Middle and long distance running posture analysis method, device and equipment and storage medium

Abstract

The invention provides a middle-distance running posture analysis method, device and equipment and a storage medium, and the method comprises the following steps: deploying a camera at the outer side of a runway to shoot a motion video, identifying and tracking a current athlete in the motion video, carrying out calibration processing on space coordinates of the current athlete, and forming a motion time sequence of coordinate information after calibration; performing quantitative analysis based on the motion time sequence to obtain a quantitative analysis result; based on the quantitative analysis result, risk early warning is prompted through threshold judgment and a machine learning model, and the machine learning model considers mutually associated balance points in quantitative analysis indexes; and acquiring exercise physiological data of the current athlete, forming structured data by the exercise physiological data, the risk early warning and the quantitative analysis result, and generating a guidance scheme based on the structured data. According to the method, a middle-distance running training mode is converted into a data quantization driving mode from a subjective experience driving mode.

Description

Technical Field

[0001] This invention relates to the field of running posture analysis technology, and in particular to a method, apparatus, device, and storage medium for analyzing running posture in middle and long distance running. Background Technology

[0002] Middle- and long-distance running refers to track and field events with a distance of 800 meters or more. Persisting in middle- and long-distance running can have a positive effect on improving metabolic rate and enhancing the immune system. However, incorrect running posture can lead to sports injuries such as muscle strain, tendinitis, and arthritis. Therefore, accurate analysis of running posture in middle- and long-distance running is crucial.

[0003] In related technologies, traditional solutions for middle- and long-distance running use exercise duration as the core training indicator, without developing an awareness of running posture analysis. Furthermore, most existing analysis solutions rely on a single data source, evaluating performance by collecting partial physiological data, lacking systematic feedback on movement behavior and postural biomechanical characteristics, and failing to comprehensively understand the athlete's condition. Moreover, current judgment methods largely rely on static threshold rules, a one-size-fits-all approach that cannot adapt to individual differences among athletes. In existing video posture analysis, common anti-shake algorithms are designed for the entire frame, without considering the characteristics of middle- and long-distance running, potentially resulting in over-smoothing when fast motion details need to be preserved, or retaining jitter when stability is required. In addition, existing machine learning-based methods are rather general and lack specificity, failing to consider the causal relationships between features during middle- and long-distance running.

[0004] Based on the above analysis of the development status of this technology field, the existing technologies lack a solution that uses computer vision technology to process images with corresponding calibration mechanisms under different conditions, and that uses multi-source data to transform the training mode of middle and long-distance running from a subjective experience-driven mode to a data quantification-driven mode. Summary of the Invention

[0005] The purpose of this invention is to provide a method, apparatus, device, and storage medium for analyzing running posture in middle and long distance running, in order to solve the aforementioned problems in the prior art.

[0006] According to a first aspect of the present invention, a method for analyzing running posture in middle- and long-distance running is provided, comprising: Cameras are deployed outside the track to capture motion videos. The current athlete is identified and tracked in the motion videos. The spatial coordinates of the current athlete are calibrated, and a motion time series of calibrated coordinate information is generated. Quantitative analysis is performed based on motion time series to obtain quantitative analysis results; Based on the results of quantitative analysis, risk warnings are issued using threshold judgment and machine learning models, respectively. The machine learning model considers the balance point of the interrelationship among the quantitative analysis indicators. Acquire current athletes' exercise physiological data, combine the exercise physiological data, risk warnings, and quantitative analysis results into structured data, and generate guidance plans based on the structured data.

[0007] According to a second aspect of the present invention, a running posture analysis device for middle and long distance running is provided, comprising: The initial processing module is used to deploy cameras outside the track to capture motion videos, identify and track the current athlete in the motion videos, calibrate the spatial coordinates of the current athlete, and form a motion time series of calibrated coordinate information. The quantitative analysis module is used to perform quantitative analysis based on motion time series and obtain quantitative analysis results. The risk warning module is used to issue risk warnings based on the results of quantitative analysis, using threshold judgment and machine learning model respectively. The machine learning model considers the balance point of the interrelationship among the quantitative analysis indicators. The training guidance module is used to acquire the athlete's current exercise physiological data, and to form structured data from the exercise physiological data, risk warnings, and quantitative analysis results. Based on the structured data, guidance plans are generated.

