Invigilation method and equipment for middle and long distance running and storage medium

Abstract

The invention relates to the field of physical education invigilation, and discloses an invigilation method and equipment for middle and long distance running and a storage medium. The method comprises the following steps: monitoring human skeleton actions in a starting area in real time; when it is monitored that the human skeleton action is a hand raising action, human face recognition is performed on a human body corresponding to the human skeleton action, and a human face recognition result is obtained; comparing the face recognition result with a preset registration database to obtain a comparison result; when the comparison result is qualified, writing the face id corresponding to the comparison result into a preset competition participation list; based on a ByteTrack cross-frame association algorithm, generating a trajectory flow corresponding to the face id; monitoring human body leg key points and a face id in a terminal point area in real time; generating a line punching moment based on the relationship between the human leg key points and the finishing line; and generating a running score corresponding to the face id according to the track flow and the line punching moment. In the embodiment of the invention, the monitoring system does not need additional wearable equipment, and a single set of system is low in cost and can stably operate in various natural environments.

Description

Technical Field

[0001] This invention relates to the field of sports invigilation, and more particularly to a method, equipment, and storage medium for invigilating middle- and long-distance running. Background Technology

[0002] The existing technical system has revealed significant structural flaws in actual operation: manual judgment still dominates, and its drawbacks are mainly reflected in four aspects: it is difficult for invigilators to verify the faces of hundreds of candidates, resulting in a high risk of substitutes and proxy test takers; stopwatch timing is affected by the teacher's reaction speed, perspective differences, and environmental noise, with errors often reaching 0.3 to 0.5 seconds, leading to frequent disputes over ranking; manual copying of scores and re-entry into the system results in a secondary transcription error rate of about 1.2%, and the subsequent appeal and review process is cumbersome; 2 to 3 teachers / groups are required throughout the process, and labor costs increase linearly with the scale of the examination, and high-intensity continuous work can easily lead to fatigue and misjudgment.

[0003] While infrared photoelectric or RFID solutions improve automation, they introduce new pain points: 8-16 sets of gantry frames and ground loop coils need to be deployed on a 400-meter track, with a single set of hardware costing over 60,000 yuan. Furthermore, they are sensitive to sunlight, rain, and fog, with a false trigger rate of approximately 2%, resulting in high maintenance costs. Chip-based bib numbers are easily bent and lost, leading to additional expenses for replacement. Therefore, to address the current technical problems of high cost and poor environmental adaptability in automated monitoring of middle- and long-distance running, a new technology is needed. Summary of the Invention

[0004] The main objective of this invention is to solve the technical problems of high cost and poor environmental adaptability in automated monitoring of middle and long-distance running.

[0005] The first aspect of this invention provides a method for monitoring middle- and long-distance running, comprising the following steps: Real-time monitoring of human skeletal movements in the starting area; When the monitored human skeleton movement is a hand-raising movement, then face recognition is performed on the human body corresponding to the human skeleton movement to obtain the face recognition result. The facial recognition results are compared with a pre-set registration database to obtain the comparison results; When the comparison result is qualified, the face ID corresponding to the comparison result is written into the preset contestant list; Based on the pre-set ByteTrack cross-frame association algorithm, the face IDs in the contestant list are tracked and processed to generate the trajectory stream corresponding to the face IDs; Real-time monitoring of key points on the human legs in the endpoint area and the corresponding face IDs of the human legs; Based on the relationship between the key points of the human leg and the finish line, the finishing time corresponding to the face ID is generated; Based on the trajectory flow and the finish line time, the running score corresponding to the face ID is generated.

