific research on basketball, exploring the best shooting techniques and strategies, and promoting the development of basketball. The present invention promotes the deep integration of sports equipment and artificial intelligence technology, and provides reference for the intelligent development of equipment in other sports.

[0042] The present invention reduces the workload of coaches. Coaches can use the data provided by the present invention to guide athletes more efficiently without spending a lot of time and energy on manual statistics and analysis. The present invention can not only analyze the results of shooting, but also analyze multiple dimensions such as shooting posture, strength, angle, etc., to comprehensively evaluate the shooting ability of athletes.

[0043] The present invention adopts a variety of advanced sensors and algorithms to ensure the accuracy and reliability of data and the stability of the system, promotes the innovation of basketball training methods and technologies, and injects new vitality into the development of basketball. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the mechanical structure of a basketball stand based on artificial intelligence provided by the present invention;

[0045] Figure 2 A schematic diagram of an intelligent data processing mechanism module of an artificial intelligence-based basketball stand provided by the present invention.

[0046] Among them, 1. longitudinal basketball pole, 2. transverse basketball pole, 3. backboard, 4. basketball hoop.

[0047] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0050] Example 1, see Figure 1 The present invention provides an artificial intelligence-based basketball stand, including a mechanical mechanism and an intelligent data processing mechanism, wherein the mechanical mechanism includes a longitudinal basket rod 1, a transverse basket rod 2, a backboard 3, and a basket ring 4, wherein the longitudinal basket rod 1 is fixed to the ground, the transverse basket rod 2 is horizontally fixed to the longitudinal basket rod 1, the backboard 3 is arranged on the transverse basket rod 2, and a basket ring 4 is fixed on the backboard 3; the longitudinal basket rod 1 is used to support the entire basketball stand structure so that it stands firmly on the ground, the transverse basket rod 2 connects the longitudinal basket rod 1 and the backboard 3 to provide horizontal support for the backboard 3, and a basket ring 4 is fixed on the backboard 3 as a target area for the basketball to enter.

[0051] Example 2, see Figure 1 and Figure 2, this embodiment is based on the above embodiment, the intelligent data processing mechanism includes a data collection module, a data preprocessing module, a shot recognition module, a shot score prediction model, a shot score prediction module, and a shot result display module; the data collection module includes a pressure sensor, an acceleration sensor, a gyroscope sensor, and a camera, the pressure sensor is arranged on the basket ring 4, the acceleration sensor and the gyroscope sensor are arranged on the backboard 3, and the camera is arranged around the basketball stand; the data collection module is composed of a plurality of sensors, which are distributed at different positions of the basketball stand, including a pressure sensor arranged on the basket ring 4, which is used to detect the pressure change when the basketball contacts the basket ring 4, and obtain the force and position information of the basketball hitting the basket ring 4; the acceleration sensor and the gyroscope sensor installed on the backboard 3 are used to collect the vibration data of the backboard 3 when the basketball hits the backboard 3 and the rebound angle of the basketball on the backboard 3 and other information; and a plurality of cameras arranged around the basketball stand, which shoot the athlete's shooting process from different angles and record the visual data such as the shooting posture and the flight trajectory of the basketball.

[0052] The data preprocessing module processes the raw data collected by the data collection module, including denoising the sensor data using a wavelet transform algorithm, and performing image enhancement and grayscale processing on the image data captured by the camera;

[0053] The shooting recognition module uses the human posture recognition based on deep learning based on the pre-processed data to recognize the athlete's body posture, arm movement, position and angle information of the basketball shot moment, and combines with the data collection module to determine whether the basketball has been shot, the direction of the shot, and whether it hits the basket 4;

[0054] The shooting score prediction model adopts a model based on long short-term memory network. The shooting score prediction model uses the shooting action feature data and basketball flight trajectory data output by the shooting recognition module as input, and after training with training data, learns the intrinsic relationship between the shooting action and the score, and predicts whether the shot can score and the probability of scoring.

[0055] The shooting score prediction module calls the shooting score prediction model, uses parallel computing technology to quickly calculate the shooting score prediction result based on the real-time shooting data provided by the shooting recognition module, and transmits the prediction result to the shooting result display module;

[0056] The shooting result display module is a display screen, and the shooting result display module is installed in a position convenient for viewing on or around a basketball stand.

