Physical exercise AI standing long jump evaluation data management system

Abstract

The invention relates to the technical field of sports artificial intelligence and data management, in particular to a physical exercise AI standing long jump evaluation data management system which comprises a dynamic environment calibration module, a multi-mode identity recognition module, a three-dimensional posture analysis module and the like. The dynamic environment calibration module uses a multispectral sensor array and a deep learning algorithm to establish a dynamic coordinate system without fixing a marker; the multi-mode identity recognition module fuses gait and face features to realize non-contact identity authentication; the three-dimensional attitude analysis module performs high-precision modeling and analysis on the long jump action through a multi-view camera and an advanced algorithm; the intelligent violation determination engine combines rules and an AI model, automatically identifies violation behaviors and can dynamically adjust determination criteria. The method gets rid of the limitation of a traditional evaluation mode, realizes high-precision and intelligent evaluation, effectively solves the problems of poor environmental adaptability, inaccurate identity authentication, unguaranteed data security and the like, and is scientific and reliable in evaluation result and safe and efficient in data management.

Description

Technical Field

[0001] The present invention relates to the technical field of sports artificial intelligence and data management, and in particular to a sports exercise AI standing long jump evaluation data management system. Background Art

[0002] In the fields of physical education, competitive training and mass fitness, standing long jump is a basic and important sport, and the accuracy and efficiency of its evaluation have always been the focus of attention. With the development of sports and the rise of the national fitness craze, the limitations of traditional standing long jump evaluation methods have become increasingly prominent.

[0003] At present, the evaluation of standing long jump mostly relies on manual refereeing or simple sensor equipment. Manual refereeing is greatly affected by subjective factors, and the judgment standards of different referees are different. It is difficult to accurately capture the details of the action, which is prone to misjudgment and omission, resulting in a lack of fairness and accuracy in the evaluation results; simple sensor equipment such as ordinary ranging sensors can only obtain the distance data of the long jump, and cannot analyze key action indicators such as take-off angle, aerial posture, landing stability, etc., which is difficult to meet the comprehensive and in-depth needs of professional training and scientific fitness for sports data.

[0004] In the context of rapid technological development, although some intelligent evaluation systems have begun to be applied in the field of sports, there are still many problems. Existing systems often rely on fixed markers for site calibration. When the environment changes, such as changes in light intensity, complex ground texture or the presence of obstacles, the calibration accuracy drops significantly, which in turn affects the entire evaluation result. In terms of identity authentication, most of them use a single card swiping, code scanning and other methods, which have the risk of identity fraud and data mismatching with users, and cannot guarantee the correspondence and security of evaluation data. In terms of data management, traditional systems lack effective data encryption and privacy protection measures, data storage is centralized, vulnerable to attacks and data leakage, and it is difficult to achieve collaborative optimization of multi-site data and model upgrades.

[0005] Therefore, in order to solve the above problems, a physical exercise AI standing long jump evaluation data management system is proposed. Summary of the invention

[0006] The purpose of the present invention is to provide a physical exercise AI standing long jump evaluation data management system to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A physical exercise AI standing long jump evaluation data management system, comprising:

[0009] Dynamic Environment Calibration Module: It collects real-time site environment data through a multi-spectral sensor array deployed around the long jump area, and establishes a dynamic coordinate system based on an adaptive calibration algorithm of deep learning. The adaptive calibration algorithm generates a virtual calibration grid by fusing environmental light intensity, ground texture features, and obstacle distribution, and does not rely on fixed markers;

[0010] Multi-modal Identity Recognition Module: It realizes contactless identity authentication by fusing gait recognition and face feature extraction technologies and combining user biometric information in the cloud database;

[0011] Three-dimensional Pose Analysis Module: It synchronously collects long jump action videos using multi-view cameras, constructs a motion trajectory model through a 3D human pose estimation algorithm, and real-time detects the takeoff angle, aerial pose, and landing stability;

[0012] Intelligent Violation Judgment Engine: Based on a rule engine and an AI time series analysis model, combined with kinematic parameters and spatial trajectory features, it automatically identifies violation behaviors including hop step, crossing the line, and body imbalance;

[0013] Distributed Data Management Unit: It uses blockchain technology to store evaluation data, supports multi-node synchronous verification and tamper-proof score records, and realizes local data processing through an edge computing module to reduce cloud dependence.

[0014] As a preferred solution, the implementation method of the dynamic environment calibration module includes:

[0015] Obtain real-time environmental information of the long jump area through a multi-spectral sensor array, including light intensity, ground texture, and obstacle distribution;

[0016] Extract environmental features based on an improved convolutional neural network CNN, and dynamically generate a virtual calibration grid adapted to different site conditions. The loss function of the improved CNN consists of two parts:

[0017] Feature reconstruction loss: Calculate the mean square error between the input environmental features and the reconstructed features, and the weight coefficient α is used to control its importance;

[0018] Environmental robustness loss: By maximizing the feature difference tolerance under light interference, the weight coefficient β is used to balance the anti-interference ability, and And The sum is 1;

[0019] Optimize the calibration parameters through a reinforcement learning algorithm to ensure the stability of the coordinate system in complex environments such as shadows and reflections.

[0020] As a preferred solution, the three-dimensional pose analysis module further includes:

[0021] Stereo images of long jump movements are synchronously captured using a multi-view camera, and a three-dimensional motion trajectory is reconstructed through an improved stereo matching algorithm. The energy function of the algorithm consists of the following three weighted terms:

[0022] Data term: Evaluate the consistency of three-dimensional poses through the projection error of multi-view images;

[0023] Smoothing term: Constrain the continuity of adjacent joint points in space;

[0024] Temporal term: Ensure the physical rationality of pose changes between adjacent frames. The weight coefficients λ1, λ2, and λ3 respectively control the contribution ratios of the three terms;

[0025] Analyze the joint motion sequence based on the Temporal Convolutional Network (TCN) to predict the centroid offset at the moment of landing;

[0026] Combine the data of the Inertial Measurement Unit (IMU), and correct the acceleration error of the aerial pose through a filtering algorithm to achieve sub-centimeter-level accuracy measurement.

[0027] As a preferred solution, the determination method of the intelligent violation determination engine includes:

[0028] Model the continuous frame action sequence through an improved Long Short-Term Memory (LSTM) network. The output of its forget gate is fused with predefined kinematic constraints, and the constraint weight coefficient γ controls the contribution of the physical prior;

[0029] Define a dynamic threshold rule library, adjust the determination criteria according to the user's age, gender, and historical data. The threshold is the weighted average of the baseline threshold and the adaptive offset, and the balance coefficient μ controls the weights of the two;

[0030] Generate an interpretable violation report, including action decomposition diagrams and improvement suggestions, and provide real-time feedback to the user terminal through an AR device.

[0031] As a preferred solution, the distributed data management unit further includes:

[0032] An intelligent contract mechanism based on blockchain to ensure the integrity and traceability of performance data during transmission and storage. The data hash is generated by concatenating the current data, a random number, and the hash value of the previous block, and a SHA-3 algorithm is used to generate a unique identifier;

[0033] Adopt a federated learning framework to aggregate anonymized training data from multiple venues, optimize the performance of the global AI model. The objective function is the weighted sum of the local losses of each venue, and a regularization term is added to prevent overfitting;

[0034] Support users to autonomously manage personal sports data through encrypted tokens (NFTs) and authorize third parties to access it.

