Health assessment method and system based on 3D human body posture image model

Abstract

The invention relates to the technical field of human body health assessment, in particular to a health assessment method and system based on a 3D human body posture image model. The method comprises the following steps: acquiring human body annular scanning data; extracting a dynamic posture sequence of a vertical position, a forward bending position and a lateral bending position of the human body annular scanning data, and performing phase shift fringe projection on the human body annular scanning data according to the dynamic posture sequence to obtain human body three-dimensional point cloud data; a sagittal plane, a coronal plane and a cross section of the human body three-dimensional point cloud data are analyzed for orthogonal projection, and a human body multi-angle projection image group is generated; and identifying skeleton contour features of the human body multi-angle projection image group so as to establish a dynamic marking system of the anatomical mark points. Through high-precision dynamic acquisition and multi-dimensional space analysis, precision, standardization and individuation of 3D human body posture health assessment are realized, and the accuracy and comprehensiveness of human body posture health assessment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of human health assessment, and in particular, to a health assessment method and system based on a 3D human body pose image model. Background Art

[0002] Early health assessments mainly relied on doctors' experience and simple two-dimensional image analysis, which were difficult to comprehensively and accurately reflect the human body pose and its health status. With the progress of deep learning and sensor technology, three-dimensional human body pose capture technology has been continuously matured, evolving from the initial marker-based motion capture system to the markerless pose recognition technology using multi-camera and multi-sensor fusion, greatly improving the accuracy and real-time performance of data acquisition. In recent years, health assessment methods based on 3D pose images, combined with machine learning models, have achieved intelligent recognition of problems such as movement dysfunction, skeletal abnormalities, and poor postures by analyzing the dynamic changes of key points of the human skeleton. However, the existing technologies for studying the healthy sleeping postures of the human body mainly analyze the pressure conditions of various parts of the human body in the lying posture through pressure sensing. This method has many limitations. On the one hand, relying solely on pressure data cannot comprehensively obtain the pose information of the human body, and it is impossible to accurately know the key data such as the angles and spatial position relationships of various parts of the human body, resulting in inaccurate and incomplete assessment of the human body pose. On the other hand, the pressure sensing technology is greatly affected by environmental factors. For example, the material and thickness of the mattress will interfere with the accuracy of the pressure data, thus affecting the final health assessment result. Summary of the Invention

[0003] Based on this, it is necessary to provide a health assessment method and system based on a 3D human body pose image model to solve at least one of the above technical problems.

[0004] To achieve the above object, a health assessment method based on a 3D human body pose image model, the method includes the following steps: Step S1: Obtain human body circular scan data; extract the dynamic pose sequences of the upright position, forward flexion position, and lateral bending position of the human body circular scan data, and perform phase-offset fringe projection on the human body circular scan data according to the dynamic pose sequences to obtain human body three-dimensional point cloud data; Step S2: Analyze the sagittal plane, coronal plane, and transverse plane of the human body three-dimensional point cloud data for orthogonal projection to generate a group of human body multi-angle projection images; identify the bone contour features of the human body multi-angle projection image group to establish a dynamic marking system for anatomical landmark points; use the dynamic marking system to calibrate the spatial pose of the bone contour features to generate spatial pose calibration parameters; Step S3: Input the spatial attitude calibration parameters into a preset healthy attitude evaluation database for multi-dimensional comparative analysis, and calculate the deviation index between the parts in the three-dimensional point cloud data of the human body and the standard value according to the analysis results, so as to construct a biomechanical load distribution map and a joint range of motion heat map; Step S4: Construct a 3D human body attitude image model based on the multi-angle projection image group of the human body; conduct attitude health evaluation on the 3D human body attitude image model respectively through the biomechanical load distribution map and the joint range of motion heat map, and perform personalized attitude correction according to the evaluation results, so as to generate a personalized attitude correction report.

[0005] The present invention realizes high-precision three-dimensional reconstruction of the human body attitude in multiple motion states through the acquisition of dynamic attitude sequences in the upright position, forward flexion position and lateral bending position, and combines the phase-shifted fringe projection technology, improving the integrity and dynamic expressiveness of the data. By using multi-angle orthogonal projections of the sagittal plane, coronal plane and transverse plane, automatically identify the bone contour features and establish a dynamic marking system, so as to calibrate the spatial attitude of the attitude, ensuring the accuracy and repeatability of subsequent analysis. By comparing the calibration parameters with the healthy attitude database in multiple dimensions, calculate the deviation index to accurately describe the difference between the individual and the standard; and further generate a biomechanical load distribution map and a joint range of motion heat map to realize multi-dimensional health evaluation. Based on the multi-angle projection maps, construct an intuitive 3D attitude image model, and combine the superimposed analysis of the heat map and the load map to enhance the result visualization and interactivity, facilitating users and doctors to intuitively understand the attitude problems. According to the individual's load distribution and range of motion analysis results, formulate a personalized correction strategy, and finally generate a customized attitude correction report, which helps to improve the accuracy and effectiveness of attitude intervention. Therefore, through high-precision dynamic acquisition and multi-dimensional spatial analysis, the present invention realizes the precision, standardization and personalization of 3D human body attitude health evaluation, improving the accuracy and comprehensiveness of human body attitude health evaluation.

[0006] Preferably, step S1 includes the following steps: Step S11: Use a scanning device to perform a 360° circular scan on the human body to obtain human circular scan data; Step S12: Extract the scan sequence of the human circular scan data, and perform inter-frame synchronous registration on the human circular scan data according to the scan sequence to generate synchronously corrected human scan frame data; Step S13: Extract the dynamic skeleton points of the human scan frame data; perform attitude annotation on the dynamic skeleton points to generate dynamic attitude sequences in the upright position, forward flexion position and lateral bending position; Step S15: Decompose the spatial attitude parameters of the dynamic attitude sequence to obtain human pose distribution control data; perform fringe phase encoding and multi-phase offset fringe map generation on the human scan frame data to obtain a multi-phase offset encoded map set; Step S16: Unwrap the multi-phase offset encoded atlas through the human body pose distribution control data to generate a dense phase map; perform triangulation and depth inversion calculations on the dense phase map to generate high-precision three-dimensional point cloud data of the human body.

[0007] In the present invention, by using a scanning device to complete a 360° circular scan of the human body, the original scan data of the entire surface surrounding the human body can be obtained, improving data integrity and providing comprehensive information support for subsequent modeling and analysis. Through the frame-by-frame synchronous registration method, the scan sequence is uniformly corrected, effectively solving the spatio-temporal error and dynamic misalignment problems caused by human movement in multi-frame data, and enhancing the stability and accuracy of point cloud reconstruction. The extraction of dynamic skeleton points and pose annotation enable the scan data to contain not only surface morphology but also dynamic pose information (upright position, forward flexion position, lateral bending position, etc.), providing a structured basis for subsequent individual pose assessment and motion function analysis. Transforming the dynamic pose sequence into pose distribution control data realizes spatial decoupling control of the human body in different postures; combining fringe phase encoding and multi-phase offset processing enhances the accuracy and anti-interference ability of depth acquisition. With the help of techniques such as phase unwrapping, dense mapping, and depth inversion, high-density and low-error three-dimensional point cloud data with sub-millimeter accuracy is generated from the phase atlas, effectively supporting subsequent spatial analysis and visualization modeling. The overall process supports dynamic scanning and motion recognition scenarios, and is applicable not only to static modeling but also to scenarios with high requirements for dynamic characteristics such as rehabilitation assessment and motion analysis, expanding the application boundary.

[0008] Preferably, in step S13, the pose annotation of the dynamic skeleton points includes: Collect the three-dimensional coordinates of the dynamic skeleton points through an optical motion capture system to obtain the three-dimensional coordinates of the skeleton points, where the sampling frequency is set to ≥30 frames / second, the spatial accuracy is ±2 mm, the coordinate unit is millimeter, and the right-handed coordinate system is referenced; Define the recognition standard for the upright position: the overall inclination angle of the spine ≤5°, the height difference between the left and right shoulders ≤10 mm, and the knee flexion angle ≤10°; Define the recognition standard for the forward flexion position: the overall anterior inclination angle of the spine ≥30° and ≤90°, the curvature radius from the neck to the lumbar spine ≤400 mm, and the hip flexion angle ≥45°; Define the recognition standard for the lateral bending position: the lateral bending angle of the spine ≥20° and ≤60°, the height difference between the left and right hips ≥15 mm, and the horizontal distance of the shoulder-hip connection deviating from the central axis ≥30 mm; Based on the defined recognition standards for the upright position, forward flexion position, and lateral bending position, perform pose classification and annotation on the three-dimensional coordinates of the skeleton points, thereby obtaining the dynamic pose sequences of the upright position, forward flexion position, and lateral bending position.

[0009] By using an optical motion capture system, under the constraints of a sampling frequency ≥ 30 frames per second and a spatial accuracy of ±2 mm, the present invention ensures that the coordinates of the human body skeleton points collected have high temporal resolution and high spatial accuracy, meeting the analysis requirements for complex dynamic posture changes. By quantitatively defining the upright position, forward flexion position, and lateral bending position (such as the overall inclination angle of the spine, shoulder height difference, joint flexion angle, etc.), an objective, repeatable, and programmable posture recognition standard system is formed, avoiding the subjectivity and inconsistency of traditional manual evaluations. Based on the matching algorithm between the three-dimensional coordinates of the skeleton points and the set recognition standards, each frame in the scanned data can be automatically classified for posture, generating high-quality posture sequence data, providing an accurate annotation basis for subsequent three-dimensional modeling and health assessment. Using recognition standards that include various geometric indicators (inclination angle, radius of curvature, joint angle, horizontal offset, etc.), the system can accurately identify complex non-standard postures such as forward flexion and lateral bending, improving the recognition sensitivity for abnormal postures and potential health risks. The three-dimensional coordinates of the skeleton points are based on the right-hand coordinate system, with the unit being millimeters, facilitating compatibility with existing three-dimensional reconstruction algorithms, anatomical databases, and posture simulation models, and supporting cross-platform and cross-system data interaction and processing.

