Motion capture method based on physical training

Abstract

The invention discloses a motion capture method based on physical training, and relates to the technical field of motion capture, and the method comprises the following steps: setting a capture area and deploying monitoring equipment; marking and identifying human body key points; synchronizing and fusing data; performing action analysis and evaluation; performing feedback and training adjustment; data storage and management; through accurate capture area setting and monitoring equipment deployment, in combination with a human body key point marking and identification technology, data synchronization and fusion are realized; in the action analysis and evaluation link, a deep learning model is used for processing the motion data in real time, and instant feedback and training adjustment suggestions are provided; the whole system is supported by data storage and management, so that the integrity and safety of training data are ensured; the technical effects that the action of the user is not greatly influenced, and the analysis precision of the action of the user is improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of motion capture, and in particular to a motion capture method based on sports training. Background Art

[0002] In the field of physical education, especially in the training of martial arts, dance and other projects, the correctness of posture is crucial. For example, in fixed-position sports such as iron ball throwing and standing long jump, the correctness of the athlete's posture directly affects the improvement of technical level and sports performance. For beginners, due to the lack of systematic training and guidance, irregular movements and postures often lead to slow technical progress and may even cause sports injuries. Therefore, it is particularly important to repeatedly review, analyze and improve one's own movements.

[0003] Traditional sports training methods rely on one-on-one guidance from professional coaches. Although the effect is remarkable, there are certain problems, such as high labor costs and insufficient teaching flexibility. In order to seek simpler and more efficient teaching and training methods, the modern sports training field is actively exploring new technical means.

[0004] In recent years, motion capture technology has been widely used in professional sports training teams and national training teams. This technology can analyze athletes' training movements in detail, accurately locate flaws and deficiencies in movements, and provide targeted improvement solutions for subsequent actual training. However, most existing motion capture technologies are based on posture sensors, requiring athletes to wear a full set of motion capture equipment. This solution not only affects the user's wearing experience, but also affects the user's movements due to equipment limitations, making it difficult to achieve the ideal level of analysis accuracy, and correcting sports movement errors becomes more challenging. Summary of the invention

[0005] The present application provides a motion capture method based on sports training, thereby solving the technical problem in the prior art that the user's movements will be affected due to equipment limitations, resulting in the difficulty in achieving an ideal level of analysis accuracy; and achieves the technical effect of not causing too much impact on the user's movements and improving the accuracy of user motion analysis.

[0006] The present application provides a motion capture method based on sports training, comprising:

[0007] S1: Capture area setting and monitoring equipment deployment;

[0008] S2: Human body key point marking and recognition;

[0009] S3: Data synchronization and fusion;

[0010] S4: Movement analysis and evaluation;

[0011] S5: Feedback and training adjustment;

[0012] S6: Data storage and management.

[0013] Furthermore, the capture area setting and monitoring equipment deployment in step S1 include:

[0014] Select the appropriate venue and determine the size and shape of the capture area to suit the needs of different sports training programs;

[0015] Select appropriate monitoring equipment and determine its quantity and location;

[0016] Ensure that the monitoring equipment covers the capture area without blind spots to ensure accurate capture of key points of the human body;

[0017] Calibrate monitoring equipment to ensure the accuracy and synchronization of data collected;

[0018] Among them, the monitoring equipment includes at least a camera and a depth sensor. The high-definition camera is used to capture two-dimensional images of human body movements; the depth sensor is used to obtain three-dimensional information of the human body.

[0019] Furthermore, the human body key point marking in step S2 includes selecting a key point model, determining marking criteria, determining marking tools and platforms, and ensuring the marking process and marking quality;

[0020] The marking process includes:

[0021] Initialize, load an image or video frame into the tool, and select a keypoint model;

[0022] Frame-by-frame annotation: For video data, it is necessary to annotate each frame, while for static images, it only needs to be annotated once;

[0023] Key point placement: drag the key point to the correct position on the image;

[0024] Check and adjust. After marking is completed, check whether the position of the key points is accurate and make adjustments.

