Sprint test intelligent timing system based on deep learning and human body tracking
Through the intelligent timing system of deep learning and human body tracking technology, the timing error and environmental interference of traditional sprint timing systems are solved, and efficient, accurate and fair sprint exam timing is achieved, reducing labor costs and improving the flexibility and fairness of the system.
Patent Information
- Application Number
- CN202410206423.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-26
- Publication Date
- 2025-08-26
AI Technical Summary
Traditional sprint timing systems have large manual timing errors, complex photoelectric timing equipment is susceptible to environmental interference, high cost and difficult to upgrade, resulting in inaccurate timing and high error rates, making it difficult to meet the efficiency and fairness needs of large-scale examinations.
Using an intelligent timing system based on deep learning and human body tracking, we recognize and track athletes through deep learning models, and combine face recognition and human body posture tracking technology to realize automatic timing, runway allocation and performance display, reducing labor costs and improving timing accuracy and fairness.
It realizes accurate timing in complex environments, eliminates substitutes, improves examination efficiency, reduces costs, facilitates system upgrades and maintenance, and ensures fairness and justice of the examination.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] The present invention relates to the field of sprint test timing, and in particular to an intelligent sprint test timing system based on deep learning and human body tracking. Background Art
[0002] Traditional sprint timing systems primarily use manual or photoelectric timing. Manual timing is subjective, typically requiring a human timekeeper to use timing equipment to accurately time the start and finish times of runners. However, this manual timing method is subject to certain subjectivity and errors, and requires significant human resources for large-scale sprint tests. Photoelectric timing is susceptible to interference from environmental factors. Traditional photoelectric timing equipment is complex and subject to numerous environmental factors, making it suitable only for simple competition scenarios and complex human environments. This results in inaccurate and unreliable timing results, and is also insufficient for efficient timing in tests. Furthermore, image timing is difficult to accurately determine due to complex outdoor lighting and frequent shadows, which requires deep learning to avoid unnecessary interference. Traditional timing systems require specialized equipment and personnel for maintenance, and due to hardware limitations, functional expansion and upgrades are difficult and expensive. In sports sprint tests, manual scoring can easily lead to high error rates, raising concerns among parents and students. Summary of the Invention To address the above issues, the present invention proposes an intelligent sprint test timing system based on deep learning and body tracking. Summary of the Invention
[0003] Specific implementation plan: The system includes components such as the sports sprint test management cloud platform, verification and runway allocation terminal, starting terminal, finish timing terminal, display terminal and management cloud platform. The terminals are connected to each other, the verification and runway allocation terminal is connected to the acquisition equipment, the starting terminal is connected to the external audio output device, and the finish timing terminal is connected to the camera to form an intelligent network.
[0004] In terms of technical implementation, the system first uses a deep learning model to identify and track athletes from camera images. This model has undergone extensive training and is capable of accurately identifying and tracking athletes. The human tracking module, receiving the output from the deep learning model, uses a highly efficient algorithm to precisely track athletes. This module is capable of handling complex scenarios, such as occlusion and motion changes between athletes. Through the Sports Sprint Exam Management Cloud Platform, facial recognition technology is used to assign lanes. Deep learning and human posture tracking technology are used to achieve precise time synchronization using SNTP. Image recognition and facial verification are used for lane assignment to prevent cheating and ensure fair and impartial exams. Deep learning and real-time human posture tracking technology are used for automatic timing, eliminating external factors that may interfere with human image recognition, as well as manual timing errors and subjective factors, ensuring accurate and impartial timing. The system also displays results on a large screen in real time to prevent students from questioning their scores. Furthermore, the system records departure and arrival times and saves relevant images and videos for candidates to appeal and arbitrate, successfully reducing labor costs and achieving the goal of replacing time judges with equipment.