[0010] Preferably, the orthogonal projection of the three-dimensional point cloud data of the human body in the sagittal plane, coronal plane, and transverse plane in step S2 includes: Normalize the center of gravity of the spatial coordinates of the three-dimensional point cloud data of the human body to generate standardized three-dimensional posture point cloud data; Extract the local principal axis of the standardized three-dimensional posture point cloud data, and use the local principal axis to calibrate the body axis vector field of the three-dimensional posture point cloud data to generate structure-aligned vector calibration data; Perform depth mapping rearrangement on the structure-aligned vector calibration data in the sagittal plane direction to generate a sagittal plane orthogonal projection map; Perform depth value aggregation and compression on the structure-aligned vector calibration data in the coronal plane direction to generate a coronal plane orthogonal projection map; Perform hierarchical slicing and resampling on the structure-aligned vector calibration data in the transverse plane direction to generate a transverse plane orthogonal projection map; Perform image group registration and angle label binding on the sagittal plane, coronal plane, and transverse plane orthogonal projection maps to generate a multi-angle projection image group of the human body.

[0011] Through spatial coordinate centroid normalization processing, the present invention effectively eliminates the influence of individual height and body shape differences on the result of pose analysis, enables the three-dimensional point cloud data to have a consistent reference framework, and improves the alignment accuracy and analysis generality between subsequent diverse body posture data. By adopting the strategy of local principal axis extraction and body axis vector field calibration, vector field calibration data conforming to the laws of human biological structure can be constructed, so that each point in the point cloud has a clear anatomical direction meaning, and the structural constraint of feature recognition and pose comparison is strengthened. Through depth mapping rearrangement, aggregation compression and hierarchical slice resampling of the sagittal plane, coronal plane and transverse plane, the spatial depth and structural features in the three-dimensional pose are effectively retained, realizing the accurate visual expression of the human body pose in the two-dimensional plane. Through image group registration and angle label binding, the output multi-angle projection maps have consistent spatial semantics and data labels, which are convenient to be input into a convolutional neural network, a pose recognition model or a biomechanics simulation platform to realize efficient automatic pose recognition and evaluation. The orthogonal projection process divides the original three-dimensional point cloud into multi-dimensional structural views organized by anatomical planes, so that abnormal pose features such as scoliosis, forward tilt or vertebral rotation can significantly enhance the contrast on a specific projection plane, improving the sensitivity and accuracy of structural abnormality recognition.

[0012] Preferably, identifying the skeletal contour features of the human body multi-angle projection image group in step S2 to establish a dynamic marking system for anatomical landmark points includes: Identifying the main skeletal contour of the human body multi-angle projection image group; Analyzing the connected domains of the human body multi-angle projection image group according to the main skeletal contour to obtain a primary skeleton distribution map; Performing key morphological turning point detection on the primary skeleton distribution map to generate candidate anatomical structure corner point data, and performing spatial consistency test based on depth prior on the candidate anatomical structure corner point data to generate a high-confidence anatomical landmark point set; Performing temporal pose dynamic matching and trajectory interpolation on the high-confidence anatomical landmark point set to generate dynamic annotation sequence data; Performing pose angle change modeling and structural stability evaluation on the high-confidence anatomical landmark point set through the dynamic annotation sequence data to generate a dynamic marking system for anatomical landmark points.

[0013] The present invention automatically identifies the main bone contours in a multi-angle projection image group, and generates a primary skeleton distribution map by combining connected component analysis, without relying on traditional manual annotation, significantly improving the automation level and efficiency of pose structure recognition, and being applicable to the processing and analysis of large-scale pose data. By adopting a key morphological turning point detection + depth prior consistency verification mechanism, it ensures that only anatomical landmark points with structural consistency and high spatial confidence are retained, effectively avoiding feature localization deviations caused by factors such as image occlusion and projection blur, and enhancing the accuracy and robustness of anatomical recognition. By performing temporal dynamic matching and trajectory interpolation algorithms on a high-confidence anatomical landmark point set, a continuous pose change annotation sequence is established, which helps to track the structural change process of an individual's pose during movement or under load, providing data support for dynamic pose assessment and disease early warning. Based on the dynamic annotation sequence data, a pose angle change modeling mechanism is established, which can accurately measure the displacement trajectory and rotation angle of each key anatomical part; at the same time, combined with structural stability assessment, the biomechanical risk characteristics during the pose change process are further characterized, enhancing the depth and reliability of healthy pose assessment.

[0014] Preferably, the spatial pose calibration of the bone contour features using the dynamic marking system in step S2 includes: Performing spatial centroid registration on the anatomical landmark point sequence in the dynamic marking system to generate initial pose alignment reference data; Performing three-dimensional rigid registration on the bone contour features, and generating a roughly registered skeleton model for the rigidly registered bone contour features using the initial pose alignment reference data; Performing landmark-based non-rigid fine-tuning optimization on the roughly registered skeleton model to generate a high-precision pose alignment model; Extracting joint angles and solving the pose transformation matrix for the high-precision pose alignment model to generate a set of local pose transformation parameters; Performing global consistency harmonization and error convergence control on the set of local pose transformation parameters to generate spatial pose calibration parameters.

[0015] By performing spatial centroid registration on the anatomical landmark point sequence in the dynamic marking system, the present invention automatically generates initial alignment reference data for the pose, effectively avoiding the error of manual point selection, which is beneficial to improving the convergence speed and initialization robustness of subsequent registration steps. By performing three-dimensional rigid registration on the bone contour features, it can quickly align the main viewing direction of an individual in space with the standard coordinate system and maintain the consistency of the basic structure in operations such as multi-frame data fusion and dynamic pose comparison. On the basis of rigid registration, non-rigid fine-tuning optimization driven by landmark points is further introduced, which can effectively repair local alignment deviations caused by motion differences, soft tissue deformation or scanning errors, and improve the registration accuracy and detail fidelity of the pose model. Based on the joint angle information and pose transformation matrix extracted from the high-precision pose alignment model, it can provide high-dimensional parameter support for complex joint motion modeling, action recognition and motion analysis, significantly improving the expression ability of human motion data. By performing global harmonic and error convergence control on the local pose transformation parameter set, it can ensure the continuity and physiological rationality of the pose transformation relationship between each joint or torso segment, thereby constructing a stable and accurate spatial pose calibration model. The generated spatial pose calibration parameters can be widely applied to the normalization processing and time series comparison of multi-frame dynamic pose data, improving the comparability and quantifiable feature expression ability of the pose evolution process, and providing key technical support for pose training and behavior analysis.

[0016] Preferably, step S3 includes the following steps: Step S31: Input the spatial pose calibration parameters into a preset healthy pose evaluation database for pose similarity matching, and generate an analysis result of standard pose matching; Step S32: Calculate the three-dimensional deviation vectors of each skeleton joint node according to the standard pose matching result data, and generate a joint deviation index matrix; Step S33: Analyze the voxel density of the human three-dimensional point cloud data, and perform mechanical structure mapping in combination with the joint deviation index matrix to generate a biomechanical load distribution map; Step S34: Perform angle interval analysis and action amplitude reconstruction based on the joint deviation index matrix to generate a joint range of motion heat map.

[0017] By inputting the spatial attitude calibration parameters into a preset healthy attitude evaluation database for attitude similarity matching, the present invention can automatically identify the structural differences from the standard healthy attitude, providing a unified control model for attitude comparison among different populations or under specific disease states. Based on the standard matching results, the three-dimensional deviation vectors of each skeleton joint node are calculated and a joint deviation index matrix is constructed, which can comprehensively quantify the magnitude, direction, and range of the attitude errors of each joint in space, effectively improving the resolution and accuracy of attitude abnormality evaluation. By fusing and modeling the voxel density analysis of the human three-dimensional point cloud data with the joint deviation index matrix, the biomechanical risk areas such as local stress concentration and load offset caused by attitude abnormality can be effectively reflected, providing a basis for clinical rehabilitation and attitude correction. By performing angle interval analysis and motion amplitude reconstruction on the joint deviation index, a heat map that intuitively reflects the activity ability of each joint is generated, which helps to quickly identify the areas with limited movement, evaluate the trend of functional degradation, and support the design and effect tracking of the rehabilitation course.

[0018] Preferably, step S31 includes the following steps: Step S311: When any of the following conditions is met for the attitude calibration parameters, it is determined that the preliminary matching fails and preliminary attitude matching failure data is generated: the displacement error of the attitude key nodes exceeds the set threshold of 10 cm; the joint angle deviation exceeds 15°; the overall attitude stability score is lower than 0.6; Step S312: When the following conditions are simultaneously met, it is determined that the attitude of the characteristic area is mismatched and characteristic attitude mismatch data is obtained: the angular coordination between the core joint groups decreases by more than 25%; the movement trajectory deviation exceeds the standard range of the historical healthy model; the matching residuals are greater than the set upper limit of 0.08 in three consecutive evaluations, where the core joint groups include the shoulders, spine, and pelvis; Step S313: When all of the following situations occur, it is determined that the overall attitude is abnormal and attitude abnormality data is obtained: the overall three-dimensional attitude alignment error exceeds 12%; the body symmetry score is lower than 0.7; the movement sequence continuity score is lower than 0.65; and the error cannot be corrected by the adaptive compensation algorithm during the matching process; Step S314: Integrate the preliminary attitude matching failure data, characteristic attitude mismatch data, and attitude abnormality data to generate the analysis results of the standard attitude matching.