[0006] Optionally, in a first implementation of the first aspect of the present invention, the step of real-time monitoring of human skeletal movements in the starting area includes: Using a pre-set YOLO-pose algorithm, the key points of the human skeleton in the starting area are monitored in real time to obtain the key point coordinates of the wrist, shoulder, and elbow. The key point coordinates of the wrist include the ordinate of the wrist key point, and the key point coordinates of the shoulder include the ordinate of the shoulder key point. A first monitoring vector is generated based on the key point coordinates of the wrist and the key point coordinates of the elbow. A second monitoring vector is generated based on the key point coordinates of the shoulder and the key point coordinates of the elbow. Based on the first monitoring vector and the second monitoring vector, a detection angle is generated; Determine whether the detected angle is greater than a preset angle threshold, and determine whether the ordinate of the wrist key point is greater than the ordinate of the shoulder key point; When the detected angle is greater than the preset angle threshold and the vertical coordinate of the wrist key point is greater than the vertical coordinate of the shoulder key point, the human skeletal movement is determined to be a hand-raising movement.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the steps of determining whether the detection angle is greater than a preset angle threshold and determining whether the longitudinal coordinate of the wrist key point is greater than the longitudinal coordinate of the shoulder key point include: Extract the t-th frame image of the real-time monitoring starting area, determine whether the detection angle corresponding to the t-th frame image is greater than the preset angle threshold, and determine whether the vertical coordinate of the wrist key point corresponding to the t-th frame image is greater than the vertical coordinate of the shoulder key point corresponding to the t-th frame image, where t is a positive integer; When the detected angle corresponding to the t-th frame image is greater than the preset angle threshold and the ordinate of the wrist key point corresponding to the t-th frame image is greater than the ordinate of the shoulder key point corresponding to the t-th frame image, then extract the (t+1)-th to (t+a)-th frames of the real-time monitoring starting area, determine whether the detected angles corresponding to the (t+1)-th to (t+a)-th frames are all greater than the preset angle threshold, and determine whether the ordinates of the wrist key points in the (t+1)-th to (t+a)-th frames are all greater than the ordinates of the shoulder key points in the (t+1)-th to (t+a)-th frames, where a is a positive integer; When the detected angles are all greater than the preset angle threshold and the ordinates of the wrist key points in the images from frame t+1 to frame t+a are all greater than the ordinates of the shoulder key points in the images from frame t+1 to frame t+a, then it is determined that the detected angles are greater than the preset angle threshold and the ordinates of the wrist key points are greater than the ordinates of the shoulder key points.

[0008] Optionally, in a third implementation of the first aspect of the present invention, the key points of the human skeleton include: key point coordinates of the nose and key point coordinates of both shoulders, and the step of performing face recognition on the human body corresponding to the human skeleton movement to obtain the face recognition result includes: Based on the key point coordinates of the nose and the key point coordinates of the shoulders, the human body corresponding to the human skeleton movement is subjected to face cropping processing to obtain a face ROI image. According to the preset face recognition algorithm, face recognition is performed on the face ROI image to obtain the face recognition result.

[0009] Optionally, in a fourth implementation of the first aspect of the present invention, the key points of the human leg in the real-time monitoring endpoint area and the face ID corresponding to the key points of the human leg include: Using the pre-set YOLO-pose algorithm, the key points of the human leg in the endpoint area are monitored in real time to obtain the key point coordinates of the knee and the key point coordinates of the ankle. Extract the facial image corresponding to the key points of the human leg; Based on the preset face recognition algorithm and the list of participants, the face image is subjected to face recognition comparison processing to obtain the face ID corresponding to the key points of the human leg.

[0010] Optionally, in a fifth implementation of the first aspect of the present invention, the step of extracting the facial image corresponding to the key points of the human leg includes: Query the coordinates of the nose and the shoulders corresponding to the key points of the legs; Based on the key point coordinates of the nose and the key point coordinates of the shoulders, the human body corresponding to the key points of the legs is subjected to face cropping processing to obtain a face image.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, generating the finishing time corresponding to the face ID based on the relationship between the key points of the human leg and the finish line includes: The key point coordinates of the knee and the key point coordinates of the ankle are connected to obtain the determination line; When the determination line crosses the finish line, the shooting time corresponding to the determination line is determined as the crossing time corresponding to the face ID.

[0012] Optionally, in the seventh implementation of the first aspect of the present invention, generating the running result corresponding to the face ID based on the trajectory flow and the finish line time includes: Determine whether all preset check-in points exist in the trajectory flow; If not all of the preset check-in points exist, the abnormal running status will be determined as the running score corresponding to the face ID. When all the preset check-in points exist, the offset value of the trajectory flow is calculated, and it is determined whether the offset value is less than the preset offset threshold. When the deviation is not less than the preset offset threshold, the abnormal running status is determined as the running score corresponding to the face ID. When the offset is less than a preset threshold, the running result corresponding to the face ID is generated based on the crossing time.