[0005] Advantages of sprint timing technology include: Accurate timing: Using deep learning and human posture tracking technology, it can accurately identify and track athletes, and accurately record the time when athletes reach the finish line even in complex environments, with minimal timing error; Eliminate cheating: Through facial recognition and image recognition technology, we verify the identity of candidates to ensure that the candidate information is consistent with the registration information, eliminate cheating and maintain the fairness of the examination; Improve efficiency: Connecting various terminals through an intelligent network enables automatic timing and score recording, reducing the workload of referees, improving examination efficiency, and shortening examination time; Convenient technology upgrades: Cloud platform management facilitates technology upgrades and function expansion, such as adding new timing algorithms and supporting new examination items, which improves the flexibility of the system. Reduce costs: By replacing time judges with equipment, the cost of manpower input is reduced, the economy of the system is improved, and the cost of examinations is reduced; Simple maintenance: Easy maintenance, simple site layout, easy installation and disassembly, maintenance personnel only need to regularly check whether the equipment is normal. BRIEF DESCRIPTION OF THE DRAWINGS
[0006] Figure 1 This is a schematic diagram of the overall structure of an intelligent timing system for sprint tests based on deep learning and human posture tracking according to an embodiment of the present invention; Figure 2 This is an intelligent timing system structure applicable to sprint tests according to an embodiment of the present invention; Figure 3 is a functional block diagram of a cloud platform according to an embodiment of the present invention; Figure 4 is a functional block diagram of a terminal for verifying runway allocation according to an embodiment of the present invention; Figure 5 is a functional block diagram of a command issuing terminal according to an embodiment of the present invention; Figure 6 is a functional block diagram of an endpoint timing terminal according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the layout of the entire sprint test examination room according to one embodiment of the present invention; Figure 8 This is a schematic diagram of the layout of the entire sprint test center terminal in one embodiment of the present invention; Figure 9 This is a schematic diagram of the camera arrangement at the end of the entire sprint test site according to one embodiment of the present invention; Figure 10 This is a diagram of the entire sprint test terminal detection interface of an embodiment of the present invention; Figure 11 This is a diagram of the entire sprint test starting terminal interface according to one embodiment of the present invention; Figure 12 This is a schematic diagram of the overall examination process according to an embodiment of the present invention; Figure 13 It is a schematic diagram of the process of processing the suspension of candidates according to one embodiment of the present invention. DETAILED DESCRIPTION
[0007] The present invention will be further described below with reference to the accompanying drawings and examples.
[0008] 1. System function definition.
[0009] 1.1 Summary of user functions: The functional modules and specific functions of the intelligent timing system for sprint tests based on deep learning and human posture tracking are listed in detail: The sprint test timing system's cloud-based exam management platform includes: a user layer (PC browser); a business layer (test area and site management, candidate management, verification management, score management, objection management, report statistics, score display, and system settings); a data layer (terminal data collection, data exchange, data verification, data storage, and monitoring storage); a terminal management layer (terminal management and platform communication); and a connection layer (4G / 5G / Wi-Fi). The verification and lane assignment terminal includes: connection to data acquisition equipment; lane assignment through facial recognition. The starting terminal includes: external audio output, starting commands, video and image storage, and start image upload. The finish timing terminal includes: one terminal per lane connected to a camera; records the finish line arrival time; provides traceable timing data; and two finish timing terminals can replicate timing for both lanes. The manual backup timing terminal includes: manual timing for system backup in the event of a failure of either finish timing terminal. The display terminal includes: real-time large-screen display of results.
[0010] Figure 1 This is a schematic structural diagram of an intelligent timing system for sports sprint tests based on deep learning and human posture tracking according to an embodiment of the present invention.
[0011] Figure 2 The present invention discloses an examination system structure suitable for intelligent timing of sprint tests according to an embodiment of the present invention.
[0012] Figure 3 This is a functional block diagram of a cloud platform according to an embodiment of the present invention.
[0013] 1.2 The entire implementation process flow.
[0014] Figure 7 It is a schematic diagram of the layout of the entire runway test examination room according to one embodiment of the present invention.
[0015] (1) Verify the identity of the candidate: Upon arrival at the venue, candidates taking the sprint test will first need to go to the identity verification station at the entrance. There, the advanced facial recognition system will quickly match the candidate's facial features with the registration information to confirm their eligibility.
[0016] (2) Lane Assignment: Once the identity is confirmed, the next step is lane assignment. The electronic management system of the stadium will automatically assign a lane to each candidate based on the real-time occupancy of the lane, ensuring that everyone has a suitable position to perform their best.
[0017] (3) Preparation before the test: Participants assigned to a specific track will go to a designated location to warm up and prepare. This period of time is crucial for participants as they need to adjust their mindset and physical condition before the official start.