[0019] By setting multiple judgment conditions such as the displacement error of key nodes, the deviation of joint angles, and the overall attitude stability score, the present invention can quickly and accurately screen out the preliminary failure cases in attitude matching, avoid redundant calculations in subsequent analysis, and improve the overall operation efficiency of the system. Through multi-dimensional dynamic monitoring of the angle coordination, motion trajectory, and matching residuals of the core joint group (shoulder, spine, pelvis), precise identification of local attitude mismatch can be achieved, thereby timely reflecting subtle but critical attitude functional disorders and enhancing the sensitivity and pertinence of anomaly detection. By integrating indicators such as three-dimensional attitude alignment error, body symmetry score, and motion sequence continuity, and combining with an adaptive compensation algorithm, it is ensured that the overall attitude is determined to be abnormal only in cases of irreparable anomalies, reducing false alarms and improving the credibility of attitude anomaly recognition. By integrating preliminary matching failures, feature mismatches, and overall anomaly data, a comprehensive analysis of the attitude matching status in multiple dimensions and at multiple levels is realized, providing reliable and comprehensive data support for subsequent attitude correction, health assessment, and clinical decision-making.

[0020] Preferably, step S4 includes the following steps: Step S41: Use the coarsely registered skeleton model to perform depth contour reconstruction on the multi-angle projection image group of the human body to construct a 3D human body attitude image model; map the biomechanical load distribution map to the surface of the 3D human body attitude image model to generate load stress distribution modeling data; Step S42: Project the joint range of motion heat map onto the skeleton structure of the 3D human body attitude image model to generate a set of dynamic range parameters; perform multi-dimensional evaluation of attitude health based on the load stress distribution modeling data and the set of dynamic range parameters to generate an attitude anomaly marking layer; Step S43: Identify the attitude optimization area based on the attitude anomaly marking layer, and combine with a preset standard attitude model to match personalized correction strategies to generate attitude correction control parameters; apply the attitude correction control parameters to the key attitude nodes of the 3D human body attitude image model to generate an attitude correction simulation result; Step S44: Generate a personalized interpretation and evaluation summary of the attitude correction simulation result, and finally output a personalized attitude correction report.

[0021] The present invention realizes the fine construction of a three-dimensional human body pose image model by using a roughly registered skeleton model to perform depth contour reconstruction on a multi-angle projection image group, providing an accurate spatial basis for subsequent biomechanical load mapping and pose function analysis. The biomechanical load distribution map and the joint range of motion heat map are respectively mapped onto the three-dimensional model surface and the skeleton structure, realizing the dynamic combined evaluation of load stress and mobility, effectively revealing potential abnormal pose areas, and enhancing the scientificity and comprehensiveness of pose health assessment. Based on the abnormal pose marking layer, the abnormal areas are intelligently identified, and personalized correction strategies are matched in combination with the standard pose model, realizing the precise positioning and targeted regulation of the abnormal parts, and improving the pertinence and effectiveness of the correction effect. The correction control parameters are applied to the key nodes of the three-dimensional pose model to complete the dynamic simulation of pose correction, facilitating the verification of the actual effect of the correction strategy in a virtual environment, reducing the clinical trial and error cost, and enhancing the scientificity of personalized intervention. Through the detailed personalized interpretation and evaluation summary generation of the pose correction simulation results, an easy-to-understand and professional pose correction report is finally output, which helps patients and clinicians to comprehensively master the pose state and correction plan.

[0022] In this specification, a health assessment system based on a 3D human body pose image model is provided for performing the above-mentioned health assessment method based on a 3D human body pose image model. The health assessment system based on a 3D human body pose image model includes: A point cloud projection module for obtaining human body circular scan data; extracting the dynamic pose sequences of the upright position, forward flexion position, and lateral bending position of the human body circular scan data, and performing phase-shifted fringe projection on the human body circular scan data according to the dynamic pose sequences to obtain human body three-dimensional point cloud data; A pose analysis module for performing orthogonal projection on the sagittal plane, coronal plane, and transverse plane of the human body three-dimensional point cloud data to generate a multi-angle projection image group of the human body; identifying the bone contour features of the multi-angle projection image group of the human body to establish a dynamic marking system for anatomical landmark points; using the dynamic marking system to perform spatial pose calibration on the bone contour features to generate spatial pose calibration parameters; A pose deviation calculation module for inputting the spatial pose calibration parameters into a preset healthy pose assessment database for multi-dimensional comparative analysis, and calculating the deviation index between the parts in the human body three-dimensional point cloud data and the standard value according to the analysis results to construct a biomechanical load distribution map and a joint range of motion heat map; A health assessment module for constructing a 3D human body pose image model based on the multi-angle projection image group of the human body; performing pose health assessment on the 3D human body pose image model through the biomechanical load distribution map and the joint range of motion heat map respectively, and performing personalized pose correction according to the assessment results, thereby generating a personalized pose correction report.

[0023] The beneficial effects of the present invention are as follows: By combining circular scanning with phase-shifted fringe projection technology, the dynamic pose sequences of the human body in the standing position, forward flexion position, and lateral bending position are accurately captured, effectively improving the integrity and dynamic capture ability of the 3D point cloud data, and providing a solid data foundation for subsequent pose analysis. By performing orthogonal projections on the 3D point cloud data in the sagittal plane, coronal plane, and transverse plane to generate a group of multi-angle projection images, and combining the skeletal contour feature recognition and the dynamic marking system of anatomical landmark points, the precise spatial calibration of the human skeletal pose is achieved, effectively improving the spatial accuracy and dynamic consistency of pose analysis. By matching the spatial pose calibration parameters with the healthy pose database and performing multi-dimensional comparisons, the deviation index of each part of the skeleton is accurately calculated, and a biomechanical load distribution map and a heat map of joint range of motion are constructed, enhancing the quantitative assessment ability of pose abnormalities and the understanding of structural mechanics. Based on the group of multi-angle projection images, a 3D human pose model is constructed, integrating the load distribution and range of motion information, realizing the comprehensive health assessment and precise correction of individual poses, and generating a personalized pose correction report. Each module works in coordination to form a complete closed-loop from human dynamic data acquisition, spatial pose analysis to health assessment and personalized intervention, greatly improving the automation, intelligence level and practical value of human pose health management. Therefore, through high-precision dynamic acquisition and multi-dimensional spatial analysis, the present invention realizes the precision, standardization and personalization of 3D human pose health assessment, and improves the accuracy and comprehensiveness of human pose health assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the step flow of a health assessment method based on a 3D human pose image model; Figure 2 For Figure 1 a detailed implementation step flow diagram of step S3 in Figure 3 For Figure 1 a detailed implementation step flow diagram of step S4 in The realization, functional features, and advantages of the objectives of the present invention will be further described in conjunction with the embodiments with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The technical method of the present invention will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve the above object, please refer to Figures 1 to 3 , a health assessment method based on a 3D human body pose image model, the method comprising the following steps: Step S1: Obtain human body circular scan data; extract the dynamic pose sequences of the standing position, forward flexion position and lateral bending position of the human body circular scan data, and perform phase-shifted fringe projection on the human body circular scan data according to the dynamic pose sequences to obtain human body three-dimensional point cloud data; Step S2: Analyze the sagittal plane, coronal plane and transverse plane of the human body three-dimensional point cloud data for orthogonal projection to generate a group of human body multi-angle projection images; identify the bone contour features of the human body multi-angle projection image group to establish a dynamic marking system for anatomical landmark points; use the dynamic marking system to calibrate the spatial pose of the bone contour features to generate spatial pose calibration parameters; Step S3: Input the spatial pose calibration parameters into a preset health pose assessment database for multi-dimensional comparative analysis, and calculate the deviation index between the parts in the human body three-dimensional point cloud data and the standard value according to the analysis results to construct a biomechanical load distribution map and a joint range of motion heat map; Step S4: Construct a 3D human body pose image model based on the human body multi-angle projection image group; perform pose health assessment on the 3D human body pose image model respectively through the biomechanical load distribution map and the joint range of motion heat map, and perform personalized pose correction according to the assessment results to generate a personalized pose correction report.

[0029] The present invention realizes high-precision three-dimensional reconstruction of human postures in multiple motion states through the acquisition of dynamic posture sequences in the upright position, forward flexion position, and lateral flexion position, combined with the phase-shifted fringe projection technology, improving the integrity and dynamic expressiveness of data. By using multi-angle orthogonal projections of the sagittal plane, coronal plane, and transverse plane, the skeletal contour features are automatically identified, and a dynamic marking system is established to calibrate the spatial posture of the posture, ensuring the accuracy and repeatability of subsequent analysis. By comparing the calibration parameters with a healthy posture database in multiple dimensions, a deviation index is calculated to accurately characterize the difference between the individual and the standard; and further, a biomechanical load distribution map and a joint range of motion heat map are generated to achieve multi-dimensional health assessment. Based on the multi-angle projection images, an intuitive 3D posture image model is constructed, and combined with the superimposed analysis of the heat map and load map, the visualization and interactivity of the results are enhanced, facilitating users and doctors to intuitively understand the posture problems. According to the analysis results of the individual's load distribution and range of motion, a personalized correction strategy is formulated for them, and finally a customized posture correction report is generated, which helps to improve the accuracy and effectiveness of posture intervention. Therefore, through high-precision dynamic acquisition and multi-dimensional spatial analysis, the present invention realizes the precision, standardization, and personalization of 3D human posture health assessment, improving the accuracy and comprehensiveness of human posture health assessment.