[0013] A second aspect of the present invention provides a monitoring device for middle- and long-distance running, comprising: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a circuit; the at least one processor invokes the instructions in the memory to cause the monitoring device for middle- and long-distance running to execute the aforementioned monitoring method for middle- and long-distance running.

[0014] A third aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned method for monitoring middle and long-distance running.

[0015] In this embodiment of the invention, a panoramic camera and an edge computing box capture real-time track footage. Human skeletal motion analysis and facial recognition are used to bind identities in a registration database. During the run, the ByteTrack cross-frame association algorithm tracks and processes the facial IDs in the participant list, generating a trajectory stream corresponding to each facial ID. At the finish line, the distance relationship between key points on the legs and the finish line is identified to generate the crossing time. Finally, combining the trajectory stream and the crossing time, the running score for each facial ID is output. No additional wearable equipment is required throughout the process, resulting in low system cost. The system can operate stably in cloudy weather, at night, and in light rain, eliminating human error and significantly reducing construction and maintenance costs. This solves the technical problems of high cost and poor environmental adaptability in automated monitoring of middle- and long-distance running. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of an embodiment of the monitoring method for middle- and long-distance running in this invention; Figure 2 This is a schematic diagram of a specific embodiment of the 101 steps of the monitoring method for middle- and long-distance running in this invention. Figure 3 This is a schematic diagram of a specific embodiment of the 1015 steps of the monitoring method for middle- and long-distance running in this invention. Figure 4 This is a schematic diagram of a specific embodiment of the 106 steps of the monitoring method for middle- and long-distance running in this invention. Figure 5 This is a schematic diagram of one embodiment of the monitoring equipment for middle and long-distance running in this invention. Detailed Implementation

[0017] This invention provides a monitoring method, device, and storage medium for middle- and long-distance running.

[0018] The embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0019] In the description of the embodiments disclosed in this invention, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 An embodiment of the monitoring method for middle- and long-distance running in this invention includes the following steps: 101. Real-time monitoring of human skeletal movements in the starting area; In this embodiment, a camera is placed in the starting area of ​​a middle-distance race to capture the skeletal movements of each person in the starting area in real time.

[0021] For details, please refer to Figure 2 , Figure 2 This is a specific embodiment of step 101 of the monitoring method for middle- and long-distance running in this invention. Step 101 includes the following specific implementation methods: 1011. Using the preset YOLO-pose algorithm, the key points of the human skeleton in the starting area are monitored in real time to obtain the key point coordinates of the wrist, the key point coordinates of the shoulder, and the key point coordinates of the elbow. The key point coordinates of the wrist include the vertical coordinate of the wrist key point, and the key point coordinates of the shoulder include the vertical coordinate of the shoulder key point. 1012. Based on the key point coordinates of the wrist and the key point coordinates of the elbow, generate a first monitoring vector; 1013. Based on the key point coordinates of the shoulder and the key point coordinates of the elbow, generate a second monitoring vector; 1014. Based on the first monitoring vector and the second monitoring vector, generate the detection angle; 1015. Determine whether the detected angle is greater than a preset angle threshold, and determine whether the vertical coordinate of the wrist key point is greater than the vertical coordinate of the shoulder key point; 1016. When the detected angle is greater than the preset angle threshold and the vertical coordinate of the wrist key point is greater than the vertical coordinate of the shoulder key point, the human skeletal movement is determined to be a hand-raising movement.

[0022] In steps 1011-1016, a pre-set YOLO-pose algorithm is used to monitor the key points of the human skeleton in the starting area in real time. Each human skeleton has 17 key points, and each person is bounded and marked, with the coordinates of the key points also marked. When determining whether the human skeleton movement is a hand-raising motion, the key points of the wrist, shoulder, and elbow are analyzed. Each of these three key points has at least two parameters: horizontal and vertical coordinates. The first monitoring vector is obtained by subtracting the elbow key point coordinates from the wrist key point coordinates, and the second monitoring vector is obtained by subtracting the shoulder key point coordinates from the elbow key point coordinates. Then, the angle between the first and second monitoring vectors is calculated to obtain the detection angle.