[0035] As can be seen from the technical solution provided by the present invention above, a sports exercise AI standing long jump evaluation data management system provided by the present invention has the following beneficial effects:

[0036] High-precision evaluation and reliable determination: The dynamic environment calibration module gets rid of the dependence on fixed markers. Through the multi-spectral sensor array and deep learning algorithm, it accurately constructs a dynamic coordinate system in a complex environment, providing an accurate spatial reference for three-dimensional pose analysis; The three-dimensional pose analysis module uses multi-view cameras and advanced algorithms to achieve sub-centimeter accuracy measurement of the long jump movement, and real-time detects key indicators such as takeoff, in the air and landing; The intelligent violation determination engine combines the rule engine and the AI time series analysis model. Based on kinematic parameters and spatial trajectory features, it automatically and accurately identifies violation behaviors, and can also dynamically adjust the determination criteria according to the individual characteristics of users to ensure that the evaluation results are scientific, fair and reliable. Compared with traditional evaluation methods, it greatly improves the accuracy and credibility of the evaluation;

[0037] Efficient identity authentication and data security: The multi-modal identity recognition module integrates gait and face features to achieve contactless identity authentication, reducing the user operation process and improving the usability; The distributed data management unit uses blockchain technology to store data, ensuring data immutability and traceability; Edge computing realizes local processing, reducing the dependence on the cloud and improving data processing efficiency; The federated learning framework aggregates multi-site data to optimize the AI model while protecting data privacy; Users can also independently manage data permissions through NFTs, comprehensively protecting data security and preventing data leakage and abuse. Compared with similar systems, it has obvious advantages in data security and privacy protection;

[0038] Intelligent analysis and personalized services: Each module of the system collaborates to generate multi-dimensional data. Through in-depth analysis of the long jump movement, it provides users with detailed movement decomposition diagrams and improvement suggestions to help users specifically improve their sports performance; At the same time, based on the user's historical data and individual characteristics, it realizes dynamic adjustment of evaluation criteria and recommendation of personalized training plans to meet the diverse needs of different users, enhance user stickiness, and promote users' scientific exercise. Compared with traditional sports evaluation systems, it is more intelligent and personalized;

[0039] Wide applicability and technical expandability: The technical solution of this system can be quickly extended to the evaluation of other sports events such as high jump and shot put by adjusting the sensor layout, defining key points of pose analysis and loading pre-trained models, effectively reducing the development cost and improving the system versatility; The advanced technologies such as multi-spectral sensing, deep learning, and blockchain integrated in it provide innovative ideas and practical experiences for the technological development in the field of sports evaluation, promoting the application and expansion of related technologies in more fields and driving the intelligent upgrade of the sports industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1This is a schematic diagram of the overall structure of an AI standing long jump evaluation data management system for physical exercise according to the present invention. Detailed implementation manners

[0041] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0042] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0043] As Figure 1 shown, an embodiment of the present invention provides an AI standing long jump evaluation data management system for physical exercise, including:

[0044] Dynamic environment calibration module: Real-time collect site environment data through a multi-spectral sensor array deployed around the long jump area, and establish a dynamic coordinate system based on an adaptive calibration algorithm of deep learning. The adaptive calibration algorithm generates a virtual calibration grid by fusing environmental light intensity, ground texture features and obstacle distribution, and does not rely on fixed markers;

[0045] Multi-modal identity recognition module: Realize contactless identity authentication by fusing gait recognition and face feature extraction technologies and combining user biometric information in the cloud database;

[0046] Three-dimensional pose analysis module: Synchronously collect long jump action videos through multi-view cameras, construct a motion trajectory model through a 3D human pose estimation algorithm, and real-time detect the takeoff angle, in-air pose and landing stability;

[0047] Intelligent violation determination engine: Automatically identify violation behaviors including stepping jumps, crossing the line, and body imbalance based on a rule engine and an AI time series analysis model, combined with kinematic parameters and spatial trajectory features;

[0048] Distributed data management unit: Store evaluation data using blockchain technology, support multi-node synchronous verification and tamper-proof score records, and implement local data processing through an edge computing module to reduce cloud dependence.

[0049] In this embodiment, the implementation method of the dynamic environment calibration module includes:

[0050] Obtain real-time environmental information of the long jump area through a multi-spectral sensor array, including light intensity, ground texture and obstacle distribution;

[0051] Extract environmental features based on an improved Convolutional Neural Network (CNN), and dynamically generate virtual calibration grids adapted to different site conditions. The loss function of the improved CNN consists of two parts:

[0052] Feature reconstruction loss: Calculate the mean square error between the input environmental features and the reconstructed features, and the weight coefficient is used to control its importance;

[0053] Environmental robustness loss: Maximize the feature difference tolerance under light interference, and the weight coefficient is used to balance the anti-interference ability, and and sum to 1;

[0054] Optimize the calibration parameters through a reinforcement learning algorithm to ensure the stability of the coordinate system in complex environments such as shadows and reflections;

[0055] Furthermore, as the "spatial positioning center" of the AI standing long jump evaluation data management system for physical exercise, the dynamic environment calibration module realizes the dynamic perception of the long jump site environment and the construction of an accurate coordinate system through the deep integration of a multi-spectral sensor array and a deep learning algorithm, ensuring the accuracy and stability of the evaluation data collection. The following will elaborate on this module from aspects such as overall function, sub-module composition, key technical principles, working process, and application value:

[0056] I. Overview of overall function:

[0057] The dynamic environment calibration module uses a multi-spectral sensor array deployed around the long jump area to collect site environment data in real time, covering information such as light intensity, ground texture, and obstacle distribution. Based on the adaptive calibration algorithm of deep learning, the module can dynamically generate virtual calibration grids, construct a dynamic coordinate system suitable for complex environments, and provide an accurate spatial reference for subsequent functions such as motion capture and pose analysis. This module gets rid of the dependence on fixed markers in traditional calibration, effectively solves the interference of environmental changes (such as shadows and reflections) on calibration accuracy, and provides reliable basic data support for standing long jump evaluation;

[0058] II. Sub-module composition and functions:

[0059] (I) Environmental data acquisition unit:

[0060] Sensor deployment: Reasonably layout a multi-spectral sensor array around and above the long jump area. This array integrates a visible light camera and an infrared camera. The visible light camera is responsible for capturing visual information such as the texture and color of the site. The infrared camera obtains the thermal imaging data of the site under low light or complex light conditions, making up for the deficiencies of the visible light camera and realizing the all-round and all-weather monitoring of the site environment;

[0061] Real-time data acquisition: The sensor array continuously collects real-time environmental information in the long jump area at a high frame rate, preliminarily encodes the collected image data, light intensity data, temperature data, etc., and sends them to the data processing unit through a high-speed data transmission channel;

[0062] (2) Feature extraction and model construction unit:

[0063] Environmental feature extraction: Based on an improved convolutional neural network (CNN), extract features from the collected environmental data; The network structure adopts a combination of multiple convolutional layers, pooling layers and fully connected layers, and automatically extracts features such as light intensity change features, ground texture features and obstacle contour features through convolutional operations; The improved CNN introduces an attention mechanism to make the network focus on the key area features that have a greater impact on the coordinate system calibration, improving the efficiency and accuracy of feature extraction;

[0064] Loss function optimization: The loss function of the improved CNN is defined as (where is the feature reconstruction loss weight coefficient, used to control the influence degree of the mean square error between the input environmental features and the reconstructed features; is the environmental robustness loss weight coefficient, and ; , is the number of samples, is the input environmental feature, is the reconstructed feature; , is the number of test samples, is the tolerance threshold, used to enhance the anti-interference ability of the model to light changes, is the environmental feature after adding noise, is the original environmental feature); By optimizing this loss function, the accurate extraction of environmental features and the improvement of model robustness are realized;

[0065] Virtual calibration grid generation: According to the extracted environmental features, dynamically generate virtual calibration grids adapted to different site conditions; The density and accuracy of the grids are automatically adjusted according to the site complexity. In areas with more obstacles or drastic texture changes, the grid division is finer to ensure the accuracy of coordinate system construction;

[0066] (3) Parameter optimization and coordinate system establishment unit:

[0067] Reinforcement learning optimization: Use reinforcement learning algorithms to optimize the parameters of the virtual calibration grid; Define a reward function that gives a positive reward when the generated coordinate system can accurately match the actual spatial position in a complex environment, and a negative reward otherwise; Through continuous interaction and learning between the agent and the environment, adjust the calibration parameters to keep the coordinate system stable in complex environments such as shadows and reflections;