[0025] Furthermore, the human body key point recognition in step S2 includes selecting a recognition algorithm, algorithm training and optimization, a recognition process, optimization of recognition results and guarantee of recognition quality;

[0026] The identification process includes:

[0027] Image input, passing the image captured by the camera as input to the trained model;

[0028] Feature extraction, extracting image features through the model;

[0029] Key point localization,The model predicts the location of key points through a fully connected layer.

[0030] Furthermore, the data synchronization and fusion in step S3 includes time synchronization and data fusion;

[0031] Data fusion includes:

[0032] Calibrate the equipment, use the calibration plate or calibration ball to calibrate the internal and external parameters of the camera and depth sensor, and obtain their calibration matrices;

[0033] Coordinate conversion, converting the coordinate systems of different devices into the same coordinate system;

[0034] Data fusion algorithm, select a suitable data fusion algorithm to merge data from different devices. The data fusion algorithm is Kalman filter or particle filter;

[0035] Real-time processing, integrating fusion algorithms into the data processing flow to ensure real-time processing and updating of data.

[0036] Furthermore, the motion analysis and evaluation in step S4 includes: motion feature extraction, motion pattern recognition, performance index calculation and motion comparison;

[0037] Among them, action feature extraction includes:

[0038] Calculate the position change of each key point from the key point data to obtain the velocity and acceleration;

[0039] Calculate the amplitude and range of motion;

[0040] Analyze the relative positions between key points and calculate joint angles;

[0041] Furthermore, the action mode recognition includes:

[0042] Preprocess the data to convert the key point trajectory data into a form suitable for the machine learning model. The converted data is time series data or feature vector;

[0043] Select a model, select a suitable action recognition model, the action recognition model is any one of a convolutional neural network, a recurrent neural network and a long short-term memory network;

[0044] Train the model using the labeled training data set;

[0045] Validation and testing, use the validation set and test set to evaluate the performance of the model.

[0046] Furthermore, the feedback and training adjustment in step S5 includes real-time feedback, training report generation and training plan adjustment;

[0047] Among them, real-time feedback includes real-time analysis of motion data; once it is detected that the action does not meet the predetermined standards, feedback is immediately provided; feedback is conveyed through visual, auditory or tactile means.

[0048] Furthermore, the data storage and management in step S6 includes data structure design, data storage, data backup, data security and access control, and data retrieval and sharing.

[0049] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0050] Through precise capture area setting and monitoring equipment deployment, combined with human key point marking and recognition technology, data synchronization and fusion are achieved; the motion analysis and evaluation link uses deep learning models to process sports data in real time, providing coaches with instant feedback and training adjustment suggestions; the entire system is supported by data storage and management, ensuring the integrity and security of training data; it solves the technical problem in the existing technology that the user's movements will be affected due to equipment limitations, making it difficult to achieve an ideal level of analysis accuracy; it achieves the technical effect of not causing too much impact on the user's movements and improving the accuracy of user motion analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the motion capture method based on sports training of the present invention;

[0052] Figure 2 The present invention is a schematic diagram of the capture area setting and monitoring equipment deployment process of the motion capture method based on sports training. DETAILED DESCRIPTION

[0053] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0054] It should be noted that the terms “vertical”, “horizontal”, “up”, “down”, “left”, “right” and similar expressions used in this document are only for illustrative purposes and do not represent the only implementation method.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0056] Example: Figure 1 and Figure 2 As shown, the motion capture method based on sports training in this application includes

[0057] S1: Capture area setting and monitoring equipment deployment;

[0058] S2: Human body key point marking and recognition;

[0059] S3: Data synchronization and fusion;

[0060] S4: Movement analysis and evaluation;

[0061] S5: Feedback and training adjustment;

[0062] S6: Data storage and management.