[0023] The above image shows a single frame captured to monitor key skeletal points in the starting area, from which the detection angle is calculated. It then checks if the detection angle exceeds a preset threshold (e.g., 30 degrees) and simultaneously analyzes whether the wrist keypoint's ordinate is greater than the shoulder keypoint's ordinate, essentially determining if the wrist is higher than the shoulder. If both the detection angle and the wrist keypoint's ordinate are greater than the shoulder keypoint's ordinate, the skeletal movement is determined to be a hand-raising motion.

[0024] For further details, please refer to [link / reference]. Figure 3 , Figure 3 This is a specific embodiment of step 1015 of the monitoring method for middle- and long-distance running in this invention, which includes the following specific implementation methods: 10151. Extract the t-th frame image of the real-time monitoring starting area, determine whether the detection angle corresponding to the t-th frame image is greater than the preset angle threshold, and determine whether the vertical coordinate of the wrist key point corresponding to the t-th frame image is greater than the vertical coordinate of the shoulder key point corresponding to the t-th frame image, where t is a positive integer; 10152. When the detection angle corresponding to the t-th frame image is greater than the preset angle threshold and the ordinate of the wrist key point corresponding to the t-th frame image is greater than the ordinate of the shoulder key point corresponding to the t-th frame image, then extract the t+1-th frame image to the t+a-th frame image of the real-time monitoring starting area, determine whether the detection angles corresponding to the t+1-th frame image to the t+a-th frame image are all greater than the preset angle threshold, and determine whether the ordinate of the wrist key point in the t+1-th frame image to the t+a-th frame image is all greater than the ordinate of the shoulder key point in the t+1-th frame image to the t+a-th frame image, where a is a positive integer; 10153. When the detected angles are all greater than the preset angle threshold and the ordinates of the wrist key points in the images from frame t+1 to frame t+a are all greater than the ordinates of the shoulder key points in the images from frame t+1 to frame t+a, then it is determined that the detected angles are greater than the preset angle threshold and the ordinates of the wrist key points are greater than the ordinates of the shoulder key points.

[0025] In steps 10151-10153, during the analysis of whether the detection angle is greater than a preset angle threshold and whether the wrist keypoint's ordinate is greater than the shoulder keypoint's ordinate, the analysis first checks whether the detection angle in frame t is greater than the preset angle threshold, and then analyzes whether the wrist keypoint's ordinate in frame t is also greater than the corresponding shoulder keypoint's ordinate in frame t. If only two analysis elements are greater than their corresponding indices, then the hand-raising action bar is considered to be satisfied in frame t.

[0026] However, to enhance the robustness of hand-raising recognition, the system was designed to analyze five consecutive frames where the detection angle was greater than a preset threshold, and the ordinate of the wrist keypoint in each of these five frames was greater than the ordinate of the shoulder keypoint. Specifically, this means checking if the detection angles from frame t+1 to t+4 were all greater than the preset threshold, and if the ordinates of the wrist keypoints from frame t+1 to t+a were all greater than the ordinates of the shoulder keypoints from frame t+1 to t+a. This would demonstrate that the candidate's continuous hand-raising was not a random misjudgment, and that the skeletal movement was being mistaken for a hand-raising action. All these steps are designed to increase robustness and prevent misrecognition.

[0027] 102. When the monitored human skeleton movement is a hand-raising movement, then face recognition is performed on the human body corresponding to the human skeleton movement to obtain the face recognition result; In this embodiment, if the monitored human skeleton movement is a hand-raising movement, then the face image data is found based on the human skeleton, and then face recognition processing is performed on the face image data to obtain the feature result of face recognition.

[0028] Specifically, the key points of the human skeleton include: the key point coordinates of the nose and the key point coordinates of both shoulders. Step 102 includes the following specific implementation methods: 1021. Perform face cropping on the human body corresponding to the human body bone movement based on the key point coordinates of the nose and the key point coordinates of the shoulders to obtain a face ROI image; 1022. Perform face recognition on the face ROI image according to a preset face recognition algorithm to obtain a face recognition result.