[0008] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the running posture analysis method for middle and long distance running as provided in the first aspect of the present disclosure.

[0009] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which an information transmission implementation program is stored, which, when executed by a processor, implements the steps of the middle- and long-distance running posture analysis method provided in the first aspect of the present disclosure.

[0010] The technical solution provided by the embodiments of the present invention has the following beneficial effects: it transforms human motion posture information into data that can be quantified and calculated, and uses two methods, threshold judgment and machine learning model, to provide risk warnings, avoid ignoring individual differences, and transforms posture analysis from a fuzzy, subjective experience-based inference mode to a highly accurate matching mode.

[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the running posture analysis method for medium and long distance running according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the camera deployment in an embodiment of the present invention; Figure 3 This is a schematic diagram of pedestrian interference filtering according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the device used in an embodiment of the present invention; Figure 5 This is a schematic diagram of the overall architecture of an embodiment of the present invention; Figure 6 This is a schematic diagram of a long-distance running posture analysis device according to an embodiment of the present invention; Figure 7 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0015] Method Implementation Examples According to an embodiment of the present invention, a method for analyzing running posture in middle- and long-distance running is provided. Figure 1 This is a flowchart of the running posture analysis method for medium and long distance running according to an embodiment of the present invention, as shown below. Figure 1 As shown, the running posture analysis method for middle and long distance running according to an embodiment of the present invention specifically includes: In step S110, cameras are deployed outside the track to capture motion video. The current athlete is identified and tracked in the motion video. The spatial coordinates of the current athlete are calibrated, and a motion time series of calibrated coordinate information is generated. Specifically, this includes: The video is captured by a camera, which is continuously and evenly positioned to ensure a continuous flow of footage. Ideally, the camera can be positioned only on one side of the key movement path; the specific configuration can be adjusted according to the actual situation. Figure 2 This is a schematic diagram of the camera site deployment according to an embodiment of the present invention, as shown below. Figure 2 The image shows how the cameras are deployed.

[0016] The system captures motion videos using cameras, and simultaneously displays metadata information including video frame rate, resolution, focal length, and timestamp during the recording process. The cameras are evenly distributed throughout the track, and facial recognition is used to match the current athlete based on the content captured by the first camera at the starting point. exist Figure 2 Camera 1 is responsible for identifying the current athlete's identity, obtaining various information related to the middle- and long-distance runners through the personnel's identity information, using a target recognition algorithm to detect the pedestrian categories in the video footage, and inputting the identified pedestrian coordinate information into the target tracking engine. The engine continuously tracks the current middle- and long-distance runners through the target tracking algorithm. Cameras 2 to 8 are responsible for recording the middle- and long-distance runners' movement videos. Adjusting the camera's focal length allows observation of each track. Preferably, the identities of people in different areas of the shot are defined: those on the middle- and long-distance running track are classified as athletes, and those outside the track are classified as interfering personnel. A target tracking algorithm is used to track all personnel in the shot, update their spatial position information, and filter out interfering personnel to ensure that the middle- and long-distance runners are always being tracked and detected. Figure 3 This is a schematic diagram of pedestrian interference filtering according to an embodiment of the present invention, as shown below. Figure 3 The diagram illustrates a scenario for athlete matching and interference filtering.

[0017] The collected motion video is input into the visual recognition model as a time frame sequence. The visual recognition model extracts the skeletal joints of the human motion posture, where the skeletal joints include the spatial coordinate information of the head, limbs and torso.