[0018] (4) Test starts: After all references are ready, the starter will give a clear start signal through the loudspeaker. Once the signal sounds, the monitoring system will start to operate and record the test footage of the contestants.
[0019] (5) Timing of the sprint process: Once the test begins, the dedicated timing system of each lane will be activated to accurately capture the start of each participant and every movement during the test, and the cameras on the sidelines will monitor the process in real time.
[0020] (6) Finish line sprint and timing: Once the candidate sprints across the finish line, the timing system will immediately capture the time of this moment, and the camera near the finish line will capture the scene of the candidate crossing the finish line. This is crucial to ensure the accuracy of the test results.
[0021] (7) Results display: After each participant’s arrival time is linked to their track and number, it will be electronically recorded and sent to the sports sprint test management cloud platform. After processing, this information will be convenient for the public to view the results in real time on the big screen.
[0022] (8) Score filing and archiving: All recorded scores and exam-related information will be uploaded and stored on the cloud-based exam management platform. This is not only used for immediate score review, but also provides a reliable basis for subsequent score inquiries, certification applications, and possible appeals.
[0023] Figure 8 This is a schematic diagram of the layout of the entire sprint test center terminal in one embodiment of the present invention; Figure 9 This is a schematic diagram of the camera arrangement at the end of the entire sprint test site according to one embodiment of the present invention; Figure 10 This is a diagram of the entire sprint test terminal detection interface of an embodiment of the present invention; Figure 11 This is a diagram of the entire sprint test starting terminal interface according to one embodiment of the present invention; Figure 12 This is a schematic diagram of the overall examination process according to an embodiment of the present invention; Figure 13 It is a schematic diagram of the process of processing the suspension of candidates according to one embodiment of the present invention. 2. Specific implementation methods
[0025] 2.1 Setting up hardware and network Arrange all hardware equipment in the test center, including the verification and lane assignment terminal, the starting terminal, the finish timing terminal, and the display terminal. These devices must be connected to the Sports Sprint Test Management Cloud Platform via a network connection, using 4G / 5G / WiFi. The verification and lane assignment terminal must be connected to a data acquisition device, such as a camera; the starting terminal must be connected to an external audio output device; and each lane must have a finish timing terminal connected to a camera.
[0026] Figure 2 This is a system structure applicable to a sprint test according to an embodiment of the present invention; Figure 3 This is a functional block diagram of a cloud platform according to an embodiment of the present invention.
[0027] 2.2 Candidate information import Before the exam, administrators need to import all candidates' basic information, including name, gender, age, and school, into the Sports Sprint Exam Management Cloud Platform and upload a facial recognition image file for each candidate. Furthermore, the exam rules and scoring criteria must be set in advance.
[0028] 2.3 Verification and runway allocation Before the exam begins, each candidate must verify their identity at the verification and lane assignment terminal. The system uses an uploaded image of the candidate's face for facial recognition to confirm their identity. The system then automatically assigns them to an available lane based on pre-set rules. This process is displayed on the large-screen display terminal and uploaded to the cloud platform. Figure 4 This is a functional block diagram of a runway allocation verification terminal according to an embodiment of the present invention.
[0029] 2.4 Starting and Timing When all candidates are ready, the starting terminal will send out a start signal, which is played directly through the speaker. At the same time, the starting terminal will start recording the time and record the picture and video of the moment of starting. Figure 5 This is a functional block diagram of a command issuing terminal according to an embodiment of the present invention.
[0030] 2.5 Finish Time After each candidate reaches the finish line, the finish timing terminal automatically detects the moment of arrival using deep learning and human posture tracking technology and ends the timing. Photos and videos of the moment of arrival at the finish line are recorded. Figure 6 This is a functional block diagram of an endpoint timing terminal according to an embodiment of the present invention.
[0031] 2.6 Display and upload results All candidates' running times will be displayed in real time on the display terminal and uploaded to the sports sprint test management cloud platform for storage and management. The management platform can also handle score statistics, objection handling, and data backup.
[0032] 2.7 Objection Handling If a candidate disagrees with their score, they can submit an objection application on the sports sprint test management cloud platform. Administrators can arbitrate based on the pictures and videos of the candidates' departure and arrival at the finish line stored on the cloud platform. Figure 13 It is a schematic diagram of the process of processing the suspension of candidates according to one embodiment of the present invention.