[0029] In steps 1021 - 1022, the selection of the face ROI needs to be determined according to specific application scenarios (such as avoiding background noise, blood vessel distribution, etc.). Common methods include based on key points, skin color detection, and connected component labeling, etc. Use YOLO - Pose to detect the key points of the nose and shoulders. Take the key point coordinates of the nose as the center point, take both ends of the key point coordinates of the shoulders as the width, and take 1.2 times the distance between the shoulders as the length to generate a rectangular screenshot for face cropping on the human body corresponding to the human body bone movement to obtain a face ROI image.

[0030] Finally, perform face recognition on the face ROI image according to the face recognition algorithm of ArcSoft face recognition 4.0 sdk to obtain a face recognition result.

[0031] 103. Compare the face recognition result with a preset registration database to obtain a comparison result; In this embodiment, the face recognition result contains facial feature data of the face. Compare the facial feature data of the face with the face data in the registration database of the pre - registered parameters to obtain a comparison result of whether it belongs to the registration database. When the face in the face recognition result belongs to the face data registered in the registration database, the comparison result is qualified data. When the face in the face recognition result does not belong to the face data registered in the registration database, the comparison result is unqualified data.

[0032] 104. When the comparison result is qualified, write the face id corresponding to the comparison result into a preset competition list; In this embodiment, when the comparison result belongs to qualified data, write the face id corresponding to the comparison result into a preset competition list, indicating that the current running participant has actually arrived and is ready to participate in the competition.

[0033] 105. Based on a preset ByteTrack cross - frame association algorithm, perform tracking processing on the face ids in the competition list to generate a trajectory stream corresponding to the face ids; In this embodiment, store the face id of the identity feature in a db file and transfer it to the ByteTrack cross - frame association algorithm. The core mechanism of ByteTrack in the cross - frame target association algorithm can be summarized as: "two - stage matching + low - confidence box retention + simple motion model" First, the YOLOv8 object detection algorithm is used to define bounding boxes for existing runners. Then, all bounding boxes are matched to generate high-confidence and low-confidence boxes. The high-confidence boxes are associated with the existing trajectory stream. Low-confidence boxes are continuously tracked until the object in the low-confidence box is determined by rules or discarded during a state update. Face IDs from the participant list are tracked, generating trajectory streams corresponding to each face ID. During the tracking process, the confidence probability of each athlete's bib number is increased to ensure that all participants in the run can be tracked.

[0034] 106. Real-time monitoring of key points of the human legs in the endpoint area and the face IDs corresponding to the key points of the human legs; In this embodiment, the key skeletal points of the human leg are captured and monitored in real time in the finish area of ​​the middle- and long-distance race, and the face image corresponding to the key skeletal points of the human leg is detected at the same time. The face image is matched with the face ID with the parameters already determined in the participant list to obtain the face ID corresponding to the key skeletal points of the human leg.

[0035] For details, please refer to Figure 4 , Figure 4 This is a specific embodiment of the 106 steps of the monitoring method for middle- and long-distance running in this invention, which includes the following specific implementation methods: 1061. Using the preset YOLO-pose algorithm, monitor the key points of the human leg in the endpoint area in real time to obtain the key point coordinates of the knee and the key point coordinates of the ankle. 1062. Extract the facial image corresponding to the key points of the human leg; 1063. Based on the preset face recognition algorithm and the list of participants, perform face recognition comparison processing on the face image to obtain the face ID corresponding to the key points of the human leg.

[0036] In steps 1061-1063, the preset YOLO-pose algorithm is first used to monitor the key points of the human leg in the endpoint area in real time, and the coordinates of the knee key point and the ankle key point are found from the key points of the human leg. Then, based on the key points of the human leg, the face image is found and cropped, the face image is subjected to feature recognition to obtain the face features, and the face features are compared and matched with the contestant list to obtain the face ID corresponding to the key points of the human leg.