[0018] The calibration process is tailored to the characteristics of middle- and long-distance running. Middle- and long-distance running is characterized by high repetitiveness of movements, strong periodicity, and stable posture. Traditional general anti-shake algorithms do not utilize these characteristics. The calibration process used in this embodiment is as follows: The visual recognition model determines whether the current frame is the athlete's foot support phase on the ground or the swing phase in the air. The Kalman filter is used to correct the athlete's spatial coordinates. If the current frame is the support phase, the captured value is given a lower weight than the Kalman filter. If the current frame is the swing phase, the captured value is given a higher weight than the Kalman filter. In a gait cycle of middle and long-distance running, the athlete's current phase is analyzed in real time. The movement characteristics are different under different conditions. During the support phase, the position of joints such as the ankle and knee in space is relatively stable and changes slowly. At this time, the results predicted by the filter model are more trusted, and the observations captured by the camera are given a lower weight, thereby strongly filtering out the lens shake that may be caused by the impact of the foot stepping on the ground. During the swing phase, the limbs are swinging rapidly, and the calibration will trust the results captured by the camera more. In this embodiment of the invention, the calibration process weighs the filter and the captured value, and preferably, further considers the impact of environmental changes on data accuracy.

[0019] Using the corrected spatial coordinates of the pelvis as anchor points to smooth the motion trajectory, we obtain calibrated coordinate information. This is because the pelvis is the most stable part of the body with the least swaying during middle and long-distance running. The relative positions of other points with the pelvis remain unchanged, thus enabling smoothing as well.

[0020] Kalman filters address jitter caused by motion characteristics, such as the high-frequency vibration of the camera caused by the impact of a foot landing, while pelvic anchors address global, consistent jitter, such as when camera instability causes the entire image to shift downwards.

[0021] In step S120, quantitative analysis is performed based on the motion time series to obtain the quantitative analysis results, specifically including: Exercise time series can convert athletes' middle- and long-distance running behaviors into calculable digital information, and calculate average stride frequency, average stride length, average flight time, average ground contact time, and skeletal joint angles based on the exercise time series. (1) Average step frequency: The moment when the key point of the foot changes from downward movement to upward movement is determined as the moment of ground contact. Let the continuous ground contact time of a single foot be 1. The total duration is The total number of steps is n-1. Calculate the average step frequency using Formula 1: Formula 1; in, This indicates the nth time a single foot touches the ground. This represents the total time until the nth time that one foot touches the ground; (2) Average stride: The straight-line distance between the points of contact when the same foot makes two consecutive contact with the ground; calculate the stride using Formula 2: Formula 2; in, This indicates the position of the i-th foot when it touches the ground. This indicates the position of the (i+1)th foot contact with the ground. (3) Average time in the air: The average time from when the foot leaves the ground to when it touches the ground again; calculate the average time in the air using Formula 3: Formula 3; in, Indicates the start frame of takeoff. This indicates the end of the takeoff frame. Indicates frame rate, Indicates the number of times the vehicle has been emptied; (4) Mean ground contact time: The average time from when the foot contacts the ground to when it leaves the ground again. Calculate the mean ground contact time using Formula 4: Formula 4; in, This indicates the start of the ground contact frame. This indicates the end of the grounding frame. Indicates frame rate, Indicates the number of times the ground touches; (5) Skeletal joint angles: Two joints are used as side points and one joint is used as the vertex to reflect the curvature of the human body posture. Among them, the head posture is calculated with the eyes and horizontal point as side points and the ears as the vertex; the torso posture is calculated with the ears and the hanging point as side points and the hip as the vertex; the arm posture is calculated with the shoulder and wrist as side points and the elbow as the vertex; the leg posture is calculated with the hip and ankle as side points and the knee as the vertex; the foot posture is calculated with the hanging point and the big toe as side points and the heel as the vertex; the body posture is calculated with the shoulder and foot as side points and the hip as the vertex.

[0022] Use Formula 5 to calculate the angle between bone joints. : Formula 5; in, and Let A and B represent the coordinates of joint edge point A and B, respectively. This represents the coordinates of the vertex O of the joint.

