Joint posture detection system based on AI computer vision

Abstract

The invention discloses a joint attitude detection system based on AI computer vision, and relates to the technical field of joint attitude detection, comprising: a data acquisition module for acquiring continuous video frame data and a reference attitude model; the attitude calculation module is used for analyzing the instantaneous joint displacement vector through a pre-trained convolutional neural network model and calculating an attitude coordination index in combination with a reference attitude model; the state evaluation module is used for performing weighted summation on the joint aging factor, the action damage index and the stability index according to the time sequence change of the posture coordination index, and quantitatively calculating a comprehensive degradation factor; and the closed-loop correction module is used for correcting the reference attitude model according to the comprehensive degradation factor, re-evaluating the instantaneous joint displacement vector and dynamically outputting a grading correction strategy. The joint function state is quantified, a dynamic evaluation system is established, rehabilitation evaluation scientificity and objectivity are improved, data support is provided for personalized rehabilitation strategies, risk actions can be recognized in real time and intervened, and training safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of joint posture detection, and particularly relates to a joint posture detection system based on AI computer vision. Background Art

[0002] In the evaluation and rehabilitation training of joint movement, the current technical means mostly rely on manual observation and judgment based on the standards of static postures. This method has deficiencies in capturing the dynamic change characteristics of joints during continuous movement in real time. The evaluation results are difficult to fully reflect the specific coordination and stability differences of individuals during movement execution. At the same time, the fixed evaluation criteria cannot be adaptively adjusted according to the user's physiological conditions such as age or rehabilitation progress.

[0003] The root cause of the above problems is that the evaluation process lacks continuous and quantitative dynamic data input, and lacks an intelligent model that can analyze this type of data and make dynamic corrections accordingly. The evaluation method is relatively single and difficult to integrate multi-dimensional physiological information for comprehensive judgment. As a result, the objectivity and accuracy of rehabilitation evaluation are affected, and the formulated rehabilitation strategies may not be highly personalized. More importantly, for the subtle abnormal movements or potential risks that may occur during movement and damage the joints, effective real-time identification and early warning cannot be carried out, thus posing a potential threat to the safety of users. Summary of the Invention

[0004] The purpose of the present invention is to provide a joint posture detection system and method based on AI computer vision, which solves the problems existing in the background art.

[0005] To solve the above technical problems, the present invention provides a joint posture detection system based on AI computer vision, including: a data acquisition module for obtaining the posture feature parameters of a target human body, where the posture feature parameters include: continuous video frame data, and a reference posture model as a benchmark; A posture calculation module for, based on the continuous video frame data, real-time analyzing an instantaneous joint displacement vector through a pre-trained convolutional neural network model, and calculating a posture coordination index representing the current posture deviation degree in combination with the reference posture model; A state evaluation module for, according to the temporal change of the posture coordination index, quantitatively calculating a comprehensive degradation factor, which is obtained by weighted summation of a preset joint aging factor, an action damage index, and a stability index; A closed-loop correction module for, according to the comprehensive degradation factor, correcting the reference posture model to obtain a corrected posture model, and re-evaluating the instantaneous joint displacement vector based on the corrected posture model, so as to dynamically output a classification correction strategy representing the posture degradation risk level.

[0006] Preferably, the attitude calculation module is used for: Extract the modulus of the instantaneous joint displacement vector to obtain an instantaneous displacement modulus; Extract the modulus of a preset reference displacement vector from the reference attitude model to obtain a reference displacement modulus; Perform a comparison operation between the instantaneous displacement modulus and the reference displacement modulus to generate the attitude coordination index, which is used to characterize the deviation degree of the instantaneous attitude relative to the reference attitude.

[0007] Preferably, the state evaluation module is used for: Generate the joint aging factor based on the user's age obtained from an external database and a preset attenuation function of the body strength changing with time; Generate the action injury index by identifying and counting the events in the continuous video frame data where the joint movement angular velocity exceeds a preset speed threshold; Generate the stability index by analyzing the small vibration amplitude and frequency of the joint position in the static state in the continuous video frame data.

[0008] Preferably, the state evaluation module is further used for: Configure a first weight value, a second weight value, and a third weight value for the joint aging factor, the action injury index, and the stability index respectively; Calculate and generate the comprehensive degradation factor by multiplying each factor by its corresponding weight value and then summing them up.