[0030] In an embodiment of the present invention, referring to Figure 1 as shown, it is a schematic flowchart of the steps of a health assessment method based on a 3D human posture image model of the present invention. In this example, the health assessment method based on a 3D human posture image model includes the following steps: Step S1: Obtain human circular scan data; extract the dynamic posture sequences of the upright position, forward flexion position, and lateral flexion position of the human circular scan data, and perform phase-shifted fringe projection on the human circular scan data according to the dynamic posture sequences to obtain human three-dimensional point cloud data; Step S2: Analyze the orthogonal projections of the sagittal plane, coronal plane, and transverse plane of the human three-dimensional point cloud data to generate a group of human multi-angle projection images; identify the skeletal contour features of the human multi-angle projection image group to establish a dynamic marking system for anatomical landmark points; use the dynamic marking system to calibrate the spatial posture of the skeletal contour features to generate spatial posture calibration parameters; Step S3: Input the spatial posture calibration parameters into a preset healthy posture assessment database for multi-dimensional comparative analysis, and calculate the deviation index between the parts in the human three-dimensional point cloud data and the standard value according to the analysis results to construct a biomechanical load distribution map and a joint range of motion heat map; Step S4: Construct a 3D human body pose image model based on the multi-angle projection image group of the human body; perform pose health assessment on the 3D human body pose image model through the biomechanical load distribution map and the joint range of motion heat map respectively, and perform personalized pose correction according to the assessment results, so as to generate a personalized pose correction report.

[0031] In the embodiment of the present invention, the human body is scanned in a 360° circular manner by using a high-precision multi-view 3D scanner to obtain data in the upright position, forward flexion position and lateral bending position respectively; an image pose recognition algorithm (such as OpenPose or MediaPipe) is used to classify and label the scanned data sequence to generate a dynamic pose sequence set. In each pose, the body surface is dynamically optically encoded by using the Phase-Shift Profilometry; the depth of each pixel is calculated by the multi-phase difference method, and combined with the parallax and geometric constraints, a high-density three-dimensional point cloud data of the human body is generated; the point cloud data is filtered, denoised and edge-complemented to output a complete human body point cloud model. The human body point cloud data is orthogonally projected on the sagittal plane, coronal plane and transverse plane; a high-resolution projection image is generated under each view angle to form a multi-angle image group covering the whole body. Based on a deep learning image semantic segmentation model (such as U-Net or HRNet), the main bone structure contours of the human body (cervical vertebra, thoracic vertebra, pelvis, etc.) are extracted; anatomically traceable landmark points are set at key nodes, and a dynamic landmark marking system (DLMS) is constructed to track the displacement changes of the landmark points in different poses. By using the position changes of the multi-frame anatomical landmark points in the DLMS, a multi-view coordinate system registration model is established; through the rigid transformation matrix and the pose inverse solution algorithm, a complete set of spatial pose calibration parameters (including rotation angle, offset vector, etc.) is generated. The pose calibration parameters are input into the healthy pose assessment database (including 3D skeleton pose models of a large number of standard populations); the deviations of different parts (such as spinal angle, pelvic tilt, etc.) from the standard values are compared, and the formula ; where is the deviation index of the th part, is the pose parameter measured at the th part, It is the standard value. The finite element estimation algorithm (FEM) is used to simulate the pressure-bearing conditions of each joint and bone on the human body point cloud; according to the angle change range and activity radius, a color heat map of joint mobility is generated; the output results are presented in the form of a visualization atlas for downstream evaluation and personalized intervention. Reconstruct a complete 3D human body pose model based on point cloud and skeleton data; fuse multi-angle projection images to improve the modeling accuracy, and perform topology correction and skin texture mapping. Overlay the biomechanical load map and joint heat map on the 3D model for comprehensive analysis; evaluate the trunk asymmetry, scoliosis degree, posture stability, etc. based on deviation indicators, load abnormal areas, and joint stiffness indices. The system generates a correction report including the following based on the analysis results: the position and direction of posture deviation, visualization heat map and load map, personalized posture training suggestions (such as stretching, corrective actions, use of assistive devices), and a template for tracking the progress of posture improvement.

[0032] Preferably, step S1 includes the following steps: Step S11: Use a scanning device to perform a 360° circular scan of the human body to obtain human circular scan data; Step S12: Extract the scan sequence of the human circular scan data, and perform inter-frame synchronous registration on the human circular scan data according to the scan sequence to generate synchronously corrected human scan frame data; Step S13: Extract the dynamic skeleton points of the human scan frame data; perform pose annotation on the dynamic skeleton points to generate dynamic pose sequences in the upright position, forward flexion position, and lateral flexion position; Step S15: Decompose the spatial pose parameters of the dynamic pose sequence to obtain human pose distribution control data; perform fringe phase encoding and multi-phase offset fringe map generation on the human scan frame data to obtain a multi-phase offset encoding atlas; Step S16: Perform phase unwrapping on the multi-phase offset encoding atlas through the human pose distribution control data to generate a dense phase mapping; perform triangulation and depth inversion calculation on the dense phase mapping to generate high-precision human three-dimensional point cloud data.

[0033] In the embodiments of the present invention, by adopting a structured light scanning device with multi-view synchronous imaging (or a ToF+RGB fusion system), a plurality of high-resolution cameras and a projection unit are arranged around the human body; the projection system is controlled to project a structured light stripe pattern onto the human body at fixed angular intervals, and at the same time, a plurality of cameras synchronously collect the reflected images of the human body surface; a synchronous triggering mechanism and a high-frame-rate camera (>60fps) are used to ensure no dynamic blur throughout the scanning process, and an original human body circular scanning image sequence is output. The viewing angle parameters and timestamps of each frame in the original scanning image sequence are extracted to establish a time-angle index model; an inter-frame synchronous registration algorithm (such as optical flow registration or depth alignment based on the ICP algorithm) is used to perform time synchronization and spatial registration on frames from different viewpoints; a synchronized and corrected human body scanning frame dataset is output, that is, the alignment result of all frames in the unified time and unified spatial coordinate system. Based on a deep learning skeleton recognition network (such as OpenPose 3D or MediaPipe Skeleton), a 3D skeleton key point sequence (shoulder, elbow, spine, pelvis, etc.) is extracted from the synchronized and corrected scanning frames; according to the temporal change trajectory of the skeleton points, a dynamic pose classifier (such as a support vector machine or an LSTM classification model) is applied to label the sequence as: upright position (Upright), forward bending position (Forward Bending), lateral bending position (Lateral Bending), and a dataset of three types of labeled dynamic pose sequences is output. For each type of dynamic pose sequence, based on the Euler angle decomposition or quaternion decomposition algorithm, spatial pose parameters (Pitch, Yaw, Roll) are extracted; each frame of the pose is represented as a six-degree-of-freedom (6DoF) pose vector to construct a complete human body pose distribution control data; at the same time, a phase encoding technology (Phase-Shift Encoding) is applied to each frame of data to generate a structured light stripe pattern: four sets of stripe patterns with a 90° phase shift from each other (i.e., 0°, 90°, 180°, 270°) are projected to obtain a multi-phase offset encoded image set; a high-quality structured light encoded image set for subsequent three-dimensional reconstruction is output. The encoded image set is spatially guided and unpacked using the human body pose distribution control data to avoid phase errors caused by pose changes; phase solving is performed for each pixel point (using: ; where ∼ are the gray values of the four phase-shifted images), and a dense phase mapping diagram is generated; triangulation and depth inversion calculations are performed based on the phase mapping diagram: through the spatial baseline of the calibrated projection-camera pair, the depth value of each pixel is calculated by combining the phase difference; the depth map is converted into three-dimensional coordinates, the point clouds from each viewpoint are fused, and a point cloud registration + fusion filtering algorithm (such as Poisson reconstruction) is used to generate the final: high-precision human body three-dimensional point cloud data (with an accuracy of up to sub-millimeter level).

[0034] Preferably, the pose annotation of the dynamic skeleton points in step S13 includes: Collect the three-dimensional coordinates of the skeleton points through an optical motion capture system to obtain the three-dimensional coordinates of the skeleton points, where the sampling frequency is set to ≥30 frames per second, the spatial accuracy is ±2 mm, the coordinate unit is millimeters, and the right-handed coordinate system is referenced; Define the recognition criteria for the upright position: the overall inclination angle of the spine ≤5°, the height difference between the left and right shoulders ≤10 mm, and the knee flexion angle ≤10°; Define the recognition criteria for the forward flexion position: the overall forward inclination angle of the spine ≥30° and ≤90°, the curvature radius from the neck to the lumbar spine ≤400 mm, and the hip flexion angle ≥45°; Define the recognition criteria for the lateral bending position: the lateral bending angle of the spine ≥20° and ≤60°, the height difference between the left and right hips ≥15 mm, and the horizontal distance of the shoulder-hip connection line deviating from the central axis ≥30 mm; Based on the defined recognition criteria for the upright position, forward flexion position, and lateral bending position, perform pose classification and annotation on the three-dimensional coordinates of the skeleton points, thereby obtaining the dynamic pose sequences of the upright position, forward flexion position, and lateral bending position.