[0023] Preferably, although middle and long-distance running has a strong periodicity compared to other sports, the characteristics of motion parameters still differ in the starting and sprinting phases. K-means clustering analysis is performed according to different motion states to divide them into starting cluster, constant speed cluster, and sprint cluster. By statistically analyzing the motion characteristics of each cluster through a large amount of training data, such as mean, variance, and skewness, a cluster feature library is established to ensure the accuracy of clustering. The matching degree between the parameters of the middle and long-distance running motion frame and the cluster feature library is calculated, and the current motion behavior is included in the corresponding motion state. If the current motion behavior belongs to the cluster but deviates, the measured value in the formula is corrected by a smooth transition method. Then, the average cadence, average stride length, average flight time and average ground contact time are calculated to further improve the accuracy of quantitative analysis.

[0024] Preferably, in middle and long-distance running, due to the different sports scenarios and sports postures, external interference from athletes and occlusion by one's own body parts are unavoidable. Under the condition of occlusion, the joint point data will produce abnormal values ​​or be lost, affecting the accuracy of sports posture analysis. In this embodiment of the invention, an interpolation compensation method is used to improve the accuracy of skeletal joint data based on the motion characteristics of middle and long-distance running, so that the posture analysis results are more consistent with the real motion state. For short frame loss (less than or equal to 5 frames) such as sprinting acceleration and body swinging, linear interpolation is used to quickly complete the data. For long frame loss (more than 5 frames) such as external moving people blocking the view, 1-2 complete in-phase motion trajectory data before and after the lost frame are extracted. Combined with the known data of the unblocked part of the lost frame, the coordinates of the lost skeletal joints are calculated by phase shift. Furthermore, the data for interpolation and filling is further constrained by middle- and long-distance running mechanisms, such as the knee joint being located between the hip and ankle joints, to further verify the validity of the coordinate data and avoid the interpolation results from violating the laws of biomechanics.

[0025] In step S130, based on the quantitative analysis results, risk warnings are issued using threshold judgment and machine learning models, respectively. The machine learning model considers the equilibrium points of interrelationships among the quantitative analysis indicators, specifically including: Based on thresholds for average cadence and average stride length, it is determined whether there is excessive striding. Excessive striding will reduce the average cadence and increase the stride length. Based on thresholds for the angle between skeletal joints, it is determined whether there is head posture, torso posture, elbow posture, knee posture and foot posture. For example, excessive knee bending will cause the angle between skeletal joints to decrease. The threshold judgment method is relatively intuitive, but further quantitative analysis is needed to obtain deeper information from the results. All quantitative analysis results, except for the angles between skeletal joints, are input into the machine learning model, which is the XGBoost model. In middle and long-distance running, some technical parameters seem contradictory, such as stride length and cadence. Increasing stride length can reduce cadence and save heart rate, but excessive stride length means longer airtime and braking impact, increasing the possibility of injury. For example, airtime and ground contact time. Longer airtime usually means greater propulsion, but it also means higher vertical fluctuations and landing impact. Therefore, in the current scenario, there is no absolutely optimal value, but rather a balance that can both improve efficiency and reduce risk. In the XGBoost model, average cadence and average stride are treated as an interrelated interaction pair, as are average flight time and average ground contact time. When constructing each tree, only features within the same interaction pair are allowed to interact; that is, average ground contact time is not allowed to be a child or parent feature of average cadence, and so on. This forces the model to explore the complex relationships within feature pairs in depth, so as to construct a more regular tree structure. The SHAP interpreter outputs the internal analysis logic of the machine learning model and outputs risk warnings. The internal analysis logic is the process of insight, such as "This running posture scores 45 points, which is far below the average score of 70 points. The main reasons are excessive stride and excessive air time, which lowered the score by 15 points and 11.5 points respectively."

[0026] Preferably, the quantitative monitoring data is classified and graded to determine the risk level of different movement postures. By judging the risk factor level, it is determined whether the current movement behavior meets the exercise standard. If there is an abnormal risk during the exercise, a risk warning is issued and a correct handling plan is given to avoid damage to the body caused by abnormal movement behavior.