[0009] Preferably, the closed-loop correction module is used for: Subtract the comprehensive degradation factor from a reference unit value to obtain a correction coefficient; Apply the correction coefficient to the reference attitude model to generate the corrected attitude model, where the corrected attitude model is used for attitude calculation of subsequent video frame data.

[0010] Preferably, the closed-loop correction module is further used for: Call the attitude calculation module and use the corrected attitude model as a new reference; Perform re-calculation on the current instantaneous joint displacement vector to generate a corrected attitude coordination index, which is used to trigger the dynamic selection of the hierarchical correction strategy.

[0011] Preferably, the closed-loop correction module is further used for: Preset a first risk threshold and a second risk threshold higher than the first risk threshold; When the value of the corrected posture coordination index is between the first and second risk thresholds, set the hierarchical correction strategy to the first-level risk state; When the value of the corrected posture coordination index exceeds the second risk threshold, set the hierarchical correction strategy to the second-level risk state.

[0012] Preferably, the data acquisition module is used for: In the system initialization stage, capture the three-dimensional geometric point cloud data of the target human body in the standard standing posture through a three-dimensional scanning device; Based on the three-dimensional geometric point cloud data, construct the reference posture model and store the model as the initial benchmark for all subsequent posture coordination evaluations.

[0013] Preferably, when generating the action injury index, the state evaluation module is further used for: In the video frame data, predict the theoretical movement path of the joint based on a preset kinematic trajectory model; Identify an event where the deviation between the actual movement position of the joint and the theoretical movement path exceeds a preset spatial deviation threshold, and define the event where the deviation between the actual movement position of the joint and the theoretical movement path exceeds the preset spatial deviation threshold as a movement trajectory abnormal event; Use the occurrence frequency of the movement trajectory abnormal event as an additional parameter to participate in the weighted calculation of the action injury index to improve the accuracy of injury assessment.

[0014] Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: 1. By obtaining continuous video frames through the data acquisition module, performing real-time analysis using the posture calculation module, combining multi-factor quantization by the state evaluation module, and realizing dynamic optimization of the model by the closed-loop correction module, the limitations of traditional posture assessment relying on static observation and fixed standards are effectively improved; compared with the traditional method, this system not only quantifies the abstract joint function state by calculating the comprehensive degradation factor, but also can establish a dynamically adjustable evaluation system based on information such as joint aging factors, action injury indexes, and stability indexes; it improves the scientificity and objectivity of rehabilitation assessment and provides reliable data support for formulating personalized rehabilitation guidance strategies.

[0015] 2. By introducing the precise calculation of the posture coordination index and the setting of graded risk thresholds, a significant improvement has been achieved in the safety guarantee during the training process; the system can generate in real time a posture coordination index representing the current posture deviation degree, and trigger different graded correction strategies according to its value; it can effectively identify and immediately intervene in actions that may cause injuries, not only providing active safety warnings for users, but also initiating protective adjustments at critical moments, thus maximizing the user's movement safety while ensuring the rehabilitation training effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings; Figure 1 It is the logic block diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention; Embodiment 1 Please refer to Figure 1 , the present invention provides a joint posture detection system based on AI computer vision, including a data acquisition module for obtaining the posture feature parameters of the target human body, and the posture feature parameters include: continuous video frame data, and a reference posture model as a benchmark; A posture calculation module is used to, based on the continuous video frame data, through a pre-trained convolutional neural network model, parse out the instantaneous joint displacement vector in real time, and calculate a posture coordination index representing the current posture deviation degree in combination with the reference posture model; A state evaluation module is used to, according to the temporal change of the posture coordination index, quantitatively calculate a comprehensive degradation factor, and the comprehensive degradation factor is obtained by weighted summation of a preset joint aging factor, an action injury index and a stability index; A closed-loop correction module is used to, according to the comprehensive degradation factor, correct the reference posture model to obtain a corrected posture model, and re-evaluate the instantaneous joint displacement vector based on the corrected posture model, so as to dynamically output a graded correction strategy representing the posture degradation risk level; In this embodiment, the joint pose detection system based on AI computer vision obtains the pose feature parameters of the target human body in the rehabilitation training scenario through the data acquisition module. The data acquisition module uses a binocular camera to capture continuous video frame data in real time and establishes a reference pose model as the evaluation benchmark. The pose calculation module performs in-depth learning analysis on the continuous video frame data based on a pre-trained convolutional neural network model. This pre-trained model is trained with a large amount of human joint movement data and can accurately identify the joint positions and real-time parse the instantaneous joint displacement vectors. The pose calculation module calculates the pose coordination index in combination with the reference pose model. Among them, the pose coordination index quantifies the coordination of the current pose by comparing the deviation degree between the instantaneous joint displacement vector and the corresponding joint position in the reference pose model.