[0063] Specifically, the capture area setting and monitoring equipment deployment in step S1 include:

[0064] 1. Choose a suitable venue and determine the size and shape of the capture area to suit the needs of different sports training programs;

[0065] 2. Select appropriate monitoring equipment and determine its quantity and location;

[0066] 3. Ensure that the monitoring equipment covers the capture area without blind spots to ensure accurate capture of key points of the human body;

[0067] 4. Calibrate the monitoring equipment to ensure the accuracy and synchronization of the collected data.

[0068] Among them, the monitoring equipment includes at least a camera and a depth sensor. The high-definition camera is used to capture two-dimensional images of human body movements; the depth sensor (such as Kinect or Intel RealSense) is used to obtain three-dimensional information of the human body.

[0069] It should be noted that a data processing center and wireless network equipment are provided inside the capture area to ensure data transmission between the monitoring equipment and the data processing center.

[0070] It should be noted that the capture area may be a measured and marked rectangular area. For example, an indoor venue or an outdoor venue may be selected according to the characteristics of the sports training program.

[0071] For example, take indoor basketball training as an example to set the capture area and deploy monitoring equipment:

[0072] Capture area settings:

[0073] The capture area was determined to be half of a basketball court, approximately 28 meters by 15 meters;

[0074] Set markers at key locations such as the three-point line and free throw line to facilitate subsequent data processing;

[0075] Monitoring equipment deployment:

[0076] Eight high-definition cameras are installed around the basketball court to ensure coverage of the entire half court; the cameras complement each other to avoid blind spots; a depth sensor is installed at each end of the basketball court to capture the three-dimensional movement information of the athletes; a wireless network is used to connect the cameras and depth sensors to the data processing center to ensure real-time data transmission.

[0077] Equipment Calibration:

[0078] Use the calibration board to calibrate the camera to ensure that the image data it collects is accurate; calibrate the depth sensor to ensure that the three-dimensional data it collects is accurate.

[0079] It should be noted that in order to ensure the synchronization between monitoring devices, the following is a detailed calibration process:

[0080] A. Time synchronization calibration:

[0081] Determine an accurate time source, such as using a high-precision time server;

[0082] Set time synchronization for each monitoring device to ensure that the time error of all devices is within millisecond level;

[0083] Check the device time synchronization status regularly, for example, automatically perform a synchronization check once a day.

[0084] B. Hardware synchronization calibration:

[0085] Use an external trigger signal source (such as a sync generator) to synchronize the acquisition time of all devices;

[0086] For the camera and depth sensor, ensure that they can start acquiring data at the same time through an external trigger signal.

[0087] C. Software synchronization calibration:

[0088] Use synchronous acquisition software, which can control the start and end time of acquisition for all devices;

[0089] Set the timestamp in the software to record the exact time of data collection for each device to facilitate subsequent data fusion processing.

[0090] Furthermore, the human body key point marking in step S2 includes:

[0091] 1. Select key point model

[0092] Before you start labeling, you first need to select a suitable human key point model. Different models may define different numbers of key points, for example:

[0093] The COCO model defines 17 key points (ears, eyes, nose, shoulders, elbows, wrists, hips, knees, ankles);

[0094] The MPII model defines 16 key points and includes the center point of the head;

[0095] The AICH model may define more or fewer key points, depending on specific needs.

[0096] 2. Determine the marking criteria

[0097] Consistency: ensuring that all annotators understand and follow the same tagging guidelines;

[0098] Accuracy: Key points should be marked as accurately as possible on visible body parts, for example, the wrist key point should be marked at the crease of the wrist;

[0099] Visibility: If a key point is not visible due to occlusion, it should be marked as specified by the model, such as at the edge of the occluder.

[0100] 3. Determine the labeling tool and platform

[0101] Use professional annotation tools, such as LabelImg, CVAT, or MakeSense, which provide intuitive interfaces and convenient operations; ensure that the tools support functions such as batch processing, multi-user collaboration, and version control.

[0102] 4. Marking process

[0103] Initialization: Load an image or video frame in the tool and select a keypoint model;

[0104] Frame-by-frame annotation: For video data, it is necessary to annotate each frame; for static images, it only needs to be annotated once;

[0105] Keypoint placement: Drag the keypoints to the correct position on the image; some tools provide auxiliary functions such as zooming, grid alignment, etc. to help place keypoints more accurately;

[0106] Check and adjust: After marking, check whether the position of key points is accurate and make adjustments if necessary.