[0035] In the embodiments of the present invention, by using an optical motion capture system (such as Vicon, OptiTrack), one or more infrared high-speed cameras are arranged to form a closed capture area; highly reflective marker points are pasted on the key bone points of the human body (head, cervical vertebra, acromion, elbow, wrist, thoracic vertebra, lumbar vertebra, hip, knee, ankle) to construct a human skeleton point marking model; the system configuration is as follows: sampling frequency: ≥30 frames / second, spatial accuracy: ±2 mm, coordinate unit: millimeter (mm), reference coordinate system: right-handed coordinate system (X forward, Y left, Z upward), and a three-dimensional coordinate data sequence of skeleton points under continuous time frames is output. Calculate the angle between the line connecting the seventh cervical vertebra (C7) and the first sacral vertebra (S1) and the Z axis, which is the overall inclination angle of the spine. Calculate the absolute height difference between the highest points of the left shoulder and the right shoulder on the Z axis, which is the height difference between the left and right shoulders. Calculate the angle between the thigh vector and the calf vector, which is the knee flexion angle. The characteristics for upright position recognition are defined by preset thresholds. Calculate the anterior inclination angle of the line connecting C7–S1 and the Z axis, which is the anterior inclination angle of the spine. Use spline interpolation to reconstruct the central axis of the spine, and the local curvature radius of the fitted curve is the curvature radius from the neck to the lumbar vertebra. Calculate the angle between the trunk and the thigh, which is the hip flexion angle. Similarly, the characteristics for forward flexion position recognition are defined by preset thresholds. Calculate the offset angle of the line connecting the midpoint of the thoracic vertebra and the center of the pelvis in the Y-Z plane, which is the scoliosis angle of the spine. Calculate the height difference between the left hip and the right hip on the Z axis, which is the height difference between the left and right hips. Calculate the maximum offset amount of the line connecting the left shoulder and the right hip projected onto the central axis (human sagittal plane) in the X-axis direction, which is the distance of the shoulder-hip connection deviating from the central axis. The characteristics for scoliosis position recognition are defined by preset thresholds. Traverse the skeleton data of each frame, and sequentially extract and determine whether each frame meets a certain recognition standard group (upright, forward flexion, scoliosis). A priority judgment mechanism is adopted: if multiple categories are matched, they are preferentially marked in the order of "scoliosis > forward flexion > upright". The classification results of consecutive frames are combined into a posture time series, and a dynamic posture series is output.

[0036] Preferably, the orthogonal projection by analyzing the sagittal plane, coronal plane and transverse plane of the human three-dimensional point cloud data in step S2 includes: Normalize the center of gravity of the spatial coordinates of the human three-dimensional point cloud data to generate standardized three-dimensional posture point cloud data; Extract the local principal axis of the standardized three-dimensional posture point cloud data, and use the local principal axis to calibrate the body axis vector field of the three-dimensional posture point cloud data to generate structure-aligned vector calibration data; Perform depth mapping rearrangement on the structure-aligned vector calibration data in the sagittal plane direction to generate a sagittal plane orthogonal projection map; Perform depth value aggregation and compression on the structure-aligned vector calibration data in the coronal plane direction to generate a coronal plane orthogonal projection map; Hierarchically slice and resample the structure alignment vector calibration data in the cross-sectional direction to generate an orthogonal projection map of the cross-section. Perform image group registration and angle label binding on the sagittal plane, coronal plane, and orthogonal projection map of the cross-section to generate a multi-angle projection image group of the human body.

[0037] In the embodiment of the present invention, the obtained three-dimensional point cloud data of the human body needs to be uniformly processed in spatial coordinates. This processing takes the overall center of gravity of the human body as the benchmark, and realizes the center of gravity normalization of the point cloud data by performing spatial translation on the data of all points. This step can ensure that subsequent analysis is based on a unified spatial reference system, which is beneficial to maintaining pose consistency and data stability. In the normalized point cloud data, the main directions reflecting the overall structure of the human body are extracted, including the longitudinal, transverse, and front-back directions of the body. The extracted directions will be used as the structural reference axes of the human body to perform spatial structure alignment on the human body pose. Subsequently, based on this structural direction, direction information corresponding to the spatial position of each point in the three-dimensional point cloud is assigned to complete the calibration of the structure alignment vector. According to the calibrated structural direction, the three-dimensional point cloud of the human body is rearranged in depth information along the left-right direction to display the human body pose characteristics observed from the side. Specifically, the point cloud data is projected onto a plane reflecting the front-back and up-down dimensions of the body to form an orthogonal projection map in the sagittal plane direction. This map clearly presents the contour changes of the spine, head, hips, etc. observed from the side. Taking the front-back direction of the body as the aggregation dimension, the point cloud data is compressed and integrated in the front-back direction to generate a projection map of the human body structure observed from the front. This map can effectively show the symmetry and deformation degree of parts such as the shoulders, chest, and pelvis in the left-right direction, and is suitable for structural evaluation and pose analysis. By performing layer-by-layer slicing in the longitudinal direction of the human body, the point cloud data at different height positions is hierarchically divided and projected to generate a cross-sectional view observed from top to bottom. This map helps to analyze features such as the width distribution, rotation angle, and local symmetry of the human body at different height levels, and is especially suitable for the analysis of the abdominal, lumbar, and pelvic structures. The projection maps generated in the above three directions are subjected to spatial alignment processing to ensure the visual unity of the central position, image size, and edge information of the images. Then, an angle recognition label is bound to each image to clarify the observation direction it represents, such as a side view, front view, top view, etc., for subsequent analysis and model processing. Finally, a multi-angle projection image group of the human body is generated to support application scenarios such as pose recognition, structural comparison, or intelligent analysis.

[0038] Preferably, identifying the bone contour features of the multi-angle projection image group of the human body to establish a dynamic marking system for anatomical landmark points includes: Identifying the main bone contours of the multi-angle projection image group of the human body; Analyzing the connected domains of the multi-angle projection image group of the human body according to the main bone contours to obtain a primary skeleton distribution map; Detect key morphological turning points in the primary skeleton distribution map, generate candidate anatomical structure corner point data, and perform a spatial consistency test based on depth prior on the candidate anatomical structure corner point data to generate a high-confidence anatomical landmark set; Perform temporal pose dynamic matching and trajectory interpolation on the high-confidence anatomical landmark set to generate dynamic annotation sequence data; Model the pose angle change and evaluate the structural stability of the high-confidence anatomical landmark set through the dynamic annotation sequence data to generate a dynamic marking system for anatomical landmarks.

[0039] In the embodiments of the present invention, by preprocessing a group of human body multi-angle projection images, including image enhancement, edge thinning, and noise filtering. Subsequently, an identification algorithm based on shape gradient change and edge continuity is adopted to extract the main bone contour regions in each projection image. The extracted main bone contours usually include structures such as the cranial edge, spinal midline, shoulder contour, pelvic boundary, and limb bone lines. These regions have obvious geometric features and are easy to separate from the background. In the obtained main bone contour image, the image connectivity analysis method is used to identify the regions with continuous pixel connection relationships, thereby dividing the connected domains corresponding to different anatomical structures. Combining the topological relationships, direction information, and contour extension characteristics of each connected domain, a rough skeleton orientation map, that is, a primary skeleton distribution map, is constructed to describe the spatial extension path of the main bony structures of the human body. For the primary skeleton distribution map, the morphological turning points on its structural path are further identified. These turning points usually correspond to the positions of bone connections, bends, or local protrusions, such as the turning point from the cervical vertebra to the thoracic vertebra, the acromion, the outer edge of the hip bone, etc. By calculating the positions, angle changes, and neighborhood geometric features of these turning points, a set of candidate anatomical structure corner point data is extracted. The above candidate corner points are verified for spatial consistency between different angle projection images. This verification is based on the principle of the reasonableness of human body posture changes, combined with human body structure symmetry, depth information hierarchy, and anatomical common sense, to eliminate abnormal points and artifact interference points, and finally a set of anatomical landmark points with relatively high credibility is selected. This set covers the key parts of the whole body and has good spatial stability and anatomical significance. Based on the human body movement changes between each frame of images, the high-confidence anatomical landmark points are dynamically trajectory-matched across time points. Through the point position tracking between consecutive frames, a dynamic trajectory of the anatomical landmark points changing with time is formed. In the case where the landmark points are missing due to factors such as occlusion and noise, a trajectory interpolation algorithm based on motion laws and structural coherence is used for completion to generate a complete dynamic annotation sequence data. Based on the dynamic annotation sequence data, the attitude angle change conditions of each anatomical landmark point at different time points are modeled, and its attitude change amplitude, frequency, and coordination are calculated. Combining the relative position relationships and motion direction trends between points, the stability and coordination degree of the anatomical structure during the dynamic process are evaluated, thereby forming an anatomical landmark dynamic marking system for action analysis and pose recognition.

[0040] Particularly importantly, the detection of key morphological turning points for the primary skeleton distribution map further includes: Performing curvature change detection on the primary skeleton distribution map to generate data of local curvature extreme points of the skeleton; Performing direction gradient analysis on the data of local curvature extreme points of the skeleton to generate candidate posture mutation point data; Performing spatial continuity filtering on the candidate posture mutation point data to generate data for screening the stability of turning points; Perform structural consistency matching on the stability screening data of turning points to generate potential anatomical structure corresponding point data; Identify the semantic geometric features of the potential anatomical structure corresponding point data to obtain candidate anatomical structure corner point data.

[0041] In the embodiment of the present invention, by performing contour tracking on continuous skeleton line segments in the primary skeleton distribution map, local geometric morphological features on each skeleton path are extracted. Using the spatial angle between continuous points, the curve change trend, and the morphological undulation, the local curvature change of the skeleton path is evaluated. Mark the positions with significant curvature changes, and extract the curvature extreme points forming a mutation trend. This processing process can initially discover the key areas where the skeleton structure deforms or the direction deflects. Based on the obtained curvature extreme points, the direction gradient information of the areas around each extreme point is further calculated, including features such as the sudden change amplitude of the main skeleton direction, the left and right deflection trends, and the discontinuity of the orientation. By setting a threshold to screen the direction change amplitude, only the points with obvious direction mutation properties are retained as candidate posture mutation points, and these points usually correspond to joint turning points, structure connection points, or posture change positions. Perform spatial continuity analysis on the distribution of candidate posture mutation points in the entire primary skeleton map. If a certain candidate point cannot form a stable structure continuation in its adjacent area or lacks consistency with the upstream and downstream skeleton directions, it is excluded. Only the points with features such as repeated appearance, continuous direction, and stable position in multiple adjacent projection maps or frame maps are retained to form the stability screening data of turning points, and this process effectively excludes false turning points caused by image noise or local errors. On the basis of stability screening, perform spatial structure matching between each candidate turning point and the known anatomical structure model of the human body. By comparing the relative distances, angular relationships, and topological paths of the branches to which the points belong, the anatomical structure attribution of each candidate point is judged. If its structural attribute has a high consistency with a certain known anatomical part, it is identified as a potential anatomical structure corresponding point, thus providing a physical basis for subsequent semantic recognition. Finally, perform further semantic recognition on the potential anatomical structure corresponding points, and conduct classification judgment by combining their geometric morphological features and anatomical functional characteristics. This process mainly determines whether it belongs to a corner structure with semantic discrimination, such as the acromion, iliac spine, knee joint, ankle point, etc., based on information such as the spatial position, morphological features, and motion significance of the points. The points screened through this recognition process are finally confirmed as candidate anatomical structure corner points, providing key node support for dynamic posture modeling.