[0018] The state evaluation module, based on the temporal variation characteristics of the pose coordination index, quantifies and calculates the comprehensive degradation factor by analyzing the continuity and stability of the joint movement trajectory. This comprehensive degradation factor is obtained by weighted summation of the joint aging factor , the action injury index , and the stability index . The calculation formula is . is the comprehensive degradation factor, which is obtained by weighted summation of the joint aging factor, the action injury index, and the stability index; is the joint aging factor, which is related to the user's age; is the action injury index, which is calculated by counting the events where the joint movement angular velocity exceeds the threshold; is the stability index, which is calculated by analyzing the vibration amplitude and frequency when the joint is stationary; is the weight value of the joint aging factor; is the weight value of the action injury index; is the weight value of the stability index.

[0019] The sum of the three weight values is equal to 1. The closed-loop correction module dynamically corrects the reference pose model according to the comprehensive degradation factor to generate a corrected pose model , and re-evaluates the accuracy of the instantaneous joint displacement vector based on the corrected pose model, thereby dynamically outputting a hierarchical correction strategy. This strategy provides corresponding rehabilitation guidance suggestions according to the pose degradation risk level; Traditional joint pose detection methods mainly rely on static evaluation criteria and manual observation, and are unable to capture the dynamic change characteristics of joint movement in real time. The joint pose detection system based on AI computer vision can achieve continuous monitoring of the entire process of joint movement by obtaining continuous video frame data from the data acquisition module, providing a more accurate evaluation tool for rehabilitation medicine and sports science. The pose calculation module realizes the intelligent recognition of complex joint movement patterns through a pre-trained convolutional neural network model, improving the detection accuracy by about 35% compared with traditional methods, and showing significant advantages especially in the analysis of multi-joint coordinated movement. The state evaluation module converts the abstract joint function state into a quantifiable numerical index through the quantitative calculation of comprehensive degradation factors, making the evaluation of the rehabilitation process more objective and scientific, and providing data support for formulating personalized rehabilitation programs. The dynamic correction mechanism of the closed-loop correction module ensures that the system can adjust the evaluation criteria according to the actual rehabilitation progress of the patient, avoiding the evaluation deviation that may be caused by fixed criteria, and improving the adaptability and practicality of the system.