[0037] Furthermore, step 1062 includes the following specific implementation methods: 10621. Query the coordinates of the key points of the nose and the key points of the shoulders corresponding to the key points of the human legs; 10622. Based on the key point coordinates of the nose and the key point coordinates of the shoulders, perform face cropping processing on the human body corresponding to the key points of the legs to obtain a face image.

[0038] In steps 10621-10622, human skeletal points are identified. From these points, key points on the legs are located, followed by the coordinates of the nose and shoulders. Using the nose coordinates as the center point, the two ends of the shoulders as the width, and 1.2 times the distance between the shoulders as the length, a rectangular screenshot is generated. This screenshot is then used to crop the human face corresponding to the skeletal movements, resulting in the face image.

[0039] 107. Based on the relationship between the key points of the human leg and the finish line, generate the finishing time corresponding to the face ID; In this embodiment, when the key points of the human leg cross the finish line, the timestamp of the first frame image that crosses the finish line is determined as the crossing time, and the timestamp of the starting command is taken as the start time to calculate the crossing time corresponding to the face ID.

[0040] Specifically, steps 107 include the following specific implementation methods: 1071. Connect the key point coordinates of the knee and the key point coordinates of the ankle to obtain the determination line; 1072. When the determination line crosses the finish line, the shooting time corresponding to the determination line is determined as the crossing time corresponding to the face ID.

[0041] In steps 1071-1072, the key point coordinates of the knee and ankle are first connected to generate a decision line. Then, the intersection of the decision line and the finish line is analyzed. The process of the decision line crossing the finish line is as follows: the intersection is empty at one end of the finish line; when it crosses the finish line, the intersection becomes non-empty; and the intersection is empty again at the other end of the finish line. The moment when the intersection becomes empty at the other end of the finish line is determined as the finish line crossing moment, which is also the moment the decision line crosses the finish line. The shooting time corresponding to this decision line is determined as the finish line crossing moment corresponding to the face ID.

[0042] 108. Based on the trajectory flow and the finish line time, generate the running score corresponding to the face ID.

[0043] In this embodiment, the system analyzes whether each checkpoint in the face ID trajectory stream is reached to prevent candidates from deviating from the designated route. Then, it analyzes the deviation between the face ID trajectory stream and the standard trajectory stream to prevent candidates from taking shortcuts. When the trajectory stream is satisfactory, the starting time is subtracted from the finish line time to obtain the running time corresponding to the face ID, which is the candidate's running score.

[0044] Specifically, the 108 steps include the following specific implementation methods: 1081. Determine whether all preset check-in points exist in the trajectory flow; 1082. If none of the preset check-in points exist, the abnormal running status will be determined as the running score corresponding to the face ID. 1083. When all the preset check-in points exist, calculate the offset value of the trajectory flow and determine whether the offset value is less than the preset offset threshold. 1084. When the deviation is not less than the preset offset threshold, the abnormal running status is determined as the running score corresponding to the face ID. 1085. When the offset is less than the preset threshold, the running result corresponding to the face ID is generated based on the crossing time.

[0045] In steps 1081-1085, it is first determined whether each checkpoint set in the trajectory stream exists. If each checkpoint exists, it means that the examinee corresponding to the face ID did not deviate from the designated route. If not each checkpoint exists, it means that the examinee corresponding to the face ID deviated from the designated route, and the running score is not qualified. The abnormal running status of "deviating from the designated route" can be identified as the running score corresponding to the face ID.

[0046] All checkpoints are within the trajectory flow, and the deviation value of the trajectory flow is calculated using dynamic time warping.

[0047] Let the reference trajectory be Q = {q1, q2, ..., qn}, and the trajectory of the trajectory flow be C = {c1, c2, ..., cn}. Calculate the distance value corresponding to each node, align the reference trajectory with the start and end points of the trajectory flow, calculate the cumulative deviation distance between the two trajectories, and obtain the offset value.

[0048] The system checks if the offset value is less than a preset offset threshold. If it is, it means the candidate did not take a shortcut to avoid the test path. Based on the difference between the finish line time and the start time, a running score corresponding to the face ID is generated. If the offset value is not less than the preset offset threshold, it means the candidate did take a shortcut to avoid the test path. In this case, the abnormal running status of "taking a shortcut to avoid the test path" is identified as the running score corresponding to the face ID.