[0042] Preferably, the spatial posture calibration of the bone contour features using the dynamic marking system in step S2 includes: Perform spatial centroid registration on the anatomical landmark point sequence in the dynamic marking system to generate initial alignment reference data for the posture; Perform three-dimensional rigid registration on the bone contour features, and generate a roughly registered skeleton model for the bone contour features after rigid registration by using the initial alignment reference data of the pose; Perform non-rigid fine-tuning optimization based on landmark points on the roughly registered skeleton model to generate a high-precision pose alignment model; Extract joint angles and solve the pose transformation matrix for the high-precision pose alignment model to generate a set of local pose transformation parameters; Perform global consistency reconciliation and error convergence control on the set of local pose transformation parameters to generate spatial pose calibration parameters.

[0043] In the embodiment of the present invention, by calculating the spatial coordinates of the sequence of anatomical landmark points collected in the dynamic marking system, the overall center of gravity position is obtained. By aligning the centers of gravity of the landmark point sequences in different poses, the spatial deviation caused by translation is eliminated, and unified initial alignment reference data of the pose is generated. This step provides a stable spatial reference for subsequent registration, ensuring the overall position consistency of different bone poses. Then, the bone contour features are transformed into a three-dimensional point cloud model, and based on the principle of rigid transformation, rotation and translation registration in three-dimensional space are performed. Using the initial alignment reference data of the pose generated in the first step, the bone contour model after rigid registration is adjusted to achieve rough spatial superposition, and a roughly registered skeleton model is obtained. This model has good structural alignment but has not yet solved the local detail differences. On the basis of the rough registration, a non-rigid registration algorithm is used to fine-tune the skeleton model. By locally constraining the positions of the key anatomical landmark points in the dynamic marking system and combining with the surface deformation of the model, the local pose of the bone contour is adjusted to more accurately match the actual human bone morphology. This process eliminates local deformation errors, improves the registration accuracy, and generates a high-precision pose alignment model. Using the high-precision pose alignment model, angle measurements are performed on the joint parts of the human bone to extract the rotation angles and spatial direction changes of each joint. According to the joint angle information, the corresponding local pose transformation matrix is solved to form a set of local pose transformation parameters. These parameters accurately describe the spatial rotation and transformation relationships of each part of the bone and are the basic data for subsequent pose analysis and adjustment. Finally, global unified reconciliation is performed on the set of local pose transformation parameters to ensure the consistency of the spatial positions and directions between local transformations. Through iterative calculation of error convergence, each transformation parameter is adjusted to achieve the optimal coordination of the overall pose. This process effectively reduces the registration error, ensures the accuracy and stability of the spatial pose calibration parameters, and finally generates reliable spatial pose calibration parameters.

[0044] Particularly importantly, performing three-dimensional rigid registration on the bone contour features and generating a roughly registered skeleton model for the bone contour features after rigid registration by using the initial alignment reference data of the pose further includes: Extract the key bone feature points of the bone contour features; Based on the key skeletal feature point set, perform three-dimensional rigid transformation estimation to generate a preliminary rigid registration matrix; Use the preliminary rigid registration matrix to perform a rigid transformation on the skeletal contour features to generate rigidly registered skeletal contour data; Through the preset pose initial alignment reference data, calculate the rough registration error metric of the rigidly registered skeletal contour data to generate registration error index data; Optimize the preliminary rigid registration matrix according to the registration error index data to generate an optimized rigid registration matrix; Use the optimized rigid registration matrix to perform a secondary transformation on the skeletal contour features to generate accurately aligned skeletal contour data; Based on the accurately aligned skeletal contour data, construct a skeleton topology model to generate a roughly registered skeleton model.

[0045] In the embodiments of the present invention, by identifying and extracting a set of representative key skeletal feature points from the skeletal contour feature data, these points usually correspond to important anatomical nodes of the human skeleton, ensuring that the subsequent registration process can achieve accurate alignment based on the key points. Based on the extracted key skeletal feature points, by calculating the best rigid transformation relationship between point sets in space, a preliminary rigid registration matrix is estimated. This matrix contains rotation and translation parameters and preliminarily describes the spatial transformation of the skeletal contour from the original pose to the target pose. Using the preliminarily estimated rigid registration matrix, perform a spatial rigid transformation on the skeletal contour features so that its overall shape and position approach the target reference pose to form preliminary rigidly registered skeletal contour data. Use the preset pose initial alignment reference data as a standard to evaluate the error of the rigidly registered skeletal contour data. Calculate the distance difference between the skeletal key points and the reference data to form a registration error index, reflecting the quality of the registration accuracy. According to the registration error index, adjust the parameters of the preliminary rigid registration matrix through an iterative optimization method to reduce the registration error. The optimization process aims to find a more accurate combination of rotation and translation to achieve a higher precision in the spatial matching between the skeletal contour features and the reference data. Using the optimized rigid registration matrix, perform a second rigid transformation on the skeletal contour features to obtain more accurately aligned skeletal contour data, ensuring that the skeletal shape highly coincides with the pose initial reference benchmark. Based on the accurately aligned skeletal contour data, establish the topological connection relationship of the skeleton, organically integrate each skeletal node and its spatial connection to generate a complete roughly registered skeleton model. This model not only retains the skeletal spatial structure but also has a basis for subsequent non-rigid fine-tuning.

[0046] As an example of the present invention, refer to Figure 2 As shown, in this example, step S3 includes: Step S31: Input the spatial pose calibration parameters into a preset healthy pose assessment database for pose similarity matching to generate an analysis result of standard pose matching; Step S32: Calculate the three-dimensional deviation vectors of each skeleton joint node based on the standard posture matching result data, and generate a joint deviation index matrix; Step S33: Analyze the voxel density of the human body three-dimensional point cloud data, and perform a mechanical structure mapping in combination with the joint deviation index matrix to generate a biomechanical load distribution map; Step S34: Perform an angle interval analysis and motion amplitude reconstruction based on the joint deviation index matrix to generate a joint range of motion heat map.

[0047] In the embodiments of the present invention, the parameters obtained through spatial attitude calibration are integrated and input into a pre-established healthy attitude evaluation database. This database contains multiple sets of standard attitude template data that have been strictly collected and screened. Each template corresponds to the three-dimensional coordinates of key skeletal nodes and attitude feature information of the human body in a specific posture. The system evaluates the matching degree between the input spatial attitude parameters and each standard template in the database by comparing frame by frame and using similarity calculation methods (such as point-to-point distance calculation, angle difference analysis, etc.). According to the matching degree ranking, the standard attitude template closest to the input attitude is selected, and the corresponding matching result data is output, including the matching standard template number, matching score, and error distribution during the matching process. Based on the best-matched standard attitude selected in step S31, for each key joint node of the human skeleton, the three-dimensional space deviation between the current attitude node and the standard node is calculated. The specific operation is as follows: Extract the skeletal node coordinates of the current attitude and the standard attitude, calculate the differences in the X, Y, and Z coordinate axes respectively to obtain the three-dimensional deviation vector of the joint. Calculate the magnitude of each joint deviation vector to form a joint deviation index. Arrange the deviation indices of all joints in the order of the skeletal structure to generate a joint deviation index matrix, and each element in the matrix represents the spatial deviation magnitude of the corresponding joint. This matrix serves as the key input data for subsequent biomechanical analysis. For the three-dimensional point cloud data of the human body, first divide the three-dimensional space into equal-volume voxel units, count the number of point cloud points contained in each voxel unit, and generate a point cloud voxel density distribution map. Then, combined with the joint deviation index matrix in step S32, according to the human skeletal structure and joint positions, map the joint deviation data to the corresponding voxel regions. During the mapping process, a weight assignment method is adopted to consider the influence range and decreasing trend of joint deviation on neighboring voxels, and calculate the force contribution of each voxel. Next, according to the principle of mechanical transmission and the human skeletal mechanical model, calculate the load value in each voxel, and the load value reflects the strength of the force in this area. Finally, present the load values of each voxel in the form of a color gradient corresponding to the spatial position to generate a biomechanical load distribution map. Using the joint deviation index matrix and dynamic attitude data, the range of motion angles of each key joint is detailedly divided. By analyzing the angle change data in the dynamic attitude sequence, count the motion amplitude and frequency of the joint in each angle interval to obtain the motion amplitude distribution. Use interpolation method to continuously reconstruct the motion amplitude in the angle interval to form a complete motion amplitude change curve. Subsequently, combined with the joint deviation index, evaluate the degree of joint movement restriction in each angle interval. Finally, display this data in the form of a heat map. The heat map distinguishes the motion amplitude and restriction degree of different joints by the depth of color to form a joint range of motion heat map, which intuitively reflects the functional state of the joints.