[0068] Dynamic coordinate system establishment: Based on the optimized virtual calibration grid, construct a dynamic coordinate system; This coordinate system takes the fixed reference point in the long jump area as the origin, and updates the axis directions and scales in real time according to environmental changes, providing a unified and accurate spatial coordinate benchmark for subsequent functions such as three-dimensional pose analysis and violation determination;

[0069] III. Key technical principles:

[0070] (I) Multi-spectral data fusion principle:

[0071] The visible light and infrared data collected by the multi-spectral sensor array have different characteristics and advantages; Through data fusion technology, the two types of data are fused at the pixel level, feature level, or decision level; In pixel-level fusion, the corresponding pixels of the visible light image and the infrared image are directly weighted and combined; In feature-level fusion, the features of the two images are extracted separately and then fused; Decision-level fusion is based on the analysis results of the two images for comprehensive decision-making; Through data fusion, more comprehensive and accurate environmental information is obtained, providing a high-quality data basis for subsequent calibration algorithms;

[0072] (II) Improved CNN feature extraction principle:

[0073] The improved CNN is based on the basic architecture of the convolutional neural network. By introducing an attention mechanism and a special network layer design, it enhances the ability to extract environmental features; The attention mechanism enables the network to automatically learn the importance of features in different regions, focusing on regions crucial for coordinate system calibration, such as the long jump take-off line and the boundary of the landing area; Special network layer designs, such as residual connections and dilated convolutions, effectively solve the problem of gradient disappearance caused by increasing network depth, improving the network's ability to express complex environmental features;

[0074] (III) Reinforcement learning parameter optimization principle:

[0075] Reinforcement learning adjusts the strategy by continuously trial-and-error and learning of the agent in the environment, with the goal of maximizing the cumulative reward; In the dynamic environment calibration module, the agent adjusts the strategy for calibration parameters, and the environment is the complex environment of the long jump field; The agent takes actions to adjust the calibration parameters according to the current environmental state (such as light intensity, obstacle distribution), obtains the corresponding reward feedback, and gradually finds the optimal calibration parameter strategy through continuous learning, keeping the dynamic coordinate system stable in various environments;

[0076] IV. Workflow of the Module:

[0077] (I) Initialization Phase:

[0078] After the dynamic environment calibration module is started, it loads the initial parameters of the improved CNN model, the initial policy of the reinforcement learning algorithm, and the preset virtual calibration grid template; establishes data communication connections with the multispectral sensor array and other modules of the system to prepare for receiving environmental data and control instructions;

[0079] (II) Data Acquisition Phase:

[0080] The multispectral sensor array continuously acquires environmental data in the long jump area at a set frequency and transmits the data to the environmental data acquisition unit in real time; this unit performs format conversion and preliminary verification on the data, and after ensuring the integrity and accuracy of the data, sends it to the feature extraction and model construction unit;

[0081] (III) Feature Extraction and Model Calculation Phase:

[0082] The feature extraction and model construction unit uses the improved CNN to extract features from the environmental data, calculates the loss function and performs backpropagation to optimize the network parameters and improve the accuracy of feature extraction; based on the extracted features, a virtual calibration grid is dynamically generated, and the grid parameters are optimized through the reinforcement learning algorithm;

[0083] (IV) Coordinate System Establishment Phase:

[0084] The parameter optimization and coordinate system establishment unit constructs a dynamic coordinate system based on the optimized virtual calibration grid; synchronizes the coordinate system information to other modules such as the three-dimensional attitude analysis module and the intelligent violation determination engine of the system to provide a spatial coordinate reference for subsequent evaluation functions;

[0085] (V) Dynamic Update Phase:

[0086] During the operation of the system, continuously monitor environmental changes. When environmental changes such as sudden changes in light intensity and movement of obstacles are detected, restart the feature extraction, model calculation, and parameter optimization processes, and update the dynamic coordinate system in real time to ensure that the coordinate system always adapts to environmental changes;

[0087] (VI) End Phase:

[0088] When the system stops running or receives a stop instruction, the dynamic environment calibration module stops data acquisition and calculation operations, saves the current calibration parameters and model status, closes the communication connection with other modules, and releases system resources;

[0089] V. Application Value of the Module:

[0090] (I) Improving Evaluation Accuracy:

[0091] Breaking free from the limitations of fixed markers, the dynamic environment calibration module can accurately construct a coordinate system in a complex and ever-changing environment, providing an accurate spatial reference for functions such as three-dimensional pose analysis and violation determination, ensuring the accuracy of standing long jump evaluation data, and reducing evaluation errors caused by environmental factors;

[0092] (2) Enhancing system adaptability:

[0093] Through dynamic perception and adaptation to different lighting conditions, ground textures, and obstacle distributions, the module enables the entire evaluation system to operate stably in various site environments, expanding the application scenarios of the system and enhancing the versatility and practicality of the system;

[0094] (3) Reducing maintenance costs:

[0095] Traditional calibration methods rely on fixed markers, which are prone to damage and loss, and require frequent maintenance and calibration; the dynamic environment calibration module does not require fixed markers, reducing the maintenance cost and calibration workload of the markers, improving the maintenance efficiency of the system, and reducing the long-term use cost.

[0096] In this embodiment, the multimodal identity recognition module, as the "digital gatekeeper" of the sports exercise AI standing long jump evaluation data management system, realizes contactless and high-precision user identity authentication through the integration of multiple biometric recognition technologies and cloud data interaction, ensuring the one-to-one correspondence between evaluation data and user information, and enhancing system security and data reliability; the following will elaborate on this module from aspects such as overall function, sub-module composition, key technical principles, working process, and application value:

[0097] I. Overview of overall function:

[0098] The multimodal identity recognition module comprehensively applies gait recognition and face feature extraction technologies, combines with the user's biometric information in the cloud database, and automatically completes identity authentication when the user enters the long jump area; the module does not require the user to operate actively, and quickly and accurately identifies the user's identity through the dynamic behavior data and static biometric data collected by multiple sensors, avoiding the cumbersome process and errors of manual registration; at the same time, the module has the ability to dynamically update the user's biometric data to adapt to the feature changes of individuals over time, laying a foundation for the accurate recording of subsequent evaluation data, personalized analysis, and user sports file management;

[0099] II. Composition and functions of sub-modules:

[0100] (1) Biometric collection unit:

[0101] Sensor Deployment: Deploy multi-view cameras and depth sensors at the entrance and surrounding areas of the long jump area; the multi-view cameras capture the user's facial images and full-body motion postures from different angles; the depth sensors obtain the three-dimensional spatial information of each part of the user's body, providing depth data support for gait recognition to ensure the integrity and accuracy of biometric data collection;

[0102] Real-time Data Collection: When the user enters the collection range, the sensor array synchronously collects multi-modal data at a high frame rate, including color face image sequences, dynamic videos of human body contours, joint movement trajectories within the gait cycle, etc., and encodes and transmits the original data to the feature extraction and analysis unit;

[0103] (2) Feature Extraction and Analysis Unit:

[0104] Face Feature Extraction: Use a convolutional neural network (CNN) based on deep learning to extract features from the collected face images; through multi-layer convolutional and pooling layer operations, extract key feature points of the face (such as the shapes and relative positions of eyes, nose, mouth, etc.), texture information (such as wrinkles, skin color, etc.), and generate high-dimensional face feature vectors; use a pre-trained face recognition model (such as FaceNet), and combine transfer learning techniques to fine-tune the model in the system application scenario to improve the accuracy and robustness of face feature extraction;

[0105] Gait Feature Extraction: Use a spatio-temporal graph convolutional network (ST-GCN) to process the human motion video and capture the spatio-temporal features of gait; construct a spatio-temporal graph from the spatial positions and time series information of human joint points in consecutive frames, and extract dynamic features such as gait cycle, stride length, and swing angle through graph convolutional operations to generate gait feature vectors; at the same time, introduce an attention mechanism to make the network focus on gait features that are crucial for identity recognition, such as unique walking postures and body swing amplitudes;