[0033] This intelligent timing system of the sports sprint test management cloud platform not only improves the efficiency of the test and reduces human errors, but also ensures the fairness of the test and can handle objections fairly.
[0034] 3. The finish timing terminal is realized through deep learning and human posture tracking.
[0035] 3.1 Data Collection: In the actual sprint test scenario, cameras are used to capture and obtain images and video data of the test takers running. This data can be used to train deep learning models and perform human posture tracking.
[0036] 3.2 Data Preprocessing: Video data is divided into keyframes, and each keyframe generates an image. Data augmentation, such as rotation, scaling, and cropping, is performed to increase the amount and diversity of data, thereby improving the generalization ability of the model.
[0037] 3.3. Model Training: Model training uses the deep learning object detection algorithm YOLO. This process requires a large amount of labeled data, including images of sprinters reaching the finish line. The goal of model training is to identify the moment when the sprinter reaches the finish line, thereby achieving accurate timing.
[0038] 3.4 Prediction and real-time timing: In the actual test, the finish timing terminal will receive real-time video data from the camera. Through the trained deep learning model and human posture tracking technology, it will predict and determine in real time whether the sprinter has reached the finish line. Once the finish line is reached, the current time will be recorded immediately to complete the timing.
[0039] 3.5 Data post-processing: The prediction results, including pictures and time of reaching the finish line, will be uploaded to the cloud platform for subsequent performance management and objection handling.
[0040] 4. Deep learning model steps for running target detection of running candidates.
[0041] The main idea is to treat the target detection problem of running candidates as a regression problem, process the entire image at one time, and perform bounding box regression directly on the image, thereby improving the processing speed.
[0042] 4.1 Data Preparation: Collect high-quality running image data, ensuring good lighting and background. Use appropriate tools to annotate the images and ensure accurate bounding boxes. Perform data augmentation, including image flipping, cropping, scaling, and color jittering, to improve the model's generalization capabilities.
[0043] 4.2 Define the model: Choose an appropriate CNN architecture and add a fully connected layer on top of the CNN to predict bounding boxes and class probabilities. You can use a pre-trained CNN model to speed up training and improve accuracy.
[0044] 4.3 Loss Function: Use an appropriate loss function, such as cross entropy loss or IoU loss, to measure the difference between the model's predictions and the true labels. A weighted loss function can be used to increase the model's attention to important samples.
[0045] 4.4 Training: Use an appropriate optimization algorithm, such as gradient descent or Adam optimization algorithm, to minimize the loss function. Use an appropriate learning rate and training batch size to prevent the model from underfitting or overfitting. Use data augmentation techniques to prevent the model from overfitting.
[0046] 4.5 Validation and Testing: Evaluate the model's performance on the validation set and adjust hyperparameters to optimize performance. Evaluate the model's final performance on the test set and fine-tune the model as needed.
[0047] Regarding the limitation of the protection scope of the present invention, those skilled in the art should understand that, based on the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the protection scope of the present invention.
Claims
1. An intelligent timing system for a sprint test based on deep learning and human body tracking, comprising a sports sprint test management cloud platform, a verification and lane allocation terminal, a starting terminal, an end timing terminal, and a display terminal. The management cloud platform is connected to each terminal, the verification and lane allocation terminal is connected to a collection device, the starting terminal is externally connected to an audio output device, and the end timing terminal of each lane is connected to a camera. The present invention allocates lanes through a cloud platform combined with face recognition, accurately calibrates time through SNTP based on deep learning and human posture tracking technology to ensure accurate timing; lane allocation is achieved through image recognition and face verification to prevent cheating and ensure fairness; automatic timing is achieved through deep learning and real-time human posture tracking to avoid interference with human image recognition due to external factors, reduce errors in manual timing and the influence of subjective factors, and be more fair, just and accurate. Results are displayed on a large screen in real time for students to view, while departure and arrival time records, pictures and videos are recorded for appeal and arbitration, reducing labor costs.
2. According to claim 1, the intelligent timing system for sprint tests based on deep learning and human body tracking, the sports sprint test management cloud platform includes candidate information, test area and test site information, evaluation criteria, score records, key images, objection handling and report generation.
3. The intelligent sprint test timing system based on deep learning and human posture tracking according to claim 1 also includes a verification and runway allocation terminal that can verify whether the candidate is taking the test on behalf of others to ensure consistency with the registration; and allocate runways in batches according to the number of runways at each test site to improve test efficiency.