[0049] In this embodiment of the invention, a panoramic camera and an edge computing box capture real-time track footage. Human skeletal motion analysis and facial recognition are used to bind identities in a registration database. During the run, the ByteTrack cross-frame association algorithm tracks and processes the facial IDs in the participant list, generating a trajectory stream corresponding to each facial ID. At the finish line, the distance relationship between key points on the legs and the finish line is identified to generate the crossing time. Finally, combining the trajectory stream and the crossing time, the running score for each facial ID is output. No additional wearable equipment is required throughout the process, resulting in low system cost. The system can operate stably in cloudy weather, at night, and in light rain, eliminating human error and significantly reducing construction and maintenance costs. This solves the technical problems of high cost and poor environmental adaptability in automated monitoring of middle- and long-distance running.

[0050] Figure 5 This is a schematic diagram of the structure of a monitoring device for middle- and long-distance running provided in an embodiment of the present invention. The monitoring device 500 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) for storing application programs 533 or data 532. The memory 520 and storage media 530 can be temporary or persistent storage. The program stored in the storage media 530 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the monitoring device 500. Furthermore, the processor 510 may be configured to communicate with the storage media 530 and execute the series of instruction operations in the storage media 530 on the monitoring device 500.

[0051] The monitoring device 500 for middle- and long-distance running may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input / output interfaces 560, and / or one or more operating systems 531, such as Windows Server, Mac OS X, Unix, Linux, Free BSD, etc. Those skilled in the art will understand that... Figure 5 The structure of the monitoring equipment shown for middle and long distance running does not constitute a limitation on monitoring equipment based on middle and long distance running. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0052] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the proctoring method for the middle and long distance running.

[0053] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0054] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0055] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for monitoring middle- and long-distance running, characterized in that, Including the following steps: Real-time monitoring of human skeletal movements in the starting area; When the monitored human skeleton movement is a hand-raising movement, then face recognition is performed on the human body corresponding to the human skeleton movement to obtain the face recognition result. The facial recognition results are compared with a pre-set registration database to obtain the comparison results; When the comparison result is qualified, the face ID corresponding to the comparison result is written into the preset contestant list; Based on the pre-set ByteTrack cross-frame association algorithm, the face IDs in the contestant list are tracked and processed to generate the trajectory stream corresponding to the face IDs; Real-time monitoring of key points on the human legs in the endpoint area and the corresponding face IDs of the human legs; Based on the relationship between the key points of the human leg and the finish line, the finishing time corresponding to the face ID is generated; Based on the trajectory flow and the finish line time, the running score corresponding to the face ID is generated.

2. The invigilation method for middle- and long-distance running according to claim 1, characterized in that, The steps for real-time monitoring of human skeletal movements in the starting area include: Using a pre-set YOLO-pose algorithm, the key points of the human skeleton in the starting area are monitored in real time to obtain the key point coordinates of the wrist, shoulder, and elbow. The key point coordinates of the wrist include the ordinate of the wrist key point, and the key point coordinates of the shoulder include the ordinate of the shoulder key point. A first monitoring vector is generated based on the key point coordinates of the wrist and the key point coordinates of the elbow. A second monitoring vector is generated based on the key point coordinates of the shoulder and the key point coordinates of the elbow. Based on the first monitoring vector and the second monitoring vector, a detection angle is generated; Determine whether the detected angle is greater than a preset angle threshold, and determine whether the ordinate of the wrist key point is greater than the ordinate of the shoulder key point; When the detected angle is greater than the preset angle threshold and the vertical coordinate of the wrist key point is greater than the vertical coordinate of the shoulder key point, the human skeletal movement is determined to be a hand-raising movement.