[0027] In step S140, the current athlete's exercise physiological data is acquired, and the exercise physiological data, risk warnings, and quantitative analysis results are combined into structured data. A guidance plan is then generated based on this structured data, specifically including: Acquire exercise physiological data through wearable devices; By combining exercise physiological data, risk warnings, quantitative analysis results, and metadata information obtained from photography, a structured data set including timestamps is formed. Fragmented information is integrated from multiple dimensions to generate structured exercise data. A rule engine is then used to generate targeted guidance plans, namely the IF-THEN logical rule engine, to make the effects of middle- and long-distance running more scientific and efficient. Preferably, the training cycle is divided into different phases, with a gradual transition between phases to achieve a smooth progression of load, intensity, and ability goals. Training modules are rationally allocated to address ability weaknesses, and each module is broken down into specific sessions, clearly defining the intensity, duration, movement standards, and key performance indicators for each session. This ensures that the training plan accurately matches the athlete's physiological characteristics and training needs.

[0028] The method further includes: In step S150, the exercise videos, quantitative analysis results, risk warnings, guidance plans, and subsequent training improvement information are integrated, such as the exercise videos, quantitative analysis results, and risk warnings during subsequent training. The integrated information is then stored in a pre-built middle- and long-distance running training file.

[0029] Preferably, information from the entire training cycle is integrated to identify potential problems and output improvement plans, generate preliminary exercise reports, and incorporate these reports into the training archive to provide a reference for subsequent training improvements.

[0030] The above technical solutions of the embodiments of the present invention will be illustrated with reference to the following accompanying drawings.

[0031] Figure 4 This is a schematic diagram of the device used in an embodiment of the present invention, such as... Figure 4 As shown, the equipment used in the long-distance running posture analysis process is illustrated. The visual recognition device is used to embed the model and algorithm involved in the long-distance running posture analysis method of the present invention. The visual recognition device uses the RTSP / RTMP protocol to read the camera video stream. The central cloud server is used to achieve communication through the network, so that the video or information can be played on the edge device.

[0032] Figure 5 This is a schematic diagram of the overall architecture of an embodiment of the present invention, as shown below. Figure 5 As shown, the main implementation process of the present invention includes: (1) loading the camera video stream; (2) loading the face recognition, target detection, and posture recognition model; (3) loading the target tracking engine; (4) filtering irrelevant interfering personnel; (5) face recognition of athletes; (6) reading personnel identity information; (7) tracking the current athlete; (8) middle and long-distance running action recognition and analysis; (9) sports risk warning and injury prevention intervention; (10) generating middle and long-distance running training plan; and (11) establishing middle and long-distance running training files.

[0033] In summary, addressing the existing problems, this invention provides a method for analyzing running posture in middle and long-distance running. During the calibration process, it leverages the high repeatability, strong periodicity, and stable posture of middle and long-distance running movements, categorizing each movement cycle into a support phase and a swing phase to specifically improve calibration effectiveness. It transforms human movement posture information into quantifiable data, employing both threshold judgment and machine learning models to provide risk warnings, avoiding the neglect of individual differences. This transforms posture analysis from a highly subjective, fuzzy, experience-based inference model to a highly accurate, precise matching model. In the machine learning process, it establishes interconnected interaction groups, ensuring that different interaction groups do not form parent-child nodes, achieving a balance between efficiency and risk reduction. Furthermore, it combines wearable devices to acquire exercise physiological data, generating targeted guidance plans to prevent subsequent training from exceeding the athlete's current capacity. This constructs a multi-dimensional data support system for middle and long-distance running training archives, enabling closed-loop review.