[0107] For example, let's mark the key points of a basketball player's shoulders:

[0108] Positioning: Locate the athlete's shoulder, usually at the acromion.

[0109] Place the key point: Place the key point in the middle of the acromion, making sure it is on the contour of the shoulder;

[0110] Verification: Verify that key points are always marked in the correct position under different viewing angles and actions;

[0111] 5. Guarantee of marking quality

[0112] Cross-validation: Have different annotators annotate the same data and then compare the results to ensure consistency;

[0113] Expert review: Let domain experts review the annotation results to ensure the accuracy of key points;

[0114] Iterative improvement: Continuously collect feedback during the annotation process, update the marking guidelines, and improve the annotation quality.

[0115] It should be noted that in order to improve the accuracy and consistency of the key points of the human body, the following are supplements to the annotation process and quality control measures:

[0116] A. Standardization of labeling criteria:

[0117] Develop a detailed annotation manual to clarify the definition, marking location and marking method of each key point;

[0118] Provide unified training to annotators to ensure that each annotator has the same understanding of the annotation guidelines.

[0119] B. Labeling quality control:

[0120] Implement multiple rounds of annotation and cross-validation, i.e. each data sample is independently annotated by multiple annotators, and then the results are compared to resolve inconsistencies;

[0121] Conduct annotation quality assessments regularly, have experts review annotation results, and promptly identify and correct deviations.

[0122] C. Annotation tool optimization:

[0123] Choose annotation tools that support real-time feedback and version control, so that annotators can adjust and correct annotation errors in a timely manner;

[0124] Auxiliary functions are integrated into the tool, such as smart prompts and automatic alignment, to reduce human errors.

[0125] D annotation process iteration:

[0126] Continuously update annotation guidelines and processes based on feedback collected during the annotation process;

[0127] By building a annotation case library, we accumulate annotation experience and solutions to common problems for reference by annotators.

[0128] Furthermore, the human body key point recognition in step S2 includes:

[0129] 1. Select recognition algorithm

[0130] Keypoint recognition usually relies on deep learning models, especially convolutional neural networks (CNNs). The following are optional algorithms:

[0131] OpenPose: A multi-stage CNN-based approach that can simultaneously detect human key points and body connections.

[0132] PoseNet: Use CNN to estimate the position of human key points directly from a single image;

[0133] AlphaPose: Similar to OpenPose, but provides faster detection speed and higher accuracy.

[0134] 2. Algorithm training and optimization

[0135] Dataset preparation: Use a labeled keypoint dataset to train the model, such as COCO, MPII, or a custom dataset;

[0136] Model architecture: Choose a suitable network architecture, such as Hourglass, CPM (Convolutional Posture Machine), or ResNet;

[0137] Loss function: Define a loss function to measure the difference between the predicted keypoints and the true keypoints, such as mean square error (MSE) or Wing Loss;

[0138] Hyperparameter tuning: Adjust hyperparameters such as learning rate, batch size, and number of iterations to optimize model performance;

[0139] 3. Identification process

[0140] Image input: The image captured by the camera is passed as input to the trained model;

[0141] Feature extraction: The model extracts image features through convolutional layers;

[0142] Keypoint localization: The model predicts the locations of keypoints through fully connected layers or other localization mechanisms.

[0143] For example, let's use OpenPose to identify key points of basketball players:

[0144] Image preprocessing: resizing the image to match the input size requirements of the OpenPose model;

[0145] Model inference: input the preprocessed image into the OpenPose model;

[0146] Keypoint Output: The model outputs the keypoint coordinates and confidence for each player in the image.

[0147] 4. Optimization of recognition results

[0148] Confidence filtering: filter out inaccurate key points predicted by the model based on the confidence threshold;

[0149] Non-maximum suppression (NMS): used to process multiple keypoint predictions to ensure that each keypoint is detected only once;

[0150] Key point smoothing: Smooth key point trajectories through time series analysis or sliding average and other methods to reduce jitter.