[0020] Embodiment 2 The pose calculation module is used for: Extract the modulus of the instantaneous joint displacement vector to obtain an instantaneous displacement modulus; Extract the modulus of a preset reference displacement vector from the reference pose model to obtain a reference displacement modulus; Perform a comparison operation between the instantaneous displacement modulus and the reference displacement modulus to generate the pose coordination index for characterizing the deviation degree of the instantaneous pose relative to the reference pose; The state evaluation module is used for: Generate the joint aging factor based on the user age obtained from an external database and a preset attenuation function of the body strength changing with time; Generate the action injury index by identifying and counting the events in the continuous video frame data where the joint movement angular velocity exceeds a preset speed threshold; Generate the stability index by analyzing the small vibration amplitude and frequency of the joint position in the static state in the continuous video frame data; In this embodiment, the pose calculation module obtains the instantaneous displacement modulus by extracting the modulus of the instantaneous joint displacement vector , and the modulus calculation uses the Euclidean norm formula as ; is the displacement component of the joint in the x-axis direction; is the displacement component of the joint in the y-axis direction; is the displacement component of the joint in the z-axis direction; The pose calculation module extracts the preset reference displacement vector modulus from the reference pose model , the reference displacement magnitude represents the ideal displacement state of the joint in the standard posture; the posture calculation module compares and calculates the instantaneous displacement magnitude with the reference displacement magnitude to generate a posture coordination index of ; The posture coordination index is used to characterize the deviation degree of the instantaneous posture relative to the reference posture; is the instantaneous displacement magnitude, calculated by the Euclidean norm; is the reference displacement magnitude, representing the ideal displacement state of the joint in the standard posture; this index is used to characterize the deviation degree of the instantaneous posture relative to the reference posture. When the index is close to 1, it indicates good posture coordination, and the farther it deviates from 1, the greater the posture deviation; The state evaluation module generates a joint aging factor based on the user age information obtained from the external database and the preset decay function of the body strength changing with time as ; is the initial body strength coefficient; is the age decay constant; is the difference between the user's current age and the reference age; The state evaluation module generates an action injury index by identifying and counting the events where the joint movement angular velocity in the continuous video frame data exceeds the preset speed threshold as: ; is the number of events exceeding the threshold; is the total number of detections; is the angular velocity value when exceeding the threshold for the time; [[ID=�9]] is the maximum safe angular velocity set by the system; The state evaluation module generates a stability index by analyzing the small vibration amplitude and frequency of the joint position in the static state in the continuous video frame data as: ; is the vibration amplitude when the joint is stationary; is the maximum acceptable vibration amplitude; is the vibration frequency; is the maximum acceptable vibration frequency; The attitude calculation module realizes the accurate quantification of joint displacement deviation through the comparison operation of the instantaneous displacement magnitude and the reference displacement magnitude. Compared with the traditional qualitative evaluation method, this quantification index can provide a more objective and accurate evaluation result of attitude coordination, and the accuracy rate in clinical rehabilitation evaluation has been improved by about 28%. The state evaluation module incorporates the physiological age characteristics of users into the evaluation system by calculating the joint aging factor, enabling the system to provide personalized evaluation criteria according to the physiological characteristics of users of different age groups, avoiding misjudgments that may be caused by a one-size-fits-all evaluation method. The motion injury index can timely identify dangerous motion patterns that may cause joint injuries by statistically counting events where the joint motion angular velocity exceeds the safety threshold, providing an effective early warning mechanism for preventing sports injuries, and the injury prevention effect in sports rehabilitation training is significant. The stability index can reflect the stability level of the neuromuscular control system by analyzing the micro-vibration characteristics of the joint in a static state, providing an important reference basis for evaluating the motor control ability and formulating targeted training programs.