[0057] Embodiment 3: This embodiment is based on the above embodiment, and the sensor data is subjected to denoising by using a wavelet transform algorithm, and the formula is:

[0058]

[0059] Where D(s, a) represents the wavelet transform coefficient, s is the translation parameter, a is the scale parameter, d(t) is the original sensor data, and ψ is the wavelet basis function.

[0060] Embodiment 4, based on the above embodiment, the image data captured by the camera is subjected to image enhancement processing, and the formula is:

[0061]

[0062] In the formula, s k is the gray value after the k-th grayscale transformation, n j is the number of pixels with gray level j in the original image, and N is the total number of pixels in the image;

[0063] At the same time, grayscale processing is performed using the formula: Gray = 0.299R + 0.587G + 0.114B; where Gray is the grayscale value, and R, G, and B are the red, green, and blue channel values ​​of the image respectively;

[0064] And perform time synchronization processing based on timestamps for different types of data.

[0065] Embodiment 5, based on the above embodiment, the determination of whether the basketball has been shot is to determine the coordinates of the starting point of the arm and the coordinates of the point where the basketball leaves the hand from the coordinates of the key points of the image identified by the camera, and then calculate the arm extension vector, which reflects the extension direction and distance of the arm during the shooting process; combine the vertical acceleration in the acceleration sensor data and the angular velocity in the gyroscope sensor data to make a comprehensive judgment; when the modulus of the arm extension vector is greater than the set arm extension distance threshold, and the vertical acceleration is greater than the vertical acceleration threshold, and the angular velocity is greater than the angular velocity threshold, it is determined that the basketball has been shot. This is because the arm is extended to a certain distance, the acceleration and angular velocity reach specific values, which comprehensively reflects that the athlete has completed the action of shooting.

[0066] Embodiment 6, based on the above embodiment, the direction of the shot is determined by taking the center coordinates of the basket ring 4 as a reference, and calculating the shooting direction vector according to the position of the basketball at the moment of shooting. The vector intuitively represents the direction from the shooting point of the basketball to the center of the basket ring 4; in order to more accurately describe the shooting direction, the direction vector is normalized to obtain a unit direction vector; then, the angle of the unit direction vector is calculated by the inverse tangent function, and this angle determines the direction of the shot. In this way, the shooting direction can be accurately quantified, providing a basis for the subsequent analysis of the accuracy and strategy of the shot.

[0067] Embodiment 7, based on the above embodiment, the embodiment determines whether the basket ring 4 is hit, and the basket ring 4 is evenly divided into K areas, and an independent pressure sensor is installed in each area; suppose the data collected by the pressure sensor of the kth (k=1, 2, ..., K) area at a certain moment is When the basketball comes into contact with the hoop 4, due to the different contact positions and forces, the pressure sensors in each area will generate corresponding electrical signal outputs according to the changes in the pressure. The output value is the pressure sensor data.

[0068] In order to accurately determine whether the basketball hits the hoop 4, two key threshold conditions are set; one is the total pressure threshold P total-thres , the threshold is an empirical value obtained based on a large number of experiments and data analysis, representing the minimum value of the sum of all pressure sensor data when the basketball and the hoop 4 are in effective contact; when at a certain moment, the sum of all regional pressure sensor data Greater than P total-thres This indicates that there is a certain degree of physical contact between the basketball and the basket ring 4, and it is preliminarily determined that the basketball may hit the basket ring 4;

[0069] The second is the single area pressure threshold P single-thres , this threshold is also determined based on a large amount of experimental data and the physical characteristics of the basket ring 4; it represents the minimum effective pressure value that a single area pressure sensor should detect when the basketball hits the basket ring 4; when the total pressure threshold condition is met, there is at least one area k such that Greater than P single-thres This further confirms that the basketball hits the hoop 4;

[0070] Only when the above two conditions are met at the same time, that is, and Only when the basketball hits the basket ring 4, it is determined that the basketball hits the basket ring 4. This multi-condition judgment method can effectively avoid misjudgment caused by the basketball just lightly brushing the edge of the basket ring 4, or the pressure sensor generating a small fluctuation due to external interference, and greatly improves the accuracy and reliability of the hit judgment.