[0151] 5. Guarantee of identification quality

[0152] Real-time monitoring: monitor the quality of key point detection in real time during the recognition process to ensure real-time feedback;

[0153] Performance evaluation: Use metrics such as PCK (keypoint accuracy) or OKS (object keypoint similarity) to evaluate model performance;

[0154] Error analysis: Conduct in-depth analysis of cases with recognition errors to find out the problems and improve the model accordingly.

[0155] Furthermore, the data synchronization and fusion in step S3 includes:

[0156] 1. Time synchronization

[0157] Implementation process:

[0158] Identify a Network Time Protocol (NTP) server as the time source. This server can be a public NTP server on the Internet or a private NTP server within the local network.

[0159] Configure the NTP client on each surveillance device (camera and depth sensor) to synchronize the time with the NTP server regularly;

[0160] Set the synchronization frequency, for example, once every minute, to ensure that the time difference between devices is minimal;

[0161] Verify the time synchronization effect and check whether the time of each device is consistent with the NTP server through the log or management interface.

[0162] For example, suppose we use a public NTP server (such as time.nist.gov) and configure an NTP client on each HD camera and depth sensor; set the time to synchronize once a minute through the command line or device management interface; check the system time of each device to confirm that they are all synchronized with the time of the NTP server with an error of milliseconds.

[0163] 2. Data Fusion

[0164] Implementation process:

[0165] Calibration equipment: Use calibration plates or calibration spheres to calibrate the internal and external parameters of the camera and depth sensor to obtain their calibration matrices;

[0166] Coordinate conversion: convert the coordinate systems of different devices into the same coordinate system. The conversion process involves rotation and translation transformation;

[0167] Data fusion algorithms: Develop or select appropriate data fusion algorithms, such as Kalman filters or particle filters, to merge data from different devices;

[0168] Real-time processing: Integrate fusion algorithms into the data processing flow to ensure real-time processing and updating of data.

[0169] For example:

[0170] Calibration equipment: Place a calibration plate of known size on the basketball court, calibrate the eight high-definition cameras and two depth sensors, and obtain their intrinsic and extrinsic parameter matrices;

[0171] Coordinate transformation: Assuming that we have captured the key points of an athlete's shoulder, we first need to transform the 3D coordinates captured by the depth sensor into the camera coordinate system. This can be done by following the steps below:

[0172] a. Use the depth sensor and calibration board to determine the position and orientation of the sensor (extrinsic parameter matrix);

[0173] b. Use a camera to capture the image of the calibration plate, and determine the camera's intrinsic and extrinsic parameter matrices based on the relationship between the image feature points and the actual physical points;

[0174] c. Apply the coordinate conversion formula to convert the depth sensor coordinates to camera coordinates.

[0175] d. Use Kalman filter to fuse data from different cameras and depth sensors. Kalman filter can predict and update the position and velocity of key points of the athlete's shoulder;

[0176] e. In the data processing center, the Kalman filter is run in real time to fuse the data captured by the camera and depth sensor into a coherent three-dimensional motion trajectory. For example, when an athlete is shooting, the system can update and display the three-dimensional position of the key points of his shoulder in real time, helping coaches and athletes analyze the accuracy and fluency of the action.

[0177] Furthermore, the action analysis and evaluation in step S4 includes:

[0178] 1. Action feature extraction

[0179] Implementation process:

[0180] Calculate the position change of each key point from the key point data to obtain the velocity and acceleration;

[0181] Calculate the amplitude and range of movement, such as the height of a jump or the width of an arm extension;

[0182] Analyze the relative positions between key points and calculate joint angles.

[0183] For example, in basketball training, when an athlete is shooting, the system will calculate the velocity and acceleration of the wrist key points to evaluate the strength and speed of the shot; the system will also measure the height change of the athlete's ankle key points when taking off to evaluate jumping ability; by calculating the angles between the shoulder, elbow and wrist key points, the arm posture during shooting is analyzed.