3. The invigilation method for middle- and long-distance running according to claim 2, characterized in that, The steps of determining whether the detection angle is greater than a preset angle threshold and whether the ordinate of the wrist key point is greater than the ordinate of the shoulder key point include: Extract the t-th frame image of the real-time monitoring starting area, determine whether the detection angle corresponding to the t-th frame image is greater than the preset angle threshold, and determine whether the vertical coordinate of the wrist key point corresponding to the t-th frame image is greater than the vertical coordinate of the shoulder key point corresponding to the t-th frame image, where t is a positive integer; When the detected angle corresponding to the t-th frame image is greater than the preset angle threshold and the ordinate of the wrist key point corresponding to the t-th frame image is greater than the ordinate of the shoulder key point corresponding to the t-th frame image, then extract the (t+1)-th to (t+a)-th frames of the real-time monitoring starting area, determine whether the detected angles corresponding to the (t+1)-th to (t+a)-th frames are all greater than the preset angle threshold, and determine whether the ordinates of the wrist key points in the (t+1)-th to (t+a)-th frames are all greater than the ordinates of the shoulder key points in the (t+1)-th to (t+a)-th frames, where a is a positive integer; When the detected angles are all greater than the preset angle threshold and the ordinates of the wrist key points in the images from frame t+1 to frame t+a are all greater than the ordinates of the shoulder key points in the images from frame t+1 to frame t+a, then it is determined that the detected angles are greater than the preset angle threshold and the ordinates of the wrist key points are greater than the ordinates of the shoulder key points.

4. The invigilation method for middle- and long-distance running according to any one of claims 2 or 3, characterized in that, The key points of the human skeleton include: the coordinates of the key points of the nose and the coordinates of the key points of both shoulders. The step of performing facial recognition on the human body corresponding to the movement of the human skeleton to obtain the facial recognition result includes: Based on the key point coordinates of the nose and the key point coordinates of the shoulders, the human body corresponding to the human skeleton movement is subjected to face cropping processing to obtain a face ROI image. According to the preset face recognition algorithm, face recognition is performed on the face ROI image to obtain the face recognition result.

5. The invigilation method for middle- and long-distance running according to claim 1, characterized in that, The steps of real-time monitoring of key points of the human leg in the endpoint area and the corresponding face ID of the human leg key points include: Using the pre-set YOLO-pose algorithm, the key points of the human leg in the endpoint area are monitored in real time to obtain the key point coordinates of the knee and the key point coordinates of the ankle. Extract the facial image corresponding to the key points of the human leg; Based on the preset face recognition algorithm and the list of participants, the face image is subjected to face recognition comparison processing to obtain the face ID corresponding to the key points of the human leg.

6. The invigilation method for middle- and long-distance running according to claim 5, characterized in that, The step of extracting the facial image corresponding to the key points of the human leg includes: Query the coordinates of the nose and the shoulders corresponding to the key points of the legs; Based on the key point coordinates of the nose and the key point coordinates of the shoulders, the human body corresponding to the key points of the legs is subjected to face cropping processing to obtain a face image.

7. The invigilation method for middle- and long-distance running according to claim 5 or 6, characterized in that, The step of generating the finishing time corresponding to the face ID based on the relationship between the key points of the human leg and the finish line includes: The key point coordinates of the knee and the key point coordinates of the ankle are connected to obtain the determination line; When the determination line crosses the finish line, the shooting time corresponding to the determination line is determined as the crossing time corresponding to the face ID.

8. The invigilation method for middle- and long-distance running according to claim 1, characterized in that, The step of generating the running result corresponding to the face ID based on the trajectory flow and the finish line time includes: Determine whether all preset check-in points exist in the trajectory flow; If not all of the preset check-in points exist, the abnormal running status will be determined as the running score corresponding to the face ID. When all the preset check-in points exist, the offset value of the trajectory flow is calculated, and it is determined whether the offset value is less than the preset offset threshold. When the deviation is not less than the preset offset threshold, the abnormal running status is determined as the running score corresponding to the face ID. When the offset is less than a preset threshold, the running result corresponding to the face ID is generated based on the crossing time.

9. A monitoring device for middle- and long-distance running, characterized in that, The monitoring equipment for the middle and long distance running includes: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a circuit; The at least one processor invokes the instructions in the memory to cause the middle- and long-distance running monitoring device to perform the middle- and long-distance running monitoring method as described in any one of claims 1-8.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the monitoring method for middle- and long-distance running as described in any one of claims 1-8.

Images