[0034] Device Examples According to an embodiment of the present invention, a running posture analysis device for middle and long distance running is provided. Figure 6 This is a schematic diagram of a running posture analysis device for medium and long distance running according to an embodiment of the present invention, as shown below. Figure 6 As shown, the running posture analysis device for middle and long distance running according to an embodiment of the present invention specifically includes: The initial processing module 60 is used to deploy cameras outside the track to capture motion video, identify and track the current athlete in the motion video, calibrate the spatial coordinates of the current athlete, and generate a motion time series of calibrated coordinate information. Specifically, it is used for: During the shooting process, metadata information including video frame rate, resolution, focal length and timestamp is synchronized. The cameras are evenly distributed in the track, and facial recognition is used to match the current athlete based on the content captured by the first camera at the starting point. The collected motion video is input into the visual recognition model as a time frame sequence. The visual recognition model extracts the skeletal joints of the human motion posture. The visual recognition model is a neural network model, and the skeletal joints include the spatial coordinate information of the head, limbs, and torso.

[0035] The visual recognition model determines whether the current frame is the athlete's foot support phase on the ground or the swing phase in the air. The Kalman filter is used to correct the athlete's spatial coordinates. If the current frame is the support phase, the captured value is given a lower weight than the Kalman filter. If the current frame is the swing phase, the captured value is given a higher weight than the Kalman filter. The corrected spatial coordinates of the pelvis were used as anchor points to smooth the motion trajectory, thus obtaining the calibrated coordinate information.

[0036] Quantitative analysis module 62 is used to perform quantitative analysis based on motion time series to obtain quantitative analysis results. Specifically, it is used for: The average cadence, average stride length, average time in the air, average time to ground contact, and angle of skeletal joints are calculated sequentially based on the motion time series. Calculate the average step frequency using Formula 1: Formula 1; in, This indicates the nth time a single foot touches the ground. This represents the total time until the nth time that one foot touches the ground; Calculate the average stride using Formula 2: Formula 2; in, This indicates the position of the i-th foot when it touches the ground. This indicates the position of the (i+1)th foot contact with the ground. Calculate the average takeoff time using Formula 3: Formula 3; in, Indicates the start frame of takeoff. This indicates the end of the takeoff frame. Indicates frame rate, Indicates the number of times the vehicle has been emptied; Calculate the average time to contact using Formula 4: Formula 4; in, This indicates the start of the ground contact frame. This indicates the end of the grounding frame. Indicates frame rate, Indicates the number of times the ground touches; Use Formula 5 to calculate the angle between bone joints. : Formula 5; in, and Let A and B represent the coordinates of joint edge point A and B, respectively. This represents the coordinates of the vertex O of the joint.

[0037] The risk warning module 64 is used to issue risk warnings based on quantitative analysis results, employing both threshold judgment and machine learning models. The machine learning model considers the balance point of interrelationships among the quantitative analysis indicators, specifically for: Based on thresholds for average cadence and average stride, it is determined whether excessive strides are being taken. Based on thresholds for the angle between skeletal joints, it is determined whether head posture, torso posture, elbow posture, knee posture, and foot posture are being assessed. All quantitative analysis results, except for the angles between skeletal joints, are input into the machine learning model, which is the XGBoost model. In the XGBoost model, average cadence and average stride are treated as a pair of related interaction groups, and average flight time and average ground contact time are treated as a pair of related interaction groups. When constructing each tree, only features within the same interaction group are allowed to interact. The interpreter outputs the internal analysis logic of the machine learning model and generates risk warnings.

[0038] Training guidance module 66 is used to acquire the athlete's current exercise physiological data, and to form structured data from the exercise physiological data, risk warnings, and quantitative analysis results. Based on this structured data, a guidance plan is generated, specifically for: Acquire exercise physiological data through wearable devices; The exercise physiological data, risk warnings, quantitative analysis results, and metadata information obtained from photography are combined into a structured data set including timestamps, and a rule engine is used to generate targeted guidance solutions.

[0039] The device further includes: The archive module 68 is used to integrate sports videos, quantitative analysis results, risk warnings, guidance plans and subsequent training improvement information, and store the integrated information in a pre-built middle and long-distance running training archive.