[0184] 2. Action pattern recognition

[0185] Implementation process:

[0186] Preprocessing data: Convert keypoint trajectory data into a form suitable for machine learning models, such as time series data or feature vectors;

[0187] Model selection: Choose a suitable action recognition model, such as convolutional neural network (CNN), recurrent neural network (RNN), or long short-term memory network (LSTM);

[0188] Training model: Use the labeled training data set to train the model;

[0189] Validation and Testing: Use validation and test sets to evaluate the performance of the model.

[0190] For example, convert the key point trajectory of the shooting action into time series data, where each time point corresponds to the position of a key point; select an LSTM-based network to identify different shooting patterns; use a dataset containing different shooting actions (such as jump shots and layups) to train the LSTM model; and evaluate the model on an independent test set to ensure that it can accurately identify different shooting actions.

[0191] 3. Performance index calculation

[0192] Implementation process:

[0193] Define performance indicators: Define relevant performance indicators according to the needs of the sport, such as shooting percentage, running speed, etc.

[0194] Collect data: Collect data related to performance indicators during training;

[0195] Calculate metrics: Calculate performance metrics using statistical methods or algorithms.

[0196] For example, in basketball training, the shooting percentage is defined as a key performance indicator; the hit rate of each shot of the athlete in training is recorded; the shooting percentage of the athlete in a certain period of time is calculated, such as 6 out of 10 shots are successful, and the hit rate is 60%.

[0197] 4. Action comparison

[0198] Implementation process:

[0199] Determine a standard action model: Select or create a standard action model as a benchmark for comparison;

[0200] Comparative analysis: comparing the athlete's actual movements with the standard movement model to identify differences;

[0201] Visualize the differences: Use charts or animations to visually show the differences in actions.

[0202] For example, the shooting action of a professional basketball player is selected as the standard model; the key point trajectories of the athlete's shooting action are compared with the standard model to find out the inconsistencies in the action; through animation playback, the differences between the athlete and the standard action are displayed, such as excessive arm angle or insufficient shooting power.

[0203] Furthermore, the feedback and training adjustment in step S5 includes:

[0204] 1. Real-time feedback

[0205] Implementation process:

[0206] As the athlete performs an action, the system analyzes the action data in real time; once it detects that the action does not meet the predetermined standards, the system immediately provides feedback; feedback can be conveyed through visual, auditory or tactile means, such as on-screen instructions, sound prompts or vibrations of wearable devices.

[0207] For example, in basketball training, when an athlete shoots, if the system detects that the wrist angle does not meet the standard shooting action, the screen will display a prompt to "adjust wrist angle" in real time; at the same time, the wearable device can vibrate to remind the athlete to adjust the action.

[0208] 2. Training report generation

[0209] Implementation process:

[0210] After training, the system summarizes the analysis results; generates a training report containing key indicators, movement evaluation and improvement suggestions; the report can be provided to coaches and athletes via email, mobile application or web interface.

[0211] For example, after training, the system generates a report that includes the athlete's shooting accuracy, average jump height, and movement consistency score; the report points out that the athlete's wrist angle is too large when shooting, and recommends special training to improve wrist control.

[0212] 3. Adjustment of training plan

[0213] Implementation process:

[0214] Based on the analysis results in the training report, the coach adjusts the training plan: increasing or decreasing the proportion of specific training items, introducing new training methods, or setting new training goals and milestones.

[0215] For example, the coach found that the athlete's shooting accuracy was lower than the standard, so he decided to increase the special training time for shooting technique; wrist strength and flexibility training, as well as repeated practice of shooting movements were added to the training plan.

[0216] 4. Personalized training

[0217] Implementation process:

[0218] Customize personalized training plans based on the athlete's individual performance and characteristics; use data analysis to identify the athlete's weaknesses and potential points; provide athletes with targeted training content and progress arrangements.