[0071] Embodiment 8, based on the above embodiment, the shooting score prediction model is based on a long short-term memory network, and uses the data output by the shooting recognition module as input to predict the shooting score;

[0072] In data processing, the shooting action feature data obtains the coordinates of the key parts of the athlete through the human posture recognition algorithm, calculates the joint angle and speed derivative features, and forms the feature vector A; the basketball flight trajectory data is measured by multiple cameras and auxiliary sensors, and the speed, acceleration and curvature are calculated after smoothing to form the feature vector T; A and T are spliced ​​and fused into a comprehensive feature vector X, and standardized preprocessing is performed to ensure that the data is at the same scale;

[0073] In the model training phase, the LSTM model structure is adopted, and the core units include input gate, forget gate, output gate and cell state. Through the gating mechanism of each layer, the input data is screened, integrated and output, and multiple layers are stacked to learn complex relationships. The training data covers shooting samples under various shooting methods, positions and physical fitness differences of athletes, and is divided into training set and validation set. The cross entropy loss function is used to measure the prediction error, and the model parameters are updated with the help of Adam optimizer. At the same time, regularization and early stopping methods are used to prevent overfitting.

[0074] When predicting, the shooting action and basketball flight trajectory data are obtained in real time, processed according to the training process and input into the trained model; the model outputs a scoring probability between 0 and 1; a threshold of 0.5 is set to determine whether the shot is successful. The probability can also be displayed directly, or the prediction stability can be improved through sliding average statistical analysis. Combined with the shooter's historical hit rate and game pressure information, comprehensive analysis and suggestions are provided to athletes and coaches.

[0075] Embodiment 9, based on the above embodiment, the shooting score prediction module obtains real-time data from the shooting recognition module, including the athlete's body posture, force situation, shooting action characteristics, and speed and acceleration information in the basketball flight trajectory; these data are integrated into a feature vector X, and standardized preprocessing is performed to eliminate data scale differences;

[0076] Enter the parallel computing stage; divide the feature vector X into multiple partitions according to the time step and feature dimension; use multi-threading combined with GPU acceleration technology, and each thread is responsible for the operation of one partition; under the powerful parallel processing capability of GPU, the computing speed is greatly improved;

[0077] During the calculation process, for each partition, a multi-head self-attention mechanism is introduced to mine the data value, and a series of matrix operations are used to generate query, key, and value matrices to calculate the attention output;

[0078] The input parallel LSTM units are processed and combined with adaptive weight adjustment items to dynamically adjust the information input, allowing the model to learn data features more accurately;

[0079] After the calculation is completed, the dynamic weight is determined according to the importance of each partition, and the final hidden state of each partition is obtained by fusing;

[0080] After the fully connected layer, the weight matrix and bias vector will be mapped into the predicted scoring probability and quickly transmitted to the shooting result display module to intuitively present the possibility of shooting scoring.

[0081] When an athlete is doing shooting training or competition, the data collection module starts working. The pressure sensor detects the pressure change when the basketball contacts the basket ring 4, the acceleration sensor and the gyroscope sensor collect the vibration data and rebound angle when the basketball hits the backboard 3, and the camera shoots the athlete's shooting process from different angles.

[0082] The data preprocessing module processes the collected raw data. First, the sensor data is denoised, then the image data is enhanced and grayed, and finally different types of data are time synchronized.

[0083] The shot recognition module uses image recognition technology and data analysis algorithms based on the preprocessed data to identify the athlete's shooting action, determine whether the basketball has been shot, the shooting direction, and whether it has hit the basket 4.

[0084] The shooting score prediction module calls the shooting score prediction model to predict whether the shot can score and the probability of scoring based on the real-time shooting data provided by the shooting recognition module.

[0085] The shooting result display module displays the predicted score results, actual score, analysis and evaluation of the shooting action and other information in real time for athletes and coaches to view.