[0048] Preferably, step S31 includes the following steps: Step S311: When the attitude calibration parameters meet any of the following conditions, it is determined that the preliminary matching fails and preliminary attitude matching failure data is generated: the displacement error of the attitude key nodes exceeds the set threshold of 10 cm; the joint angle deviation exceeds 15°; the overall attitude stability score is lower than 0.6; Step S312: When the following conditions are met simultaneously, it is determined that the attitude of the feature region is mismatched and feature attitude mismatch data is obtained: the angular coordination between the core joint groups decreases by more than 25%; the movement trajectory deviates beyond the standard range of the historical healthy model; the matching residuals are greater than the set upper limit of 0.08 in three consecutive evaluations, where the core joint groups include the shoulders, spine, and pelvis; Step S313: When all of the following situations occur, it is determined that the overall attitude is abnormal and attitude abnormal data is obtained: the overall three-dimensional attitude alignment error exceeds 12%; the body symmetry score is lower than 0.7; the movement sequence continuity score is lower than 0.65; and the error cannot be corrected by the adaptive compensation algorithm during the matching process; Step S314: Integrate the preliminary attitude matching failure data, feature attitude mismatch data, and attitude abnormal data to generate the analysis result of the standard attitude matching.

[0049] In the embodiments of the present invention, by detecting the input spatial attitude calibration parameters, it is determined whether the displacement error of each key bone node exceeds a preset threshold of 10 cm; at the same time, the angle deviation of each joint is calculated to determine whether it exceeds the limit of 15 degrees; in addition, the system also calculates the stability score of the overall attitude based on multi-dimensional attitude data. If the score is lower than 0.6, it is determined that the preliminary matching fails. At this time, the system generates and saves the preliminary attitude matching failure data including the failure reason and relevant attitude data. When the system detects that the angular coordination between the core joint groups (including the shoulders, spine, and pelvis) has decreased by more than 25% compared to the historical healthy model, and the movement trajectories of the core joints show obvious deviations beyond the allowable range of the standard healthy model, and at the same time, in three consecutive matching evaluations, the matching residual values continue to be greater than the set upper limit of 0.08, the system determines that there is a posture mismatch in the characteristic area. The system will record this state and generate the characteristic posture mismatch data including the above abnormal indicators and corresponding attitude data for subsequent posture correction for specific areas. The system comprehensively judges the overall attitude quality. When the alignment error of the overall three-dimensional attitude exceeds 12%, the body symmetry score is lower than 0.7, and the continuity score of the motion sequence is lower than 0.65, and at the same time, when the system fails to correct the error through the adaptive compensation algorithm, it is determined that the overall attitude is abnormal. At this time, the system will generate the attitude abnormal data that details the overall attitude abnormality, including the specific error values, score situations, and specific information on the compensation failure, for further in-depth analysis. Integrate the preliminary attitude matching failure data, characteristic posture mismatch data, and overall attitude abnormal data generated in steps S311, S312, and S313. Through data fusion and comprehensive analysis, form an analysis result report on standard attitude matching, which details the judgment basis, abnormal types, and relevant parameters in the report, providing an accurate basis for subsequent attitude evaluation and correction.

[0050] As an example of the present invention, refer to Figure 3 As shown, in this example, step S4 includes: Step S41: Use the rough registration skeleton model to perform depth contour reconstruction on the multi-angle projection image group of the human body to construct a 3D human body posture image model; map the biomechanical load distribution map to the surface of the 3D human body posture image model to generate load stress distribution modeling data; Step S42: Project the joint range of motion heat map onto the skeleton structure of the 3D human body posture image model to generate a set of dynamic range parameters; perform multi-dimensional evaluation of posture health based on the load stress distribution modeling data and the set of dynamic range parameters to generate a posture abnormality marker layer; Step S43: Identify the posture optimization area based on the posture anomaly marking layer, and combine it with the preset standard posture model to match the personalized correction strategy, generating posture correction control parameters; Apply the posture correction control parameters to the key posture nodes of the 3D human body posture image model to generate a posture correction simulation result; Step S44: Generate a personalized interpretation and evaluation summary of the posture correction simulation result, and finally output a personalized posture correction report.

[0051] In the embodiment of the present invention, by using the previously constructed rough registration skeleton model as the basic framework, depth contour reconstruction is performed on the multi-angle projection image group of the human body. This process combines the contour information of the multi-view projection images and the spatial constraints of the skeleton model to achieve the precise reconstruction of the three-dimensional surface morphology of the human body, thereby constructing a complete 3D human body posture image model. Subsequently, the data of the biomechanical load distribution map is projected onto the surface of this three-dimensional model through spatial mapping technology. Combining the geometric features of the model, the load stress data is converted into surface stress distribution modeling data, forming a three-dimensional posture model containing mechanical information. Project the heat distribution information in the joint range of motion heat map onto the skeleton structure of the 3D human body posture image model to form a dynamic range of motion parameter set. This parameter set specifically represents the range of motion and the restricted range of each key joint in three-dimensional space. Then, the system combines the load stress distribution modeling data and the dynamic range of motion parameter set, and uses a multi-dimensional evaluation algorithm to comprehensively analyze the health status of the human body posture, identify the abnormal areas and potential risks in the posture, and then generate a posture anomaly marking layer to provide visual assistance for subsequent posture optimization. Based on the generated posture anomaly marking layer, the system automatically identifies the posture areas that need to be optimized. By comparing and matching with the preset standard posture model and combining the individual differences of the user, a personalized posture correction strategy is formulated. The system calculates and generates posture correction control parameters, and the specific parameters include the adjustment direction and amplitude of the key posture nodes, etc. Subsequently, these control parameters are applied to the key posture nodes of the 3D human body posture image model to simulate the posture effect after correction, generating a posture correction simulation result to demonstrate and verify the feasibility of the correction plan. Conduct a detailed analysis of the posture correction simulation result, and combine the initial posture data and health assessment results of the individual to generate a personalized interpretation text and an evaluation summary. The content of this report covers the description of the correction effect, the key adjustment parts, potential health improvement suggestions, etc., forming a final personalized posture correction report for the user and relevant health management personnel to reference and apply.

[0052] In this specification, a health assessment system based on a 3D human body posture image model is provided for performing the above-mentioned health assessment method based on a 3D human body posture image model. The health assessment system based on a 3D human body posture image model includes: A point cloud projection module for obtaining human body circular scan data; extracting dynamic pose sequences of the upright position, forward flexion position, and lateral bending position of the human body circular scan data, and performing phase-shifted fringe projection on the human body circular scan data according to the dynamic pose sequences to obtain three-dimensional human body point cloud data; A pose analysis module for performing orthogonal projection on the sagittal plane, coronal plane, and transverse plane of the three-dimensional human body point cloud data to generate a group of multi-angle projection images of the human body; identifying the bone contour features of the group of multi-angle projection images of the human body to establish a dynamic marking system for anatomical landmark points; using the dynamic marking system to perform spatial pose calibration on the bone contour features to generate spatial pose calibration parameters; A pose deviation calculation module for inputting the spatial pose calibration parameters into a preset healthy pose evaluation database for multi-dimensional comparative analysis, and calculating the deviation index between the parts in the three-dimensional human body point cloud data and the standard value according to the analysis results to construct a biomechanical load distribution map and a joint range of motion heat map; A health evaluation module for constructing a 3D human body pose image model based on the group of multi-angle projection images of the human body; performing pose health evaluation on the 3D human body pose image model respectively through the biomechanical load distribution map and the joint range of motion heat map, and performing personalized pose correction according to the evaluation results to generate a personalized pose correction report.

[0053] The beneficial effects of the present invention are as follows: By combining circular scanning with phase-shifted fringe projection technology, the dynamic pose sequences of the human body in the upright position, forward flexion position, and lateral bending position are accurately captured, effectively improving the integrity and dynamic capture ability of the three-dimensional point cloud data, and providing a solid data foundation for subsequent pose analysis. By performing orthogonal projection on the sagittal plane, coronal plane, and transverse plane of the three-dimensional point cloud data to generate a group of multi-angle projection images, and combining bone contour feature recognition with a dynamic marking system for anatomical landmark points, accurate spatial calibration of the human body bone pose is realized, effectively improving the spatial accuracy and dynamic consistency of pose analysis. Using the spatial pose calibration parameters to match and multi-dimensionally compare with the healthy pose database, accurately calculating the deviation index of each part of the skeleton, and constructing a biomechanical load distribution map and a joint range of motion heat map, enhancing the quantitative evaluation ability of pose abnormalities and the understanding of structural mechanics. Based on the group of multi-angle projection images, a three-dimensional human body pose model is constructed, integrating load distribution and range of motion information, realizing comprehensive health evaluation and accurate correction of individual poses, and generating a personalized pose correction report. Each module works together to form a complete closed loop from human body dynamic data acquisition, spatial pose analysis to health evaluation and personalized intervention, greatly improving the automation, intelligence level and practical value of human body pose health management. Therefore, the present invention realizes the precision, standardization and personalization of 3D human body pose health evaluation through high-precision dynamic acquisition and multi-dimensional spatial analysis, improving the accuracy and comprehensiveness of human body pose health evaluation.

[0054] Therefore, in any aspect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0055] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A health assessment method based on a 3D human body pose image model, characterized in that, It includes the following steps: Step S1: Obtain human body circular scan data; extract the dynamic pose sequences of the upright position, forward flexion position, and lateral bending position of the human body circular scan data, and perform phase-offset fringe projection on the human body circular scan data according to the dynamic pose sequences to obtain human body three-dimensional point cloud data; Step S2: Analyze the sagittal plane, coronal plane, and transverse plane of the human body three-dimensional point cloud data for orthogonal projection to generate a group of human body multi-angle projection images; Identify the bone contour features of the human body multi-angle projection image group to establish a dynamic marking system for anatomical landmark points; use the dynamic marking system to calibrate the spatial pose of the bone contour features to generate spatial pose calibration parameters; Step S3: Input the spatial pose calibration parameters into a preset healthy pose assessment database for multi-dimensional comparative analysis, and calculate the deviation index between the parts in the human body three-dimensional point cloud data and the standard value according to the analysis results to construct a biomechanical load distribution map and a joint range of motion heat map; Step S4: Construct a 3D human body pose image model based on the human body multi-angle projection image group; perform pose health assessment on the 3D human body pose image model respectively through the biomechanical load distribution map and the joint range of motion heat map, and perform personalized pose correction according to the assessment results, thereby generating a personalized pose correction report.