[0219] For example, through analysis, it was found that a certain athlete often used excessive force when shooting, resulting in reduced accuracy; the coach customized a special training program for the athlete, including lightweight shooting exercises, to improve his strength control when shooting; the athlete practiced according to the personalized training program and gradually improved his shooting accuracy in subsequent training.

[0220] Furthermore, the data storage and management in step S6 includes:

[0221] 1. Data structure design

[0222] Implementation process:

[0223] Define data models, including athlete information, training data, motion capture data, and evaluation results; design database table structures to ensure data integrity and consistency.

[0224] Specifically, create an athletes table that contains basic information such as athlete ID, name, age, height, weight, etc.

[0225] Create a training_sessions table to record the date, time, athlete ID and training items of each training session;

[0226] Create a motion_data table to store key point coordinates, timestamps, training session IDs, and other information.

[0227] 2. Data Storage

[0228] Implementation process:

[0229] After motion capture and analysis are completed, the data is stored in the database; transactions are used to ensure the atomicity, consistency, isolation, and durability (ACID properties) of data writes.

[0230] For example, after a basketball training session, the system writes the athlete's shooting action key point data, speed, acceleration and other information into the motion_data table; the system ensures that if an error occurs during the writing process, the entire transaction will be rolled back to avoid data corruption.

[0231] 3. Data backup

[0232] Implementation process:

[0233] Back up the database regularly and automatically to prevent data loss; choose an appropriate backup strategy, such as full backup, incremental backup, or differential backup.

[0234] Specifically, you can set the system to automatically perform a full database backup every night and store the backup file in the cloud or on an external storage device; perform an incremental backup once a week, and only back up the data that has changed since the last full backup.

[0235] 4. Data security and access control

[0236] Implementation process:

[0237] Implement user authentication mechanisms to ensure that only authorized users can access data; set different permission levels to limit user access and operations on sensitive data.

[0238] Specifically, coaches, athletes and other staff log in to the system with usernames and passwords; coaches can access all training data and analysis reports, while athletes can only view their own training progress and feedback.

[0239] 5. Data retrieval and sharing

[0240] Implementation process:

[0241] Provides a data retrieval interface that allows users to query data based on specific conditions; implements data export function to facilitate data analysis and sharing.

[0242] Specifically, coaches can use the system to query the training performance of specific athletes in the past month and filter out data for specific training items; the system allows coaches to export training reports in PDF or Excel format for sharing in team meetings.

[0243] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0244] 1. It solves the technical problem in the prior art that the user's actions will be affected by the equipment limitations, resulting in the difficulty in achieving the ideal level of analysis accuracy; it achieves the technical effect of not causing too much impact on the user's actions and improving the accuracy of user action analysis;

[0245] 2. Avoid the discomfort caused by external equipment to the user and improve the user's sensory perception and movement fluency during training;

[0246] 3. Optimization of capture area setting and monitoring equipment deployment: By selecting appropriate sites and monitoring equipment and ensuring that there is no blind spot coverage between devices, the accuracy of key point capture and the integrity of training data are improved;

[0247] 4. Accuracy of key point marking and recognition: Select appropriate key point models, determine consistent marking criteria, and use professional annotation tools to improve the accuracy and consistency of key point marking;

[0248] 5. Real-time and accuracy of data synchronization and fusion: The network time protocol (NTP) is used for time synchronization, and data fusion is performed through algorithms such as the Kalman filter, ensuring the real-time and accuracy of data;

[0249] 6. Deep learning applications for motion analysis and evaluation: Deep learning models are used to extract motion features, recognize patterns, and calculate performance indicators, which improves the accuracy of motion analysis and the evaluation of training effects;

[0250] 7. Real-time and personalized feedback and training adjustments: Provide real-time feedback, generate personalized training reports, and adjust training plans based on analysis results, improving training results and athlete performance;

[0251] 8. Integrity and security of data storage and management: Design a reasonable data structure, implement data backup and access control to ensure data integrity and security;

[0252] 9. Convenience of data retrieval and sharing: Provides data retrieval interface and data export functions to facilitate users to query and analyze data, and supports data sharing and dissemination.