[0040] In summary, addressing the existing problems, this invention, a running posture analysis device for middle and long-distance running, leverages the high repeatability, strong periodicity, and stable posture characteristics of middle and long-distance running movements during the calibration process. It categorizes each movement cycle into two phases: the support phase and the swing phase, specifically improving calibration effectiveness. It transforms human movement posture information into quantifiable data, employing both threshold judgment and machine learning models to provide risk warnings, avoiding the neglect of individual differences. This transforms posture analysis from a highly subjective, fuzzy, experience-based inference model to a highly accurate, precise matching model. During machine learning, it establishes interconnected interaction groups, ensuring that different interaction groups do not form parent-child nodes, achieving a balance between efficiency and risk reduction. Furthermore, it combines wearable devices to acquire exercise physiological data, generating targeted guidance plans to prevent subsequent training from exceeding the athlete's current capacity. This constructs a multi-dimensional data support system for middle and long-distance running training archives, enabling closed-loop review.

[0041] Electronic device examples Figure 7This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device 700 may include at least one processor 710 and a memory 720. The processor 710 can execute instructions stored in the memory 720. The processor 710 is communicatively connected to the memory 720 via a data bus. In addition to the memory 720, the processor 710 can also be communicatively connected to an input device 730, an output device 740, and a communication device 750 via the data bus.

[0042] The processor 710 can be any conventional processor, such as a commercially available CPU. The processor may also include graphics processing units (GPUs), field-programmable gate arrays (FPGAs), systems on chips (SoCs), application-specific integrated circuits (ASICs), or combinations thereof.

[0043] The memory 720 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0044] In this embodiment of the present disclosure, the memory 720 stores executable instructions, and the processor 710 can read the executable instructions from the memory 720 and execute the instructions to implement all or part of the steps of any of the long-distance running posture analysis methods in the above exemplary embodiments.

[0045] Computer-readable storage medium embodiments In addition to the methods and apparatus described above, exemplary embodiments of this disclosure may also be a computer program product or a computer-readable storage medium storing the computer program product, wherein the computer program product includes computer program instructions that can be executed by a processor to implement all or part of the steps described in any of the long-distance running posture analysis methods in the exemplary embodiments described above.

[0046] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. Programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages, and scripting languages ​​(e.g., Python). The program code can be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0047] Computer-readable storage media may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media include: static random access memory (SRAM) having one or more electrically connected wires; electrically erasable programmable read-only memory (EEPROM); erasable programmable read-only memory (EPROM); programmable read-only memory (PROM); read-only memory (ROM); magnetic storage; flash memory; magnetic disk or optical disk; or any suitable combination thereof.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing running posture in middle- and long-distance running, characterized in that, include: Cameras are deployed outside the track to capture motion videos. The current athlete is identified and tracked in the motion videos. The spatial coordinates of the current athlete are calibrated, and a motion time series of calibrated coordinate information is formed. Quantitative analysis is performed on the aforementioned motion time series to obtain the quantitative analysis results; Based on the quantitative analysis results, risk warnings are issued using threshold judgment and machine learning models, respectively. The machine learning model considers the balance point of the interrelationship among the quantitative analysis indicators. The current athlete's exercise physiological data is acquired, and the exercise physiological data, the risk warning, and the quantitative analysis results are combined into structured data. A guidance plan is generated based on the structured data.

2. The method according to claim 1, characterized in that, The method further includes: The exercise videos, quantitative analysis results, risk warnings, guidance plans, and subsequent training improvement information are integrated and stored in a pre-constructed middle- and long-distance running training archive.

3. The method according to claim 1, characterized in that, The deployment of cameras outside the track to capture motion videos, and the identification and tracking of the current athlete in the motion videos, specifically includes: The system captures motion videos using cameras, and simultaneously displays metadata information including video frame rate, resolution, focal length, and timestamp during the capture process. The cameras are evenly distributed throughout the track, and facial recognition is used to match the current athlete based on the content captured by the first camera at the starting point. The collected motion video is input into a visual recognition model as a time frame sequence. The visual recognition model extracts the skeletal joints of the human motion posture, wherein the skeletal joints include the spatial coordinate information of the head, limbs and torso.