4. According to the intelligent timing system for sprint tests based on deep learning and human posture tracking described in claim 3, the verification and runway allocation terminal verifies whether the candidate has taken the test on behalf of others to ensure that it is consistent with the registration, including scanning the admission ticket and verifying the candidate information, comparing it with the photographed face, judging whether the verification is consistent, prompting the verification result, generating and uploading the verification result, and allocating the current batch of fixed runways; the candidate information includes the candidate information and configuration local storage and the host control update synchronization of the candidate information and configuration, and feedback of the results; the verification information includes generating a verification record, local storage and upload.
5. The intelligent sprint test timing system based on deep learning and human posture tracking according to claim 1 is characterized by: The sports sprint test management cloud platform includes a user layer, a business layer, a data layer, and a terminal management layer. The user layer includes a PC browser, the business layer includes test area and test site management, candidate management, verification management, score management, objection management, report statistics, large-screen display, and system settings. The data layer includes terminal data collection, data exchange, data verification, data storage, and monitoring storage. The terminal management layer includes terminal management and platform communication. The connection layer includes 4G / 5G / wifi.
6. According to the intelligent timing system for sprint tests based on deep learning and human posture tracking described in claim 2, the candidate management of the sports sprint test management cloud platform includes the management of candidate basic information through addition, deletion, modification and query, supports the import of registration files, supports the import of image files for face recognition, and examination processing includes the modification of information; examination area and examination site management includes the management of basic examination site information through addition, deletion, modification and query, and maintains scoring rules; verification management includes candidate verification records and track records; score management includes terminal device collection records, candidate test records, and test scores including support for key video key frame screenshots; report statistics include test score statistics, test item data statistics, test site data statistics, each school's test list or non-test list, and school progress statistics; System settings include permission role management, login account management, and log management; terminal management includes terminal management; terminal communication package verification information upload, command image upload, start time upload, and candidate track allocation information upload; Data storage includes candidate information data, score data and score printing data; Third-party data exchange includes connecting DBF files with the registration system and connecting DBF and EXCEL files with the score system.
7. The intelligent sprint test timing system based on deep learning and human posture tracking according to claim 1, which uses SNTP for precise time calibration, is characterized by: The start command terminal and the end point detection terminal are precisely calibrated with the cloud platform through SNTP to ensure accurate timing.
8. The intelligent sprint test timing system based on deep learning and human posture tracking according to claim 1, which realizes automatic timing through deep learning and real-time human posture tracking, is characterized by: The deep learning-based object detection algorithm captures image data in real-world exam scenarios. Each video frame generates an image. Data augmentation is then performed, transforming the raw image data to increase the data volume and diversity, thereby improving the model's robustness and generalization capabilities. Data cleaning is then performed to filter and remove low-quality, duplicate, and incorrectly labeled data, thereby improving model training effectiveness and accuracy. Outdoor lighting conditions are complex, with numerous shadows and shadows, and achieving the desired accuracy without labeled learning is difficult. The system utilizes a multi-branch feature extraction network to extract features at different scales, thereby improving the model's detection capabilities for objects of varying sizes. The system also uses a pre-trained model for initialization, accelerating model convergence and improving accuracy. For each image, human posture is detected, specifically the athlete's body movement, and the time between touches is recorded. After extensive training, the model can accurately identify and track athletes. The human tracking module, receiving the output of the deep learning model, uses an efficient algorithm to accurately track the athletes. This module can handle complex scenarios, such as occlusion and motion changes between athletes. The results are displayed on a large screen in real time and announced. The system first uses a deep learning model to identify and track athletes from images captured by the camera.
9. The intelligent sprint test timing system based on deep learning and human posture tracking according to claim 1, characterized in that: The system records the arrival time, pictures and videos for appeals and arbitration. The issuing terminal records the picture of the start time of the order and the video after the order is issued; the end point detection terminal records the picture and time when the candidate arrives at the finish line.
10. The intelligent sprint test timing system based on deep learning and human posture tracking according to claim 8, characterized in that: Each terminal timed two lanes, with a camera placed on the line between the two lanes to prevent collisions between runners. An additional terminal was added to duplicate the timing of both lanes, with the platform calculating the timing data from both terminals to ensure system accuracy, high availability, and high stability.