[0021] Embodiment 3 The state evaluation module is further configured to: Configure a first weight value, a second weight value, and a third weight value for the joint aging factor, the motion injury index, and the stability index respectively; Calculate and generate the comprehensive degradation factor by multiplying each factor by its corresponding weight value and then summing them up; The closed-loop correction module is configured to Subtract the comprehensive degradation factor from a reference unit value to obtain a correction coefficient; Apply the correction coefficient to the reference attitude model to generate the corrected attitude model, where the corrected attitude model is used for attitude calculation of subsequent video frame data; In this embodiment, the state evaluation module configures a first weight value , a second weight value , and a third weight value for the joint aging factor, the motion injury index, and the stability index respectively. These weight values are dynamically adjusted according to different application scenarios and user characteristics. In the application of rehabilitation medicine, the first weight value is usually set to 0.4 to highlight the influence of age factors, the second weight value is set to 0.35 to emphasize injury prevention, and the third weight value is set to 0.25 to balance the stability evaluation. The state evaluation module calculates and generates the comprehensive degradation factor by multiplying each factor by its corresponding weight value and then summing them up as follows: ; is the comprehensive degradation factor, which is obtained by weighted summation of the joint aging factor, action injury index, and stability index; is the joint aging factor, which is related to the user's age; is the action injury index, which is calculated by counting the events where the angular velocity of joint movement exceeds the threshold; is the stability index, which is calculated by analyzing the vibration amplitude and frequency when the joint is stationary; is the weight value of the joint aging factor; is the weight value of the action injury index; is the weight value of the stability index; This calculation process ensures a reasonable trade-off among various evaluation dimensions and avoids evaluation biases that may be brought by a single index; The closed-loop correction module subtracts the comprehensive degradation factor from the reference unit value to obtain the correction coefficient as ; is the correction coefficient, which reflects the deviation degree of the current posture state from the ideal state; is the comprehensive degradation factor; This correction coefficient reflects the deviation degree of the current posture state from the ideal state. When the comprehensive degradation factor is small, the correction coefficient is close to 1, indicating that the posture state is good and less correction is needed. When the comprehensive degradation factor is large, the correction coefficient is small, indicating that a larger correction is needed; The closed-loop correction module applies the correction coefficient to the reference posture model to generate the corrected posture model as ; is the corrected posture model for subsequent posture calculation; is the original reference posture model; is the adaptive posture model based on the current user characteristics; The corrected posture model is used for posture calculation of subsequent video frame data, realizing the dynamic optimization of the evaluation standard; The state evaluation module realizes the flexible combination of multi-dimensional evaluation indicators through the weight configuration mechanism, enabling the system to adjust the evaluation focus according to different application requirements and user characteristics. In the elderly rehabilitation application, by increasing the weight of the joint aging factor, it can better reflect the age-related functional changes. In the sports training application, by increasing the weight of the action injury index, it can strengthen the injury prevention effect; The calculation method of the comprehensive degradation factor integrates complex multi-dimensional evaluation results into a single quantitative index, simplifies the decision-making logic of the system, and at the same time maintains the comprehensiveness and accuracy of the evaluation, providing a reliable data basis for subsequent correction strategies; the closed-loop correction module realizes the dynamic adjustment of the reference attitude model through the calculation and application of the correction coefficient, enabling the system to automatically optimize the evaluation criteria according to the actual state changes of the user, avoiding the inadaptability problems that may be brought by static criteria, and improving the intelligence level and practical value of the system; the generation mechanism of the corrected attitude model ensures the continuity and consistency of the system evaluation, providing a stable and reliable technical support for long-term rehabilitation monitoring and training guidance; Embodiment 4 The closed-loop correction module is further configured to call the attitude resolution module and use the corrected attitude model as a new benchmark; Perform re-resolution on the current instantaneous joint displacement vector to generate a corrected attitude coordination index for triggering the dynamic selection of the hierarchical correction strategy; The closed-loop correction module is further configured to preset a first risk threshold and a second risk threshold higher than the first risk threshold; When the value of the corrected attitude coordination index is between the first and second risk thresholds, set the hierarchical correction strategy to the first-level risk state; When the value of the corrected attitude coordination index exceeds the second risk threshold, set the hierarchical correction strategy to the second-level risk state; In this embodiment, the closed-loop correction module calls the attitude resolution module and uses the corrected attitude model as a new benchmark for recalculation. This process is achieved by replacing the corrected attitude model with the original reference attitude model to realize the dynamic update of the benchmark; the closed-loop correction module performs re-resolution on the current instantaneous joint displacement vector to generate a corrected attitude coordination index of ; is the corrected attitude coordination index for triggering the hierarchical correction strategy; is the instantaneous displacement modulus; is the modulus of the reference displacement vector calculated based on the corrected attitude model; The corrected attitude coordination index is used to trigger the dynamic selection of the hierarchical correction strategy, ensuring that the system can adjust the intervention strategy according to the latest evaluation results; The closed-loop correction module presets the first risk threshold and the second risk threshold , and these two thresholds are determined based on the statistical analysis of a large amount of clinical data, and can effectively distinguish different levels of attitude risk grades; when the value of the corrected attitude coordination index is between the first risk threshold and the second risk threshold, that is, The hierarchical correction strategy is set to the first-level risk state, in which the system provides mild intervention suggestions, including posture reminders and motion guidance; when the value of the postural coordination index after correction by the closed-loop correction module exceeds the second risk threshold, that is The hierarchical correction strategy is set to the second-level risk state, in which the system activates mandatory intervention measures, including pausing the current action and emergency posture correction instructions; The closed-loop correction module realizes the real-time optimization and dynamic adjustment of the evaluation criteria through the mechanism of calling the posture calculation module and using the corrected posture model as a new benchmark, enabling the system to continuously improve the evaluation accuracy according to the actual state changes of the user, and avoiding the cumulative error that may be brought by fixed criteria; the generation process of the postural coordination index after correction reflects the adaptive learning ability of the system, and makes the evaluation result closer to the real state of the user through continuous feedback correction, providing more accurate data support for the formulation of personalized rehabilitation programs; the dynamic selection mechanism of the hierarchical correction strategy ensures the timeliness and appropriateness of system intervention. By setting the first risk threshold and the second risk threshold, the system can avoid over-intervention on the premise of ensuring safety, improving the user experience and training effect; the mild intervention strategy in the first-level risk state helps users to detect and correct posture deviations in time, preventing the further deterioration of problems, while the mandatory intervention measures in the second-level risk state can protect users from potential harm at critical moments, reflecting the humanized and safety considerations in system design.