[0086] The present invention solves the problem that athletes cannot obtain detailed analysis of their own shooting data in time. Through the collaborative work of various modules, it can collect and analyze various data such as shooting posture, strength, angle, etc., and display the results intuitively to athletes. The present invention solves the problem that it is difficult for coaches to provide targeted guidance to athletes to improve their shooting techniques through precise data. Coaches can formulate personalized training plans based on the data presented by the shooting result display module. The present invention overcomes the problem of single function of traditional basketball stands, gives basketball stands intelligent data processing and feedback functions, and improves the professionalism and scientificity of basketball training and competition. The present invention solves the problems of noise interference and data asynchrony in the process of data collection and processing. The accuracy and relevance of the data are guaranteed through the denoising and time synchronization processing of the data preprocessing module. The present invention solves the problem of inaccurate shooting score prediction. The shooting score prediction model constructed by adopting a deep learning algorithm improves the accuracy and reliability of the prediction.

[0087] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0088] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0089] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A basketball stand based on artificial intelligence, characterized in that: It includes a mechanical mechanism and an intelligent data processing mechanism, wherein the mechanical mechanism includes a longitudinal basketball pole, a transverse basketball pole, a backboard, and a basketball hoop, and the intelligent data processing mechanism includes a data collection module, a data preprocessing module, a shot recognition module, a shot score prediction model, a shot score prediction module, and a shot result display module; The longitudinal basket pole is fixed on the ground, the transverse basket pole is horizontally fixed on the longitudinal basket pole, the backboard is arranged on the transverse basket pole, and a basket ring is fixed on the backboard; The data collection module includes a pressure sensor, an acceleration sensor, a gyroscope sensor, and a camera. The pressure sensor is arranged on the basketball hoop, the acceleration sensor and the gyroscope sensor are arranged on the backboard, and the camera is arranged around the basketball stand. The data preprocessing module processes the raw data collected by the data collection module, including denoising the sensor data using a wavelet transform algorithm, and performing image enhancement and grayscale processing on the image data captured by the camera; The shot recognition module uses human posture recognition based on deep learning based on the pre-processed data to recognize the athlete's body posture, arm movement, position and angle information of the moment the basketball is shot, and combines with the data collection module to determine whether the basketball is shot, the direction of the shot, and whether it hits the basket; The shooting score prediction model adopts a model based on long short-term memory network. The shooting score prediction model uses the shooting action feature data and basketball flight trajectory data output by the shooting recognition module as input, and after training with training data, learns the intrinsic relationship between the shooting action and the score, and predicts whether the shot can score and the probability of scoring. The shooting score prediction module calls the shooting score prediction model, uses parallel computing technology to quickly calculate the shooting score prediction result based on the real-time shooting data provided by the shooting recognition module, and transmits the prediction result to the shooting result display module; The shooting result display module is a display screen, and the shooting result display module is installed in a position convenient for viewing on or around a basketball stand.

2. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The sensor data is denoised using a wavelet transform algorithm, and the formula is: Where D(s, a) represents the wavelet transform coefficient, s is the translation parameter, a is the scale parameter, d(t) is the original sensor data, and ψ is the wavelet basis function.

3. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The image enhancement processing is performed on the image data captured by the camera, and the formula is: Where sk is the grayscale value after the k-th grayscale transformation, nj is the number of pixels with grayscale j in the original image, and N is the total number of pixels in the image; At the same time, grayscale processing is performed using the formula: Gray = 0.299R + 0.587G + 0.114B; In the formula, Gray is the grayscale value, R, G, and B are the red, green, and blue channel values ​​of the image respectively; And perform time synchronization processing based on timestamps for different types of data.

4. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The method for judging whether the basketball has been shot is to determine the coordinates of the arm starting point and the coordinates of the point where the basketball leaves the hand from the coordinates of the key points of the image recognized by the camera, and then calculate the arm extension vector, which reflects the extension direction and distance of the arm during the shooting process; and make a comprehensive judgment based on the vertical acceleration in the acceleration sensor data and the angular velocity in the gyroscope sensor data; when the modulus of the arm extension vector is greater than the set arm extension distance threshold, and the vertical acceleration is greater than the vertical acceleration threshold, and the angular velocity is greater than the angular velocity threshold, it is determined that the basketball has been shot.

5. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The direction of the shot is determined by taking the center coordinates of the basket ring as a reference, and calculating the shooting direction vector according to the position of the basketball at the moment of release. The shooting direction vector intuitively represents the direction from the release point of the basketball to the center of the basket ring; in order to more accurately describe the shooting direction, the direction vector is normalized to obtain a unit direction vector; the angle of the unit direction vector is calculated by the inverse tangent function, and the angle of the unit direction vector is used to determine the direction of the shot.

6. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The method for determining whether the basketball hoop is hit is to divide the basketball hoop into K areas evenly, and an independent pressure sensor is installed in each area; suppose the data collected by the pressure sensor of the kth (k=1, 2, ..., K) area at a certain moment is When the basketball comes into contact with the hoop, due to the different contact positions and forces, the pressure sensors in each area will generate corresponding electrical signal outputs according to the changes in the pressure. The output value of the electrical signal is the pressure sensor data. In order to accurately determine whether the basketball hits the hoop, two key threshold conditions are set; one is the total pressure threshold P total-thres The total pressure threshold is an empirical value obtained based on a large number of experiments and data analysis. It represents the minimum value of the sum of all pressure sensor data when the basketball and the hoop are in effective contact. Greater than P total-thres This indicates that there is a certain degree of physical contact between the basketball and the hoop, and it is preliminarily judged that the basketball may hit the hoop; The second is the single area pressure threshold P single-thres The single area pressure threshold is determined based on a large amount of experimental data and the physical properties of the basketball hoop; the single area pressure threshold represents the minimum effective pressure value that a single area pressure sensor should detect when the basketball hits the basketball hoop; when the total pressure threshold condition is met, there is at least one area k such that Greater than P single-thres This further confirmed that the basketball hit the rim; Only when the above two conditions are met at the same time, that is, and Only then is it determined that the basketball has hit the hoop.

7. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The shooting score prediction model is based on a long short-term memory network and uses the data output by the shooting recognition module as input to predict the shooting score. In the data processing stage, the shooting action feature data obtains the coordinates of the key parts of the athlete through the human posture recognition algorithm, calculates the joint angle and velocity derivative features, and forms the feature vector A; the basketball flight trajectory data is measured by multiple cameras and sensors, and the speed, acceleration and curvature are calculated after smoothing to form the feature vector T; A and T are spliced ​​and fused into a comprehensive feature vector X, and standardized preprocessing is performed to ensure that the data is at the same scale; In the model training phase, the LSTM model structure is adopted, and the core units include input gate, forget gate, output gate and cell state. Through the gating mechanism of each layer, the input data is screened, integrated and output, and multiple layers are stacked to learn complex relationships. The training data covers shooting samples under various shooting methods, positions and physical fitness differences of athletes, and is divided into training set and validation set. The cross entropy loss function is used to measure the prediction error, and the model parameters are updated with the help of Adam optimizer. At the same time, regularization and early stopping methods are used to prevent overfitting. When predicting, the shooting action and basketball flight trajectory data are obtained in real time, processed according to the training process and input into the trained model; the model outputs a scoring probability between 0 and 1; a threshold of 0.5 is set to determine whether the shot is successful. The probability can also be displayed directly, or the prediction stability can be improved through sliding average statistical analysis. Combined with the shooter's historical hit rate and game pressure information, comprehensive analysis and suggestions are provided to athletes and coaches.

8. The basketball stand based on artificial intelligence according to claim 1, characterized in that: The shooting score prediction module obtains real-time data from the shooting recognition module, including the athlete's body posture, force shooting action characteristics, and speed and acceleration information in the basketball flight trajectory; these data are integrated into a feature vector X, and standardized preprocessing is performed to eliminate data scale differences; Enter the parallel computing stage; divide the feature vector X into multiple partitions according to the time step and feature dimension; use multi-threading combined with GPU acceleration technology, and each thread is responsible for the operation of one partition; under the powerful parallel processing capability of GPU, the computing speed is greatly improved; During the calculation process, for each partition, a multi-head self-attention mechanism is introduced to mine the data value, and a series of matrix operations are used to generate query, key, and value matrices to calculate the attention output; The input parallel LSTM units are processed and combined with adaptive weight adjustment items to dynamically adjust the information input, allowing the model to learn data features more accurately; After the calculation is completed, the dynamic weight is determined according to the importance of each partition, and the final hidden state of each partition is obtained by fusing; After the fully connected layer, the weight matrix and bias vector will be mapped into the predicted scoring probability and quickly transmitted to the shooting result display module to intuitively present the possibility of shooting scoring.

Images