[0106] Feature Fusion: Adopt a weighted fusion strategy to fuse the extracted face feature vectors and gait feature vectors; dynamically adjust the fusion weights according to the recognition accuracies of the two features in different scenarios; for example, in an environment with sufficient light and clear faces, appropriately increase the weight of face features; in the case of long distance or face occlusion, increase the weight of gait features to generate the final multi-modal feature vector;

[0107] (3) Identity Authentication and Data Management Unit:

[0108] Cloud data comparison: Send the fused multi-modal feature vectors to the cloud database for comparison with the stored user biometric templates; use methods such as cosine similarity calculation or Euclidean distance measurement to evaluate the similarity between feature vectors; set an identity authentication threshold, and when the similarity exceeds the threshold, determine that the user identity match is successful; if the threshold is not reached, trigger a secondary authentication process, such as prompting the user to enter the registered mobile phone number or verifying other alternative biometric features;

[0109] User data management: After successful authentication, the module binds the user identity information with the current evaluation data and updates the relevant data in the user's exercise profile; at the same time, if it is detected that there are certain differences between the user's biometric features (such as minor changes in facial features caused by hairstyle changes, gait changes caused by long-term exercise) and the original template and meet the update conditions, automatically upload the new feature data to the cloud database to replace the old template and achieve dynamic update of the user's biometric features;

[0110] Security protection: Use encryption algorithms (such as AES-256) to encrypt the transmitted biometric data to prevent the data from being stolen or tampered with during transmission; in the cloud database storage link, use blockchain technology to store the user identity information in a distributed manner to ensure the immutability and traceability of the data and improve the security of the identity authentication system;

[0111] III. Key technical principles:

[0112] (I) Multi-modal biometric fusion principle:

[0113] Multi-modal biometric fusion is based on the concept of information complementarity, combining the advantages of two biometric features, face and gait; facial features have stability and uniqueness, suitable for identity recognition under close-range and clear images; gait features can be collected at a distance and in a non-cooperative state, and contain individual behavior habit information; through the feature-level fusion method, the two features are integrated at the data level to generate a feature vector containing more identity information, making up for the recognition defects of single-modal features in complex scenarios and improving the accuracy and reliability of identity authentication;

[0114] (II) Deep learning feature extraction principle:

[0115] Deep learning models (such as CNN, ST-GCN) automatically learn complex patterns and feature representations in the data through multi-layer neural networks; in facial feature extraction, CNN extracts semantic features at different levels through the sliding operation of convolutional kernels on the image; in gait feature extraction, ST-GCN regards human motion as a spatio-temporal graph and captures the dynamic relationship of joint points in the spatial and temporal dimensions through graph convolution operations; the model is trained with a large amount of labeled data to optimize the network parameters, enabling the network to accurately extract features that are discriminative for identity recognition;

[0116] (III) Cloud Collaboration Authentication Principle:

[0117] The cloud database stores a large number of user biometric templates. During the identity authentication process, the local module uploads the extracted feature vectors to the cloud for comparison; the cloud server uses distributed computing and parallel processing technologies to quickly retrieve and match user templates and return the authentication result; at the same time, a two-way data synchronization mechanism is established between the cloud and the local module to ensure the real-time update and consistency of user biometric data, realizing seamless identity authentication across devices and scenarios;

[0118] IV. Workflow of the Module:

[0119] (I) Initialization Phase:

[0120] After the multi-modal identity recognition module is started, it loads the pre-trained face and gait recognition model parameters, establishes a secure communication connection with the cloud database, and initializes the feature fusion weight parameters and identity authentication threshold; it completes the self-check and calibration of the sensor array to ensure the normal operation of the acquisition device;

[0121] (II) Data Acquisition Phase:

[0122] When it detects that the user enters the long jump area, the multi-view cameras and depth sensors of the biometric acquisition unit start to synchronously collect face images and human motion data, and transmit the collected raw data to the feature extraction and analysis unit at a fixed frequency;

[0123] (III) Feature Extraction and Fusion Phase:

[0124] The feature extraction and analysis unit processes the face image and human motion video respectively, extracts the face feature vector and gait feature vector; adopts a dynamic weighted fusion strategy to merge the two feature vectors into a multi-modal feature vector to provide data support for subsequent identity authentication;

[0125] (IV) Identity Authentication Phase:

[0126] Send the multi-modal feature vector to the cloud database to calculate the similarity with the stored user biometric template; if the similarity is higher than the set threshold, it is determined that the identity authentication is successful and the user identity information is obtained; if the similarity does not meet the standard, trigger the secondary authentication process, requiring the user to provide additional identity verification information until the authentication is passed or the authentication operation is ended;

[0127] (V) Data Binding and Update Phase:

[0128] After successful authentication, bind the user identity information with the standing long jump evaluation data of this time and record it in the user's sports file; at the same time, analyze the difference between the newly collected biometric features and the original template. If the difference reaches the update standard, upload the new feature data to the cloud database to update the user's biometric template;

[0129] (6) End stage:

[0130] When the system stops running or receives a stop instruction, the multi-modal identity recognition module stops data collection and authentication operations, disconnects from the cloud database, releases system resources, and saves the current model parameters and configuration information;

[0131] V. Application value of the module:

[0132] (1) Improve user experience:

[0133] Implement contactless and automated identity authentication. Users do not need to carry identity credentials or perform manual operations, reducing waiting time, improving the convenience and fluency of using the system, and providing users with more intelligent and efficient sports evaluation services;

[0134] (2) Ensure data accuracy:

[0135] Through multi-modal biometric recognition, ensure the accurate matching of evaluation data and user identity, avoid data confusion or incorrect recording, provide a reliable data basis for user sports performance analysis and training effect evaluation, and make sports guidance and suggestions more targeted;

[0136] (3) Enhance system security:

[0137] Adopt encryption transmission and blockchain storage technologies to protect the security of user biometric data and prevent identity information leakage and misappropriation; the multi-modal feature fusion and dynamic update mechanism improve the anti-counterfeiting ability of identity authentication, effectively resist forgery attacks, and ensure the security and stability of system operation;

[0138] (4) Support personalized services:

[0139] Based on accurate identity authentication and user sports file management, the system can provide different users with personalized sports plans, training suggestions and performance analysis reports, meet the diverse sports needs of users, promote users to continuously use the system for physical exercise, and improve user stickiness and satisfaction.

[0140] In this embodiment, the three-dimensional pose analysis module further includes:

[0141] Use multi-view cameras to synchronously capture stereo images of the long jump action and reconstruct the three-dimensional motion trajectory through an improved stereo matching algorithm. The energy function of the algorithm is composed of the following three items weighted:

[0142] Data item: Evaluate the consistency of three-dimensional pose through the projection error of multi-view images;

[0143] Smoothing term: Constrain the continuity of adjacent joint points in space;

[0144] Temporal term: Ensure the physical rationality of the pose change between adjacent frames, and the weight coefficients 、 、 respectively control the contribution ratios of the three terms;

[0145] Analyze the joint motion sequence based on the Temporal Convolutional Network (TCN) to predict the centroid offset at the moment of landing;

[0146] Combine the data of the Inertial Measurement Unit (IMU), correct the acceleration error of the aerial pose through a filtering algorithm, and achieve sub-centimeter-level accuracy measurement;

[0147] Furthermore, as the "motion digital mirror" of the AI standing long jump evaluation data management system for physical exercise, the three-dimensional pose analysis module performs high-precision three-dimensional modeling and dynamic analysis of the long jump action through the fusion of multi-view visual perception and advanced algorithms, providing key motion data support for the evaluation system; The following will elaborate in detail from aspects such as overall function, sub-module composition, key technical principles, working process, and application value:

[0148] I. Overview of overall function:

[0149] The three-dimensional pose analysis module uses multi-view cameras to synchronously collect long jump action videos, constructs a motion trajectory model with the help of 3D human pose estimation algorithms, and can detect key indicators such as takeoff angle, aerial pose, and landing stability in real time and accurately; This module can not only capture the motion trajectories of each joint point of the human body in three-dimensional space, but also quantitatively evaluate the action quality in combination with kinematic principles, provide action detail feedback for users, and at the same time provide core data for intelligent violation determination to ensure the scientificity and fairness of the evaluation results;

[0150] II. Composition and functions of sub-modules:

[0151] (I) Multi-view image acquisition unit:

[0152] Device deployment: Deploy multi-view cameras in a circular or multi-angle distribution around the long jump area to ensure that the takeoff point, airborne path, and landing area are all within the shooting field of view; The camera parameters are uniformly calibrated, including shooting frame rate, resolution, focal length, etc., to ensure the time synchronization and spatial consistency of multi-view images;

[0153] Data acquisition: When the user starts the long jump motion, multi-view cameras synchronously acquire motion videos at a high frame rate (such as 60fps), and send the original image data to the data processing unit in real time through a high-speed data transmission link, providing the original material for subsequent posture analysis;

[0154] (2) Three-dimensional motion trajectory reconstruction unit:

[0155] Stereo matching processing: An improved stereo matching algorithm is used to process multi-view images, and temporal consistency constraints are introduced to optimize the matching accuracy; its energy function is (where represents the three-dimensional pose parameters of the th frame, including the coordinates of each joint point of the human body in three-dimensional space; is a data item, and the image consistency is measured by calculating the projection error of multi-view images. The smaller the error, the more accurate the matching; is a smoothing term, which is used to constrain the continuity of adjacent joint points in spatial position and avoid trajectory mutations; is a temporal term, which ensures that the posture changes between adjacent frames conform to physical laws and prevents jumps that do not conform to motion logic; , , are weight coefficients, which are used to adjust the relative importance of each energy term and are optimized and determined through a large amount of experimental data); by minimizing the energy function, precise matching of corresponding points in multi-view images is achieved, and then the motion trajectory of the human body in three-dimensional space is reconstructed;

[0156] Trajectory optimization: The reconstructed three-dimensional trajectory is filtered to remove abnormal data points caused by noise or matching errors, and the Kalman filter or median filter algorithm is used to smooth the trajectory curve to make the motion trajectory more fitting;

[0157] (3) Key index analysis unit:

[0158] Takeoff angle calculation: According to the three-dimensional pose data at the takeoff instant reconstructed, calculate the spatial vector angle between the center of gravity of the human body and the takeoff point and the landing point, and determine the takeoff angle in combination with the ground coordinate system; by analyzing the deviation between the takeoff angle and the ideal angle, evaluate the rationality of the takeoff motion;

[0159] In-air posture evaluation: Based on the Temporal Convolutional Network (TCN), analyze the joint motion sequence and predict the centroid offset at the landing instant; the TCN network can effectively capture the time dependence of the motion sequence, learn the law of the human body's in-air posture change, and judge whether there is posture imbalance or unreasonable motion;

[0160] Landing stability detection: Combining inertial measurement unit (IMU) data with three-dimensional trajectory information, using the Kalman filter algorithm to correct the acceleration error of the aerial attitude, and achieving sub-centimeter-level precision measurement; By analyzing the joint angles, body inclination, and impact force distribution at the moment of landing, evaluating the landing stability, and judging whether there are risks such as falling and slipping;

[0161] (4) Data output unit:

[0162] Format the analyzed data such as takeoff angle, aerial attitude parameters, and landing stability indicators, and output them in the form of numbers, charts, etc.; At the same time, transmit the key data to the intelligent violation determination engine and the result display module, providing a basis for violation determination and an intuitive action analysis report for users;

[0163] III. Key technical principles:

[0164] (1) Multi-view stereo matching principle:

[0165] Multi-view stereo matching is based on the principle of triangulation. By finding the corresponding pixels of the same object point in multiple images, the three-dimensional coordinates are calculated using the internal and external parameters of the camera; The improved stereo matching algorithm introduces a temporal consistency constraint, considering the motion continuity between adjacent frames, constructing an energy function by combining the data term, smooth term, and temporal term, and solving the optimal matching result through an optimization algorithm to improve the accuracy and stability of three-dimensional reconstruction;

[0166] (2) Temporal Convolutional Network (TCN) analysis principle:

[0167] The Temporal Convolutional Network (TCN) is a deep learning model specialized for processing sequence data. Through causal convolution and dilated convolution operations, it can effectively capture long-range temporal dependencies without losing the time series information; In three-dimensional pose analysis, TCN takes the joint motion sequence as input, learns the patterns and rules of the action changing over time, and realizes the prediction of future poses and the evaluation of the current pose;

[0168] (3) Kalman filter correction principle:

[0169] The Kalman filter is an optimal estimation algorithm based on the linear system state space model. It estimates the system state through two steps: prediction and update; In three-dimensional pose analysis, the Kalman filter is used to fuse IMU data and visual measurement data to correct the acceleration error of the aerial attitude; According to the state equation and observation equation of the system, continuously update the state estimation value to improve the accuracy of pose measurement;

[0170] IV. Working process of the module:

[0171] (1) Initialization stage:

[0172] After the 3D attitude analysis module is started, it completes the parameter configuration and calibration of the multi-view cameras, loads the initial parameters of the stereo matching algorithm, TCN model, and Kalman filtering algorithm; establishes a data communication channel with other modules to prepare for receiving image data and control commands;

[0173] (2) Data acquisition stage:

[0174] The multi-view image acquisition unit monitors the long jump area in real time. When a user's takeoff action is detected, the multi-view cameras are synchronously started for acquisition, and the video data is quickly transmitted to the 3D motion trajectory reconstruction unit;

[0175] (3) Trajectory reconstruction stage:

[0176] The 3D motion trajectory reconstruction unit performs stereo matching processing on the multi-view images, calculates the 3D attitude parameters by optimizing the energy function, reconstructs the human motion trajectory, and performs filtering optimization to remove data noise and outliers;

[0177] (4) Key index analysis stage:

[0178] The key index analysis unit calculates the takeoff angle, analyzes the in-air attitude, and detects the landing stability based on the reconstructed trajectory data; uses the TCN network to predict the centroid offset, and combines IMU data and the Kalman filtering algorithm to improve the measurement accuracy;

[0179] (5) Data output stage:

[0180] The data output unit sorts and formats the analysis results, outputs them to the intelligent violation determination engine and the result display module, and stores the key data for subsequent query and analysis at the same time;

[0181] (6) End stage:

[0182] When the system stops running or receives a stop command, the 3D attitude analysis module stops data acquisition and processing operations, saves the algorithm parameters and intermediate results, closes the communication connection with other modules, and releases system resources;

[0183] V. Application value of the module:

[0184] (1) Improving the professionalism of evaluation:

[0185] Through high-precision 3D attitude analysis, various technical indicators of the long jump action can be quantified, providing users with professional and detailed action analysis reports, helping users understand their own action defects, improving training methods targeted, and enhancing sports performance;

[0186] (2) Enhancing the accuracy of determination:

[0187] Provide reliable data support for intelligent violation determination. Based on accurate analysis of takeoff angle, aerial posture, and landing stability, it can accurately identify violation behaviors such as stepping jumps and crossing the line, ensuring the fairness and authority of the evaluation results;

[0188] (III)Support personalized training:

[0189] According to the movement characteristics and evaluation data of each user, the system can formulate personalized training plans for users, recommend suitable training intensities and methods, meet the training needs of different users, and improve training effects and user satisfaction;

[0190] (IV)Promote technological innovation:

[0191] This module integrates advanced technologies such as multi-perspective visual perception, deep learning, and sensor fusion, provides practical experience and innovative ideas for the technological development in the field of sports evaluation, and promotes the application and expansion of related technologies in other sports events.