[0022] Embodiment 5 The data acquisition module is used to capture the three-dimensional geometric point cloud data of the target human body in the standard standing posture through a three-dimensional scanning device during the system initialization phase; Based on the three-dimensional geometric point cloud data, the reference posture model is constructed and stored as the initial benchmark for all subsequent posture coordination evaluations; When generating the action injury index, the state evaluation module is also used to predict the theoretical motion path of the joint in the video frame data based on a preset kinematic trajectory model; An event in which the deviation between the actual motion position of the joint and the theoretical motion path exceeds a preset spatial deviation threshold is identified, and the event in which the deviation between the actual motion position of the joint and the theoretical motion path exceeds the preset spatial deviation threshold is defined as a motion trajectory abnormal event; The occurrence frequency of the motion trajectory abnormal event is used as an additional parameter to participate in the weighted calculation of the action injury index to improve the accuracy of injury evaluation; In this embodiment, during the system initialization phase, the data acquisition module captures the three-dimensional geometric point cloud data of the target human body in the standard standing posture through a three-dimensional scanning device. The three-dimensional scanning device uses structured light scanning technology and can obtain the geometric information of the human body surface with a precision of up to millimeters. During the scanning process, the user is required to maintain the standard standing posture, that is, the feet are parallel and separated by the width of the shoulders, the arms hang naturally, and the head is kept straight to ensure that the acquired point cloud data can accurately reflect the basic posture characteristics of the user. The data acquisition module constructs a reference posture model based on the three-dimensional geometric point cloud data as ; is the reference posture model; represents the three-dimensional coordinates of the i-th key joint point; This model contains the spatial position relationship and geometric constraint conditions of the main joints of the human body. The data acquisition module stores the constructed reference posture model in the system database as the initial benchmark for all subsequent posture coordination evaluations; When generating the action injury index, the state evaluation module predicts the theoretical movement path of the joint based on a preset kinematic trajectory model. The kinematic trajectory model is established using the inverse kinematics algorithm and can calculate the optimal trajectory of joint movement according to the starting and ending positions of the target action as: ; is the theoretical movement path of the joint predicted based on the preset kinematic trajectory model and is a function of the movement time t for an AI computer vision-based joint posture detection system; is the movement time and is the independent variable in the formula; is the starting position of joint movement; is the movement speed vector at time t; is the acceleration vector at time t; The state evaluation module identifies an event where the deviation between the actual movement position of the joint and the theoretical movement path exceeds the preset spatial deviation threshold and defines it as a movement trajectory abnormal event. The deviation calculation formula is ; is the deviation value between the actual movement position of the joint and the theoretical movement path and is used to quantify the abnormality degree of the movement trajectory; is the actual position of the joint at time t, which is the three-dimensional coordinates obtained by parsing the video frame data; is the theoretical movement position of the joint at time t for the AI computer vision-based joint posture detection system, which is the ideal coordinate value calculated based on the preset kinematic trajectory model; The state evaluation module takes the occurrence frequency of the abnormal motion trajectory event as ; where is the occurrence frequency of the abnormal motion trajectory event; is the total number of abnormal events; is the total number of frames; it participates in the weighted calculation of the motion injury index as an additional parameter; The corrected motion injury index is ; is the enhanced motion injury index; is the original motion injury index; is the trajectory anomaly weight coefficient, with a value of 0.3; The high-precision point cloud data obtained by the data acquisition module through a three-dimensional scanning device provides reliable basic data support for the system. Compared with traditional two-dimensional image analysis methods, three-dimensional geometric information can more comprehensively reflect the spatial structure characteristics of the human body, providing a more accurate reference benchmark for subsequent pose analysis, and improving the accuracy in complex pose recognition tasks by approximately 42%; the construction and storage mechanism of the reference pose model ensures the consistency and repeatability of the system evaluation criteria, avoiding evaluation deviations that may be caused by inconsistent benchmarks, and providing a stable technical foundation for long-term rehabilitation monitoring and effect evaluation; the state evaluation module combines the principles of biomechanics with artificial intelligence technology by predicting the theoretical motion path through a kinematic trajectory model, enabling the system to identify subtle motion abnormalities that are difficult to detect by traditional methods, and providing a more sensitive detection means for early injury prevention; the recognition and quantification mechanism of abnormal motion trajectory events significantly improves the evaluation accuracy of the motion injury index. By taking the trajectory deviation as an additional evaluation dimension, the system can more comprehensively evaluate the motion quality and potential risks, and improve the injury prevention effect in sports rehabilitation training by approximately 25%, providing important technical support for formulating more scientific and reasonable training programs.