[0253] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A motion capture method based on sports training, characterized in that: include: S1: Capture area setting and monitoring equipment deployment; S2: Human body key point marking and recognition; S3: Data synchronization and fusion; S4: Movement analysis and evaluation; S5: Feedback and training adjustment; S6: Data storage and management.

2. The motion capture method based on sports training as claimed in claim 1, characterized in that: The capture area setting and monitoring equipment deployment in step S1 include: Select the appropriate venue and determine the size and shape of the capture area to suit the needs of different sports training programs; Select appropriate monitoring equipment and determine its quantity and location; Ensure that the monitoring equipment covers the capture area without blind spots to ensure accurate capture of key points of the human body; Calibrate monitoring equipment to ensure the accuracy and synchronization of data collected; Among them, the monitoring equipment includes at least a camera and a depth sensor. The high-definition camera is used to capture two-dimensional images of human body movements; the depth sensor is used to obtain three-dimensional information of the human body.

3. The motion capture method based on sports training as claimed in claim 1, characterized in that: The human body key point marking in step S2 includes selecting a key point model, determining marking criteria, determining marking tools and platforms, and ensuring the marking process and marking quality; The marking process includes: Initialize, load an image or video frame into the tool, and select a keypoint model; Frame-by-frame annotation: For video data, it is necessary to annotate each frame, while for static images, it only needs to be annotated once; Key point placement: drag the key point to the correct position on the image; Check and adjust. After marking is completed, check whether the position of the key points is accurate and make adjustments.

4. The motion capture method based on sports training as claimed in claim 3, characterized in that: The human body key point recognition in step S2 includes selecting a recognition algorithm, algorithm training and optimization, a recognition process, optimization of recognition results and guarantee of recognition quality; The identification process includes: Image input, passing the image captured by the camera as input to the trained model; Feature extraction, extracting image features through the model; Key point localization,The model predicts the location of key points through a fully connected layer.

5. The motion capture method based on sports training as claimed in claim 1, characterized in that: The data synchronization and fusion in step S3 include time synchronization and data fusion; Data fusion includes: Calibrate the equipment, use the calibration plate or calibration ball to calibrate the internal and external parameters of the camera and depth sensor, and obtain their calibration matrices; Coordinate conversion, converting the coordinate systems of different devices into the same coordinate system; Data fusion algorithm, select a suitable data fusion algorithm to merge data from different devices. The data fusion algorithm is Kalman filter or particle filter; Real-time processing, integrating fusion algorithms into the data processing flow to ensure real-time processing and updating of data.

6. The motion capture method based on sports training as claimed in claim 4, characterized in that: The motion analysis and evaluation in step S4 includes: motion feature extraction, motion pattern recognition, performance index calculation and motion comparison; Among them, action feature extraction includes: Calculate the position change of each key point from the key point data to obtain the velocity and acceleration; Calculate the amplitude and range of motion; Analyze the relative positions between key points and calculate joint angles.

7. The motion capture method based on sports training as claimed in claim 6, characterized in that: The action mode recognition includes: Preprocess the data to convert the key point trajectory data into a form suitable for the machine learning model. The converted data is time series data or feature vector; Select a model, select a suitable action recognition model, the action recognition model is any one of a convolutional neural network, a recurrent neural network and a long short-term memory network; Train the model using the labeled training data set; Validation and testing, use the validation set and test set to evaluate the performance of the model.

8. The motion capture method based on sports training as claimed in claim 1, characterized in that: The feedback and training adjustment in step S5 include real-time feedback, training report generation and training plan adjustment; Among them, real-time feedback includes real-time analysis of motion data; once it is detected that the action does not meet the predetermined standards, feedback is immediately provided; feedback is conveyed through visual, auditory or tactile means.

9. The motion capture method based on sports training as claimed in claim 1, characterized in that: The data storage and management in step S6 includes data structure design, data storage, data backup, data security and access control, and data retrieval and sharing.

Images