[0192] In this embodiment, the determination method of the intelligent violation determination engine includes:

[0193] Model the continuous-frame action sequence through an improved LSTM network, and the output of its forgetting gate is fused with predefined kinematic constraints, and the constraint weight coefficient γ controls the contribution of physical priors;

[0194] Define a dynamic threshold rule library, adjust the determination criteria according to the user's age, gender, and historical data, and the threshold is the weighted average of the reference threshold and the adaptive offset, and the balance coefficient μ controls the weights of the two;

[0195] Generate an interpretable violation report, including action decomposition diagrams and improvement suggestions, and provide real-time feedback to the user side through AR devices;

[0196] Furthermore, as the "rule guardian" of the AI standing long jump evaluation data management system for physical exercise, the intelligent violation determination engine deeply integrates the rule engine and the AI time series analysis model, and based on kinematic parameters and spatial trajectory features, automatically and intelligently determines violation behaviors during the long jump process. The following will elaborate on this module from aspects such as overall function, sub-module composition, key technical principles, working process, and application value:

[0197] I. Overview of overall function:

[0198] The intelligent violation determination engine automatically identifies violation behaviors such as taking a hop step, crossing the line, and body imbalance by analyzing the real-time human motion data and spatial trajectory information during the long jump process, and combining with a predefined rule library and an AI algorithm model. The engine can not only quickly and accurately determine the type of violation, but also dynamically adjust the determination criteria based on individual user characteristics (age, gender, historical data, etc.) to ensure the scientificity and fairness of the determination results. At the same time, it generates a visual and interpretable violation report and feeds it back to the user in real time through an AR device to help the user understand the reason for the violation in time and improve the movement.

[0199] II. Sub-module composition and functions:

[0200] (1) Data reception and preprocessing unit:

[0201] Data access: Establish data communication interfaces with other modules in the system such as the three-dimensional pose analysis module and the dynamic environment calibration module to receive in real time the human three-dimensional pose data (joint angles, centroid position, etc.), motion trajectory data, takeoff point and landing point coordinate data, and environment calibration information during the long jump process.

[0202] Data cleaning and conversion: Clean the received data to remove noise and outliers, such as unreasonable data points caused by sensor errors or matching errors. At the same time, uniformly convert data in different formats into a standard format that meets the processing requirements of the engine to provide a high-quality data basis for subsequent analysis.

[0203] (2) Violation behavior analysis unit:

[0204] Action sequence modeling: Use an improved LSTM network to model the continuous frame action sequence. Its gating mechanism introduces kinematic prior knowledge, and the update formula is (where is the output of the forget gate, used to control the retention degree of the information at the previous moment; is the weight matrix of the forget gate, which determines the influence weight of the input information on the output of the forget gate; is the hidden state at the previous moment, containing the information of the previous action sequence; is the input at the current moment, that is, the human motion data of the current frame; is the bias vector of the forget gate used to adjust the benchmark of the forget gate output; is a predefined kinematic constraint matrix, storing kinematic rule information such as joint angle range and centroid velocity limit; is the constraint weight coefficient, which controls the influence degree of the kinematic constraint on the output of the forget gate; is the activation function, usually using the sigmoid function, which maps the output value to between 0 and 1); Through this model, learn the time series characteristics of the long jump action and capture the coherence and abnormal changes of the action.

[0205] Dynamic threshold calculation: Define a dynamic threshold rule library, adjust the judgment criteria according to the user's age, gender, and historical data. The threshold calculation formula is (where is the final judgment threshold; is the reference threshold, a general standard obtained by statistically analyzing a large amount of normal long jump action data; is the adaptive offset based on the user's historical data, reflecting the individual movement characteristics and ability differences of the user; is the balance coefficient, and the value range is , which is used to adjust the weights of the reference threshold and the adaptive offset, and can be dynamically adjusted according to actual needs); Through the dynamic threshold, the judgment criteria can be more in line with the individual situation of the user and reduce misjudgment;

[0206] Identification of violation behaviors: Compare and analyze the processed data with the dynamic threshold and kinematic rules; For example, detect abnormal movement of the center of mass before takeoff to determine whether there is a hop step; Determine whether the takeoff line is crossed according to the position relationship between the takeoff point coordinates and the takeoff line; Evaluate whether there is body imbalance based on the change of body joint angles and the stability of the center of mass; Once it is found that the data exceeds the threshold range or does not conform to the kinematic rules, it is determined that a violation behavior has occurred, and the type and time point of the violation are recorded;

[0207] (3) Report generation and feedback unit:

[0208] Generation of violation reports: When a violation behavior is detected, generate a detailed violation report; The report content includes a breakdown diagram of the violation action, marking the specific time node, body part, and action deviation of the violation; At the same time, combined with the kinematic principles and rules, give an analysis of the cause of the violation behavior and improvement suggestions to help the user understand the problem and conduct targeted training;

[0209] Real-time feedback: Display the violation report to the user in real time through the AR device. The user can intuitively see the virtual annotation and prompt information of their own violation actions, realizing "feedback immediately after jumping" and correcting wrong actions in a timely manner; In addition, synchronously transmit the violation judgment result to the data management module of the system for subsequent performance recording and statistical analysis;

[0210] III. Key technical principles:

[0211] (1) Principle of improved LSTM network:

[0212] Based on the traditional LSTM structure, the improved LSTM network incorporates kinematic prior knowledge into the gating mechanism. The calculation of the forget gate, input gate, and output gate is intervened through the kinematic constraint matrix, so that the network can follow the physical laws of human movement when learning action sequences. For example, when the joint angle exceeds the normal range, the network will adjust the information transmission according to the constraint matrix, more accurately identify abnormal action patterns, and improve the detection ability of illegal behaviors.

[0213] (II) Principle of dynamic threshold adjustment:

[0214] Dynamic threshold adjustment is based on statistics and personalized learning concepts; the baseline threshold provides a general judgment standard, and the adaptive offset analyzes the user's historical data (such as action parameters and performance of multiple long jumps) to explore the user's individual action characteristics and ability level; the balance coefficient It is used to balance the relationship between the two, so that the threshold can reflect the general rules and adapt to individual differences, so as to achieve accurate judgment;

[0215] (III) Principles of integration of rule engine and AI:

[0216] The rule engine pre-sets the basic rules and kinematic constraints of the long jump competition, such as take-off rules and landing requirements; the AI ​​time series analysis model (such as the improved LSTM network) learns complex action patterns and abnormal features from a large amount of action data; after the two are integrated, the rule engine provides a clear judgment boundary for the AI ​​model, and the AI ​​model makes up for the shortcomings of the rule engine in recognizing complex action patterns, forming complementary advantages and achieving comprehensive and accurate judgment of violations;

[0217] 4. Module workflow:

[0218] (I) Initialization phase:

[0219] After the intelligent violation judgment engine is started, it loads the pre-trained improved LSTM network model parameters, dynamic threshold rule library, kinematic constraint matrix and other configuration information; establishes data communication connection with other modules in the system, and prepares to receive relevant data during the long jump;

[0220] (II) Data reception and processing stage:

[0221] The data receiving and preprocessing unit receives data from the three-dimensional posture analysis module and the like in real time, performs cleaning and format conversion, and transmits the processed data to the violation analysis unit;

[0222] (III) Violation determination stage:

[0223] The violation behavior analysis unit uses an improved LSTM network to model and analyze the action sequence, combines the dynamically calculated threshold and kinematic rules to identify violation behaviors; once a violation is determined, it records information such as the type and time of the violation, and transmits the result to the report generation and feedback unit;

[0224] (4) Report generation and feedback stage:

[0225] The report generation and feedback unit generates a detailed report based on the violation information, real-time feedbacks the report to the user through the AR device, and synchronizes the determination result to other modules of the system at the same time;

[0226] (5) End stage:

[0227] When the system stops running or receives a stop instruction, the intelligent violation determination engine stops data processing and determination operations, saves the model parameters and determination records, closes the communication connection with other modules, and releases system resources;

[0228] V. Application value of the module:

[0229] (1) Ensure the fairness of evaluation:

[0230] Through an automated and standardized violation determination process, eliminate the interference of human factors, ensure that all users are evaluated under the same rules, guarantee the fairness and impartiality of the evaluation results, and enhance the credibility of the evaluation system;

[0231] (2) Improve the training effect of users:

[0232] Real-time and intuitive violation feedback helps users promptly discover and correct action errors, deepen the understanding and memory of correct actions, accelerate the transformation of training results, and effectively improve the long jump technical level and sports performance of users;

[0233] (3) Reduce operating costs:

[0234] Reduce the need for manual referees, lower labor costs and time costs; at the same time, the high efficiency and accuracy of intelligent determination reduce repeated evaluations or dispute handling caused by determination disputes, and improve the operation efficiency of the system;

[0235] (4) Promote the scientific development of sports:

[0236] Accumulate a large amount of violation data and action analysis results, provide data support for sports training research and sports rule optimization, help deeply understand the technical key points and violation causes of long jump sports, and promote the scientific development of sports training and competition rules.