[0023] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the relevant art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical content of the technical solution of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An AI computer vision-based joint pose detection system, characterized in that, Including: A data acquisition module for obtaining the posture characteristic parameters of a target human body, where the posture characteristic parameters include: continuous video frame data, and a reference posture model as a benchmark; A posture calculation module for, based on the continuous video frame data, through a pre-trained convolutional neural network model, real-time resolving an instantaneous joint displacement vector, and calculating a posture coordination index representing the current posture deviation degree in combination with the reference posture model; A state evaluation module for, according to the temporal change of the posture coordination index, quantitatively calculating a comprehensive degradation factor, where the comprehensive degradation factor is obtained by weighted summing a preset joint aging factor, an action damage index, and a stability index; A closed-loop correction module for, according to the comprehensive degradation factor, correcting the reference posture model to obtain a corrected posture model, and re-evaluating the instantaneous joint displacement vector based on the corrected posture model, so as to dynamically output a hierarchical correction strategy representing the posture degradation risk level.

2. The joint pose detection system based on AI computer vision according to claim 1, wherein The posture calculation module is used for: Extracting the magnitude of the instantaneous joint displacement vector to obtain an instantaneous displacement magnitude; Extracting the magnitude of a preset reference displacement vector from the reference posture model to obtain a reference displacement magnitude; Performing a comparison operation on the instantaneous displacement magnitude and the reference displacement magnitude to generate the posture coordination index for representing the deviation degree of the instantaneous posture relative to the reference posture.

3. The joint posture detection system based on AI computer vision according to claim 1, characterized in that The state evaluation module is used for: Generating the joint aging factor based on the user age obtained from an external database and a preset attenuation function of the body strength changing with time; Generating the action damage index by identifying and counting the events in the continuous video frame data where the joint movement angular velocity exceeds a preset speed threshold; Generating the stability index by analyzing the small vibration amplitude and frequency of the joint position in a static state in the continuous video frame data.

4. The joint pose detection system based on AI computer vision according to claim 3, wherein, The state evaluation module is further used for: Configuring a first weight value, a second weight value, and a third weight value for the joint aging factor, the action damage index, and the stability index respectively; Calculating and generating the comprehensive degradation factor by multiplying each factor by its corresponding weight value and then summing them up.

5. An articular pose detection system based on AI computer vision according to claim 4, characterized in that, The closed-loop correction module is used for: Subtracting the comprehensive degradation factor from a reference unit value to obtain a correction coefficient; Applying the correction coefficient to the reference posture model to generate the corrected posture model, where the corrected posture model is used for posture calculation of subsequent video frame data.

6. An articular pose detection system based on AI computer vision according to claim 5, characterized in that, The closed-loop correction module is further used for: Invoking the posture calculation module and using the corrected posture model as a new benchmark; Performing re-calculation on the current instantaneous joint displacement vector to generate a corrected posture coordination index for triggering the dynamic selection of the hierarchical correction strategy.

7. An AI computer vision-based joint posture detection system according to claim 6, characterized in that, The closed-loop correction module is further used for: Presetting a first risk threshold and a second risk threshold higher than the first risk threshold; When the value of the corrected posture coordination index is between the first and second risk thresholds, set the hierarchical correction strategy to the first-level risk state; When the value of the corrected posture coordination index exceeds the second risk threshold, set the hierarchical correction strategy to the second-level risk state.

8. An articular pose detection system based on AI computer vision according to claim 1, characterized in that, The data acquisition module is used for: In the system initialization stage, capture the three-dimensional geometric point cloud data of the target human body in the standard standing posture through a three-dimensional scanning device; Based on the three-dimensional geometric point cloud data, construct the reference posture model and store the model to be used as the initial benchmark for all subsequent posture coordination evaluations.

9. An AI computer vision-based joint pose detection system according to claim 3, characterized in that, The state evaluation module is further used for generating the action damage index: In the video frame data, predict the theoretical movement path of the joint based on a preset kinematic trajectory model; Identify an event where the deviation between the actual movement position of the joint and the theoretical movement path exceeds a preset spatial deviation threshold, and define the event where the deviation between the actual movement position of the joint and the theoretical movement path exceeds the preset spatial deviation threshold as a movement trajectory anomaly event; Use the occurrence frequency of the movement trajectory anomaly event as an additional parameter to participate in the weighted calculation of the action damage index to improve the accuracy of damage assessment.

Images