4. The method according to claim 1, characterized in that, The calibration process for the current athlete's spatial coordinates specifically includes: The visual recognition model determines whether the current frame is the athlete's foot support phase on the ground or the swing phase in the air. The Kalman filter is used to correct the athlete's spatial coordinates. If the current frame is the support phase, the captured value is assigned a lower weight than the Kalman filter. If the current frame is the swing phase, the captured value is assigned a higher weight than the Kalman filter. The corrected spatial coordinates of the pelvis were used as anchor points to smooth the motion trajectory, thus obtaining the calibrated coordinate information.

5. The method according to claim 1, characterized in that, The quantitative analysis based on the motion time series to obtain the quantitative analysis results specifically includes: Based on the aforementioned motion time series, the average cadence, average stride length, average time in the air, average time to ground contact, and angle of the skeletal joints are calculated sequentially. Calculate the average step frequency using Formula 1: Official 1; in, This indicates the nth time a single foot touches the ground. This represents the total time until the nth time that one foot touches the ground; Calculate the average stride using Formula 2: Official 2; in, This indicates the position of the i-th foot when it touches the ground. This indicates the position of the (i+1)th foot contact with the ground. The average takeoff time was calculated using Formula 3: Official 3; in, Indicates the start frame of takeoff. This indicates the end of the takeoff frame. Indicates frame rate, Indicates the number of times the vehicle has been emptied; The average ground contact time was calculated using Formula 4: Official 4; in, This indicates the start of the ground contact frame. This indicates the end of the grounding frame. Indicates frame rate, Indicates the number of times the ground touches; The angle between the bone joints is calculated using Formula 5. : Official 5; in, and Let A and B represent the coordinates of joint edge point A and B, respectively. This represents the coordinates of the vertex O of the joint.

6. The method according to claim 5, characterized in that, The risk warning based on the quantitative analysis results, using threshold judgment and machine learning model respectively, specifically includes: Based on the thresholds of the average step frequency and the average stride, it is determined whether excessive stride is being taken; based on the thresholds of the angle between the skeletal joints, it is determined whether the head posture, torso posture, elbow posture, knee posture, and foot posture are being determined. All quantitative analysis results, except for the angles between skeletal joints, are input into the machine learning model, which is an XGBoost model. In the XGBoost model, the average step frequency and the average stride are treated as an interrelated interaction group, and the average flight time and the average ground contact time are treated as an interrelated interaction group. When constructing each tree, only features within the same interaction group are allowed to interact. The interpreter outputs the internal analysis logic of the machine learning model and outputs the risk warning.

7. The method according to claim 1, characterized in that, The process of acquiring the current athlete's exercise physiological data, forming structured data from the exercise physiological data, the risk warning, and the quantitative analysis results, and generating a guidance plan based on the structured data specifically includes: The exercise physiological data is acquired through wearable devices; The exercise physiological data, the risk warning, the quantitative analysis results, and the metadata information obtained from the shooting are combined into a structured data including a timestamp, and a rule engine is used to generate a targeted guidance plan.

8. A running posture analysis device for middle and long distance running, characterized in that, include: The initial processing module is used to deploy cameras outside the track to capture motion videos, identify and track the current athlete in the motion videos, calibrate the spatial coordinates of the current athlete, and form a motion time series of calibrated coordinate information. The quantitative analysis module is used to perform quantitative analysis based on the motion time series and obtain the quantitative analysis results. The risk warning module is used to issue risk warnings based on the quantitative analysis results using threshold judgment and machine learning models, wherein the machine learning model considers the balance point of the interrelationship among the quantitative analysis indicators. The training guidance module is used to acquire the current athlete's exercise physiological data, form structured data from the exercise physiological data, the risk warning, and the quantitative analysis results, and generate a guidance plan based on the structured data.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the running posture analysis method for middle and long distance running as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an information transmission implementation program, which, when executed by a processor, implements the steps of the middle- and long-distance running posture analysis method as described in any one of claims 1 to 7.

Images