[0237] In this embodiment, the distributed data management unit further includes:

[0238] The blockchain-based smart contract mechanism ensures the integrity and traceability of performance data during transmission and storage. The data hash is generated by concatenating the current data, a random number, and the hash value of the previous block, and using the SHA-3 algorithm to generate a unique identifier;

[0239] Adopt the federated learning framework to aggregate anonymized training data from multiple sites, optimize the performance of the global AI model. The objective function is the weighted sum of the local losses of each site, and a regularization term is added to prevent overfitting;

[0240] Support users to autonomously manage personal sports data through encrypted tokens NFT and authorize third parties to access;

[0241] Furthermore, the distributed data management unit, as the "data center guardian" of the AI standing long jump evaluation data management system for physical exercise, comprehensively uses technologies such as blockchain and federated learning to achieve secure storage, efficient processing, and collaborative optimization of evaluation data, ensuring the integrity, traceability, and privacy of the data; The following will elaborate on this module from aspects such as overall function, sub-module composition, key technical principles, work process, and application value:

[0242] I. Overview of overall function:

[0243] The distributed data management unit uses blockchain technology to store standing long jump evaluation data, ensuring that the data cannot be tampered with and is traceable during transmission and storage. At the same time, a multi-node synchronization verification mechanism is used to ensure the accuracy and consistency of the data; With the help of the edge computing module, local data processing is realized, reducing the dependence on the cloud and improving the data processing efficiency; In addition, the federated learning framework is used to aggregate anonymized training data from multiple sites, optimize the performance of the global AI model, support users to autonomously manage personal sports data through encrypted tokens (NFT), realize the secure sharing and authorized access of data, and provide solid support for the stable operation of the system and the mining of data value;

[0244] II. Composition and functions of sub-modules:

[0245] (I) Blockchain data storage unit:

[0246] Data on the chain: When the standing long jump evaluation is completed, the evaluation data (including user identity information, performance data, motion analysis data, violation determination results, etc.) is packaged to generate a data block; The data block contains information such as data content, timestamp, and the hash value of the previous block;

[0247] Hash generation and verification: Use the hash generation algorithm (where, represents the data concatenation operation, which is used to combine different types of data together; is a random number, and its value is continuously changed to adjust the hash value to ensure that the hash result meets specific requirements; is the hash value of the previous block, which is used to construct the chain structure of the blockchain to ensure the continuity and traceability of data; is the evaluation data stored in the current block, covering various types of information related to users; is the generated hash value of the current block, which is the unique identifier of the data in this block) performs a hash operation on the data block to generate a unique hash value; through a multi-node synchronization verification mechanism, each node verifies the hash value of the data block to ensure that the data has not been tampered with;

[0248] Smart contract execution: Based on the smart contract mechanism of the blockchain, rules for data storage, access, modification, etc. are preset in advance; when specific conditions are met (such as data being uploaded to the chain, user authorization for access, etc.), the corresponding contract terms are automatically executed to ensure the integrity and traceability of the score data during transmission and storage, and prevent illegal operations;

[0249] (2) Federated learning optimization unit:

[0250] Data collection and anonymization: Establish connections with the evaluation systems of multiple sites to collect anonymized training data generated at each site, including user long jump motion data, environmental data, violation data, etc.; perform anonymization processing on the collected data to remove user sensitive information and only retain the feature data related to model training to ensure data privacy and security;

[0251] Model collaborative training: Adopt a federated learning framework. Each site trains an AI model (such as models for dynamic environment calibration, three-dimensional pose analysis, etc.) locally based on its own data and only uploads the model parameters (instead of the original data) to the central server; the central server uses the objective function (where, is the model parameter, which is the variable that needs to be optimized for the models trained at each site; is the data volume of the th site, reflecting the contribution of the data at this site to the overall training; is the total data volume of all sites, which is used to calculate the proportion of the data volume of each site; is the local loss function of the th site, measuring the training error of the model on the data at this site; is the regularization coefficient, which is used to control the model complexity and prevent overfitting) to aggregate and update the model parameters uploaded by each site, generate globally optimized model parameters, and distribute them to each site; each site uses the updated parameters to continue training to optimize the performance of the global AI model through multiple rounds of iteration;

[0252] Model Evaluation and Feedback: Regularly evaluate the optimized global model, and judge the model performance by its performance on the test dataset (such as evaluation accuracy, violation determination accuracy, etc.); According to the evaluation results, feedback the model training effect to each venue, and guide each venue to adjust the training strategy or data processing method to further improve the model performance;

[0253] (3) Data Permission Management Unit:

[0254] NFT Token Generation: Generate encrypted tokens (NFTs) for each user as the unique identifier and management voucher for the user's personal sports data; The NFT token is bound to the user's biometric information or account information to ensure the uniqueness and non-replicability of the token;

[0255] Permission Management and Authorization: Users can independently manage the access permissions of their personal sports data through the NFT token, and can authorize third parties (such as sports coaches, research institutions, etc.) to access some or all of the data; During the authorization process, the system records the authorization information, including the authorized object, authorized data range, authorization time, etc., to ensure the legality and controllability of data access;

[0256] Data Access Control: When a third party applies to access user data, the system verifies its access permissions; Only with the effective authorization of the user can the third party access the corresponding data, and the access process is strictly monitored to prevent data leakage or abuse;

[0257] (4) Edge Computing Processing Unit:

[0258] Local Data Processing: Deploy an edge computing module at the data generation end (such as the device end of the long jump venue) to perform preliminary processing on the collected raw data (such as multi-spectral sensor data, camera video data, etc.), including data cleaning, feature extraction, format conversion, etc., to reduce the data transmission volume and reduce the cloud computing pressure;

[0259] Real-time Response Support: For some tasks with high real-time requirements (such as some data processing for real-time violation determination, preliminary analysis of action postures, etc.), the edge computing module performs rapid processing locally and timely feedbacks the results to the user or relevant modules to improve the system's response speed and user experience;

[0260] Data Caching and Synchronization: The edge computing module sets up a data caching mechanism to temporarily store some processed data locally; When the network conditions permit, the cached data is synchronized to the cloud blockchain storage unit to ensure the integrity and consistency of the data;

[0261] III. Key Technical Principles:

[0262] (1) Blockchain Technical Principles:

[0263] Blockchain adopts distributed ledger technology, stores data on multiple nodes, and each node stores a complete copy of the ledger; through hash algorithms and chain structures, it ensures the immutability of data. Once data is recorded on the chain, any modification will cause the hash value to change and thus be detected by other nodes; based on the decentralized characteristics of the blockchain, smart contracts automatically execute pre-set rules, realizing the automation and standardization of data operations, and ensuring the security and credibility of data;

[0264] (2) Principle of federated learning:

[0265] Federated learning is a distributed machine learning framework that allows participating parties to collaboratively train machine learning models without sharing raw data; by training models locally and uploading parameters, the central server aggregates the parameters to achieve the joint optimization of the models; this method not only protects data privacy but also fully utilizes the value of multi-party data, improving the generalization ability and performance of the models;

[0266] (3) Principle of edge computing:

[0267] Edge computing sinks computing and data processing tasks from the cloud to network edge devices (such as sensors, cameras, local servers, etc.); by processing at the source where data is generated, it reduces data transmission latency, reduces the cloud load, and improves the real-time performance and reliability of the system; at the same time, edge computing is combined with cloud computing to form a cloud-edge collaborative computing mode, optimizing the resource utilization efficiency;

[0268] IV. Workflow of the module:

[0269] (1) Initialization phase:

[0270] After the distributed data management unit is started, it loads the blockchain network configuration information, federated learning framework parameters, data permission management rules, and the initialization settings of the edge computing module; establishes data communication connections with other system modules (such as the evaluation module, user management module, etc.), and prepares to receive and process data;

[0271] (2) Data generation and processing phase:

[0272] When the standing long jump evaluation is completed, the evaluation data is first transmitted to the edge computing processing unit; the edge computing module performs preliminary processing on the data such as cleaning and feature extraction, conducts local calculations for tasks with high real-time requirements, and caches the processed data;

[0273] (3) Data storage phase:

[0274] The evaluated data after cleaning is packaged to generate data blocks. The blockchain data storage unit calculates the block hash value through a hash generation algorithm and conducts multi-node synchronous verification. After passing the verification, the data blocks are stored according to the chain structure of the blockchain, and at the same time, the smart contract is executed to record the operations related to data storage.

[0275] (IV) Model Optimization Phase:

[0276] The federated learning optimization unit collects anonymized training data from each site. Each site trains the AI model locally and uploads the parameters. The central server aggregates and updates the model parameters according to the objective function, generates a globally optimized model, and distributes it to each site for a new round of training to continuously optimize the model performance.

[0277] (V) Data Permission Management Phase:

[0278] The data permission management unit generates NFT tokens for users. Users manage their personal data access permissions through the tokens. When a third party applies to access the data, the system verifies the permissions, controls the data access according to the authorization situation, and records the access operations.

[0279] (VI) Data Synchronization and End Phase:

[0280] The edge computing processing unit synchronizes the cached data to the blockchain storage unit at an appropriate time to ensure data integrity. When the system stops running or receives a stop instruction, the distributed data management unit stops the data processing operations, saves the relevant configurations and data, closes the communication connections, and releases the system resources.

[0281] V. Application Value of the Module:

[0282] (I) Ensuring Data Security and Reliability:

[0283] Blockchain technology ensures the immutability and traceability of the evaluated data. The multi-node verification mechanism prevents data errors or malicious modifications, provides a solid guarantee for the authenticity and integrity of the data, and enhances the credibility and availability of the data.

[0284] (II) Improving Model Performance:

[0285] The federated learning framework breaks data silos, aggregates data from multiple sites to optimize the AI model, enabling the model to learn richer features and patterns, improving the accuracy and generalization ability of the model in different scenarios, and enhancing the intelligent level of the system.

[0286] (III) Protecting User Privacy:

[0287] Through anonymization and strict data permission management, users can independently control access to their personal data, protect privacy and security during the data sharing process, meet users' needs for data privacy protection, and enhance users' trust in the system;

[0288] (4) Improve system efficiency:

[0289] Edge computing reduces the amount of data transmission and the computing pressure on the cloud, enables rapid data processing and real-time response, improves the overall operating efficiency of the system and the user experience, reduces the system operation cost, and enables the system to better handle application scenarios with high concurrency and high real-time requirements.

[0290] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A sports exercise AI standing long jump evaluation data management system, characterized in that: Including: Dynamic environment calibration module: It collects site environment data in real time through a multi-spectral sensor array deployed around the long jump area, and establishes a dynamic coordinate system based on an adaptive calibration algorithm of deep learning. The adaptive calibration algorithm generates a virtual calibration grid by fusing environmental light intensity, ground texture features, and obstacle distribution, and does not rely on fixed markers; Multi-modal identity recognition module: It realizes contactless identity authentication by fusing gait recognition and face feature extraction technologies and combining user biometric information in the cloud database; Three-dimensional pose analysis module: It uses multi-view cameras to synchronously collect long jump action videos, constructs a motion trajectory model through a 3D human pose estimation algorithm, and detects the takeoff angle, air pose, and landing stability in real time; Intelligent violation determination engine: Based on a rule engine and an AI time series analysis model, combined with kinematic parameters and spatial trajectory features, it automatically identifies violation behaviors including hopping, crossing the line, and body imbalance; Distributed data management unit: It uses blockchain technology to store evaluation data, supports multi-node synchronous verification and tamper-proof score records, and realizes local data processing through an edge computing module to reduce dependence on the cloud; 2. The AI standing long jump evaluation data management system for physical exercise according to claim 1, wherein: The implementation method of the dynamic environment calibration module includes: Obtaining real-time environmental information of the long jump area through a multi-spectral sensor array, including light intensity, ground texture, and obstacle distribution; Extracting environmental features based on an improved convolutional neural network CNN and dynamically generating virtual calibration grids adapted to different site conditions. The loss function of the improved CNN consists of two parts: Feature reconstruction loss: Calculating the mean square error between the input environmental features and the reconstructed features, and the weight coefficient α is used to control its importance; Environmental robustness loss: By maximizing the feature difference tolerance under light interference, the weight coefficient β is used to balance the anti-interference ability, and the sum of α and β is 1; Optimizing calibration parameters through a reinforcement learning algorithm to ensure the stability of the coordinate system in complex environments such as shadows and reflections.

3. The AI standing long jump evaluation data management system for physical exercise according to claim 1, wherein: The three-dimensional pose analysis module further includes: Using multi-view cameras to synchronously capture stereo images of long jump actions and reconstructing a three-dimensional motion trajectory through an improved stereo matching algorithm. The energy function of the algorithm is composed of the following three items weighted: Data item: Evaluating the consistency of three-dimensional poses through the projection error of multi-view images; Smoothing item: Constraining the continuity of adjacent joint points in space; Temporal item: Ensuring the physical rationality of the pose change between adjacent frames, and the weight coefficients λ1, λ2, and λ3 respectively control the contribution ratios of the three items; Analyzing the joint motion sequence based on a temporal convolutional network TCN and predicting the centroid offset at the moment of landing; Combining inertial measurement unit IMU data and correcting the acceleration error of the air pose through a filtering algorithm to achieve sub-centimeter-level accuracy measurement.

4. The AI standing long jump evaluation data management system for physical exercise according to claim 1, wherein: The determination method of the intelligent violation determination engine includes: Modeling the continuous frame action sequence through an improved LSTM network, and the output of its forget gate is fused with predefined kinematic constraints, and the constraint weight coefficient γ controls the contribution of physical priors; Define a dynamic threshold rule library, adjust the judgment criteria according to the user's age, gender and historical data. The threshold is the weighted average of the baseline threshold and the adaptive offset, and the balance coefficient μ controls the weights of the two; Generate an interpretable violation report, including an action decomposition diagram and improvement suggestions, and provide real-time feedback to the user side through an AR device.

5. The AI standing long jump evaluation data management system for physical exercise according to claim 1, characterized in that: The distributed data management unit further includes: A blockchain-based smart contract mechanism to ensure the integrity and traceability of performance data during transmission and storage. The data hash is generated by concatenating the current data, a random number, and the previous block hash value, and using the SHA-3 algorithm to generate a unique identifier; Adopt a federated learning framework to aggregate anonymized training data from multiple sites, optimize the performance of the global AI model. The objective function is the weighted sum of the local losses of each site, and a regularization term is added to prevent overfitting; Support users to independently manage their personal sports data through the encrypted token NFT and authorize third parties to access it.

Images