2. The health assessment method based on the 3D human body pose image model according to claim 1, wherein, Step S1 includes the following steps: Step S11: Use a scanning device to perform a 360° circular scan on the human body to obtain human body circular scan data; Step S12: Extract the scan sequence of the human body circular scan data, and perform inter-frame synchronous registration on the human body circular scan data according to the scan sequence to generate synchronously corrected human body scan frame data; Step S13: Extract the dynamic skeleton points of the human body scan frame data; perform pose annotation on the dynamic skeleton points to generate dynamic pose sequences of the upright position, forward flexion position, and lateral bending position; Step S15: Decompose the spatial pose parameters of the dynamic pose sequence to obtain human body pose distribution control data; perform stripe phase encoding and multi-phase offset fringe map generation on the human body scan frame data to obtain a multi-phase offset encoding map set; Step S16: Perform phase unwrapping on the multi-phase offset encoding map set through the human body pose distribution control data to generate a dense phase mapping map; perform triangulation and depth inversion calculation on the dense phase mapping map to generate high-precision human body three-dimensional point cloud data.

3. The health assessment method based on the 3D human body pose image model according to claim 2, wherein, The pose annotation of the dynamic skeleton points in Step S13 includes: Collect the three-dimensional coordinates of the skeleton points of the dynamic skeleton points through an optical motion capture system to obtain the three-dimensional coordinates of the skeleton points, where the sampling frequency is set to ≥30 frames / second, the spatial accuracy is ±2 mm, the coordinate unit is millimeters, and the right hand coordinate system is used as a reference; Define the upright position recognition standard: the overall inclination angle of the spine ≤5°, the height difference between the left and right shoulders ≤10 mm, and the knee flexion angle ≤10°; Define the forward flexion position recognition standard: the overall forward inclination angle of the spine ≥30° and ≤90°, the curvature radius from the neck to the lumbar spine ≤400 mm, and the hip flexion angle ≥45°; Define the lateral bending position recognition standard: the lateral bending angle of the spine ≥20° and ≤60°, the height difference between the left and right hips ≥15 mm, and the horizontal distance of the shoulder-hip connection line deviating from the central axis ≥30 mm; Based on the defined recognition criteria for the upright position, forward flexion position, and lateral curvature position, the three-dimensional coordinates of the skeletal points are classified and labeled for posture, thereby obtaining the dynamic posture sequences of the upright position, forward flexion position, and lateral curvature position.

4. The health assessment method based on the 3D human body pose image model according to claim 1, characterized in that, In step S2, the orthogonal projections of the sagittal plane, coronal plane, and cross-section of the human three-dimensional point cloud data include: Normalize the center of gravity of the spatial coordinates of the human three-dimensional point cloud data to generate standardized three-dimensional posture point cloud data; Extract the local principal axis of the standardized three-dimensional posture point cloud data, and use the local principal axis to calibrate the body axis vector field of the three-dimensional posture point cloud data to generate structure-aligned vector calibration data; Perform depth mapping rearrangement on the structure-aligned vector calibration data in the sagittal plane direction to generate a sagittal plane orthogonal projection map; Perform depth value aggregation and compression on the structure-aligned vector calibration data in the coronal plane direction to generate a coronal plane orthogonal projection map; Perform hierarchical slicing and resampling on the structure-aligned vector calibration data in the cross-section direction to generate a cross-section orthogonal projection map; Perform image group registration and angle label binding on the sagittal plane, coronal plane, and cross-section orthogonal projection maps to generate a human multi-angle projection image group.

5. The health assessment method based on the 3D human body pose image model according to claim 1, wherein, In step S2, identify the bone contour features of the human multi-angle projection image group to establish a dynamic marking system for anatomical landmark points, including: Identify the main bone contour of the human multi-angle projection image group; Analyze the connected regions of the human multi-angle projection image group according to the main bone contour to obtain a primary skeleton distribution map; Perform key morphological turning point detection on the primary skeleton distribution map to generate candidate anatomical structure corner point data, and perform spatial consistency verification based on depth prior on the candidate anatomical structure corner point data to generate a high-confidence anatomical landmark point set; Perform temporal posture dynamic matching and trajectory interpolation on the high-confidence anatomical landmark point set to generate dynamic annotation sequence data; Perform posture angle change modeling and structural stability evaluation on the high-confidence anatomical landmark point set through the dynamic annotation sequence data to generate a dynamic marking system for anatomical landmark points.

6. The health assessment method based on the 3D human body pose image model according to claim 1, wherein In step S2, use the dynamic marking system to perform spatial posture calibration on the bone contour features, including: Perform spatial center of gravity registration on the anatomical landmark point sequence in the dynamic marking system to generate initial alignment reference data for posture; Perform three-dimensional rigid registration on the bone contour features, and use the initial alignment reference data for posture to generate a roughly registered skeleton model for the rigidly registered bone contour features; Perform landmark-based non-rigid fine-tuning and optimization on the roughly registered skeleton model to generate a high-precision posture alignment model; Extract joint angles and solve the posture transformation matrix for the high-precision posture alignment model to generate a set of local posture transformation parameters; Perform global consistency reconciliation and error convergence control on the set of local posture transformation parameters to generate spatial posture calibration parameters.

7. The health assessment method based on the 3D human body pose image model according to claim 1, wherein Step S3 includes the following steps: Step S31: Input the spatial posture calibration parameters into a preset healthy posture evaluation database for posture similarity matching to generate an analysis result of standard posture matching; Step S32: Calculate the three-dimensional deviation vectors of each skeleton joint node according to the standard posture matching result data to generate a joint deviation index matrix; Step S33: Analyze the voxel density of the human body three-dimensional point cloud data, and perform mechanical structure mapping in combination with the joint deviation index matrix to generate a biomechanical load distribution map; Step S34: Perform angle interval analysis and motion amplitude reconstruction based on the joint deviation index matrix to generate a joint range of motion heat map.

8. The health assessment method based on the 3D human body pose image model according to claim 7, characterized in that, Step S31 includes the following steps: Step S311: When any of the following conditions are met for the pose calibration parameters, it is determined that the preliminary matching fails and preliminary pose matching failure data is generated: the displacement error of the pose key nodes exceeds the set threshold of 10 cm; the joint angle deviation exceeds 15°; the overall pose stability score is lower than 0.6; Step S312: When the following conditions are met simultaneously, it is determined that the feature region pose mismatch occurs and the feature pose mismatch data is obtained: the angle coordination between the core joint groups decreases by more than 25%; the motion trajectory deviation exceeds the standard range of the historical healthy model; the matching residuals are greater than the set upper limit of 0.08 in three consecutive evaluations, where the core joint groups include the shoulders, spine, and pelvis; Step S313: When all of the following situations occur, it is determined that the overall pose is abnormal and the pose abnormality data is obtained: the overall three-dimensional pose alignment error exceeds 12%; the body symmetry score is lower than 0.7; the motion sequence continuity score is lower than 0.65; and the error cannot be corrected by the adaptive compensation algorithm during the matching process; Step S314: Integrate the preliminary pose matching failure data, the feature pose mismatch data, and the pose abnormality data to generate the analysis results of the standard pose matching.

9. The health assessment method based on the 3D human body pose image model according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Use the rough registration skeleton model to perform depth contour reconstruction on the multi-angle projection image group of the human body to construct a 3D human pose image model; map the biomechanical load distribution map to the surface of the 3D human pose image model to generate load stress distribution modeling data; Step S42: Project the joint range of motion heat map onto the skeleton structure of the 3D human pose image model to generate a dynamic range of motion parameter set; perform multi-dimensional assessment of pose health based on the load stress distribution modeling data and the dynamic range of motion parameter set to generate a pose abnormality marker layer; Step S43: Identify the pose optimization region based on the pose abnormality marker layer, and perform personalized correction strategy matching in combination with the preset standard pose model to generate pose correction control parameters; apply the pose correction control parameters to the key pose nodes of the 3D human pose image model to generate a pose correction simulation result; Step S44: Generate a personalized interpretation and evaluation summary of the pose correction simulation result, and finally output a personalized pose correction report.

10. A health assessment system based on a 3D human body pose image model, characterized in that, A health assessment system based on a 3D human pose image model for performing the health assessment method as described in claim 1, the health assessment system based on a 3D human pose image model includes: A point cloud projection module for acquiring human body circular scan data; extracting the dynamic pose sequences of the upright position, forward flexion position, and lateral flexion position of the human body circular scan data, and performing phase shift fringe projection on the human body circular scan data according to the dynamic pose sequences to obtain human body three-dimensional point cloud data; The posture analysis module is used to perform orthogonal projection on the sagittal plane, coronal plane and transverse plane of the human body three-dimensional point cloud data to generate a group of human body multi-angle projection images; identify the skeletal contour features of the human body multi-angle projection image group to establish a dynamic marking system for anatomical landmark points; use the dynamic marking system to perform spatial posture calibration on the skeletal contour features to generate spatial posture calibration parameters; The posture deviation calculation module is used to input the spatial posture calibration parameters into a preset healthy posture evaluation database for multi-dimensional comparative analysis, and calculate the deviation index between the parts in the human body three-dimensional point cloud data and the standard value according to the analysis results to construct a biomechanical load distribution map and a joint range of motion heat map; The health assessment module is used to construct a 3D human body posture image model based on the human body multi-angle projection image group; perform posture health assessment on the 3D human body posture image model respectively through the biomechanical load distribution map and the joint range of motion heat map, and perform personalized posture correction according to the assessment results, thereby generating a personalized posture correction report.

Images