Physiological mechanism driving and data fusion method and system for sitting posture human body modeling

Abstract

The present application provides a physiological mechanism driving and data fusion method and system for seated human body modeling, which relates to the cross technical field of digital twin and biomechanics. The method comprises: mapping human macroscopic parameters to biomechanical attribute parameters; generating a basic three-dimensional human body grid based on human macroscopic parameters; generating a deformed three-dimensional human body grid by combining the basic three-dimensional human body grid and the human macroscopic parameters; solving the optimal posture based on the vehicle geometric parameters, the task scene identifier and the biomechanical attribute parameters; finally integrating the deformed three-dimensional human body grid and the optimal posture to generate a high-fidelity seated digital twin, realizing end-to-end automatic modeling from macroscopic parameters to digital twin, and solving the core problems of shape and posture mismatch, insufficient population coverage and poor task adaptability.

Description

Technical Field

[0001] This invention relates to the field of digital twin and biomechanics interdisciplinary technology, and in particular to a physiological mechanism-driven and data fusion method and system for seated human body modeling. Background Technology

[0002] In the field of vehicle design and evaluation, digital human models are key tools for visual field analysis, accessibility assessment, and comfort prediction. However, existing technologies have significant limitations: First, posture prediction models and body shape models are usually independent, resulting in a mismatch between the generated virtual human form and posture, making it difficult to apply to detailed simulations such as seat contact pressure distribution; second, mainstream statistical body shape models are mostly built based on standing postures or data from younger populations, failing to accurately represent the unique body shapes of older individuals, especially those in supported sitting positions due to muscle relaxation and spinal morphology changes; finally, traditional posture models are mostly designed for single driving tasks, lacking adaptability to diverse operational scenarios. These shortcomings severely restrict the accuracy and value of digital models in high-end industrial applications. Existing solutions, such as marker-based motion capture systems, are poorly portable and prone to errors, while some parametric models ignore the biomechanical rationality of sitting postures.

[0003] Therefore, the industry urgently needs a solution that can automatically and accurately generate a sitting posture digital twin that matches specific driver physiological characteristics, vehicle geometry, and task scenarios. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, the first aspect of this invention proposes a physiological mechanism-driven and data fusion method for seated human body modeling, comprising:

[0005] S1: Based on the input macroscopic human body parameters, the physiological mechanism-driven model is used to process the parameters and generate biomechanical property parameters including equivalent muscle activation threshold, spinal segment bending stiffness coefficient and biomechanical posture type probability vector.

[0006] S2: Based on macroscopic human body parameters, a parametric morphological generation model is used to generate a basic three-dimensional human body mesh that matches the user's physiological characteristics.

[0007] S3: Based on the basic 3D human body mesh and macroscopic human body parameters, a deformable 3D human body mesh integrating the nonlinear compression deformation characteristics of soft tissue under sitting posture is generated through processing by a soft tissue deformation simulation model.

[0008] S4: Based on vehicle geometric parameters, task scenario identifiers, and biomechanical attribute parameters, the optimal posture that matches the user's physiological characteristics, vehicle interior geometry, and dynamic task scenario is obtained through processing by a task adaptive posture prediction model.

[0009] S5: Based on the deformed three-dimensional human body grid and the optimal posture, a high-fidelity sitting posture digital twin is generated through a system coupling model.

[0010] Compared with the prior art, the present application has the following advantages:

[0011] The present application effectively solves the core problems of morphology and posture mismatch, incomplete crowd representation and poor task adaptability in the background art. First, the physiological mechanism driven model generates biomechanical attribute parameters including equivalent muscle activation threshold, spine segment bending stiffness coefficient and biomechanical posture type probability vector based on the input human macroscopic parameters. These parameters reflect the physiological characteristics of individuals and provide personalized driving basis for subsequent morphology generation and posture prediction, ensuring that the model can adapt to users of different ages and body types and overcoming the defects of insufficient crowd coverage of traditional models. The parameterized morphology generation model generates a basic three-dimensional human body grid that matches the physiological characteristics of the user based on human macroscopic parameters, establishing an accurate initial morphology and laying a foundation for subsequent deformation simulation. The soft tissue deformation simulation model generates a deformed three-dimensional human body grid that integrates the nonlinear compression deformation characteristics of soft tissue in a sitting posture based on the basic three-dimensional human body grid and human macroscopic parameters, accurately simulating the physical deformation of the hips and thighs in a sitting posture and solving the problem of traditional models that cannot truly represent the supporting sitting posture. The task-adaptive posture prediction model obtains an optimal posture that matches the physiological characteristics of the user, the geometry of the vehicle interior, and the dynamic task scene based on the vehicle geometry parameters, task scene identifiers, and biomechanical attribute parameters, and through comprehensive consideration of the vehicle environment and multiple task scenarios, it realizes adaptive optimization for diversified tasks, making up for the shortcomings of traditional posture models that are designed for single tasks. Finally, the system coupling model generates a high-fidelity sitting posture digital twin based on the deformed three-dimensional human body grid and the optimal posture, realizing the dynamic unification of morphology and posture and fundamentally eliminating the disadvantages of morphology and posture mismatch.

[0012] Throughout the entire process, the steps are closely linked: the biomechanical attribute parameters output by the physiological mechanism driven model directly drive the personalized processing of the parameterized morphology generation model and the task-adaptive posture prediction model, ensuring that both the morphology and the posture are based on the same physiological basis; the parameterized morphology generation model and the soft tissue deformation simulation model jointly construct the real morphology, the task-adaptive posture prediction model optimizes the posture, and the system coupling model finally fuses the morphology and the posture, forming an end-to-end automated modeling, significantly improving the accuracy and practicality of the sitting posture modeling. BRIEF DESCRIPTION OF DRAWINGS

[0013] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0014] Figure 1 The diagram shown is a flowchart of a physiological mechanism-driven and data fusion method for seated human body modeling provided by an embodiment of the present invention.

[0015] Figure 2 The diagram shown is a deep neural network structure diagram of a physiological mechanism-driven model provided in an embodiment of the present invention.

[0016] Figure 3 The diagram shown is a structural schematic of a physiological mechanism-driven and data fusion system for seated human body modeling provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0018] The specific embodiments of the present invention will be described below.

[0019] Example 1

[0020] like Figure 1 As shown, this invention proposes a physiological mechanism-driven and data fusion method for seated human body modeling, including:

[0021] S1: Based on the input macroscopic human body parameters, the physiological mechanism-driven model is used to process the parameters and generate biomechanical property parameters including equivalent muscle activation threshold, spinal segment bending stiffness coefficient and biomechanical posture type probability vector.

[0022] S2: Based on macroscopic human body parameters, a parametric morphological generation model is used to generate a basic three-dimensional human body mesh that matches the user's physiological characteristics.

[0023] S3: Based on the basic 3D human body mesh and macroscopic human body parameters, a deformable 3D human body mesh integrating the nonlinear compression deformation characteristics of soft tissue under sitting posture is generated through processing by a soft tissue deformation simulation model.

[0024] S4: Based on the vehicle geometry parameters, task scene identifiers and biomechanical attribute parameters, the task adaptive posture prediction model is processed to obtain the optimal posture matched with the user's physiological characteristics, vehicle interior geometry and dynamic task scene;

[0025] Among them, "vehicle geometry parameters" refer to the geometric size parameters of the vehicle interior, including seat height, seat inclination, steering wheel position, pedal position, instrument panel distance, etc., which are used to define the sitting posture environment constraints.

[0026] "Task scene identifier" refers to a discrete variable representing different work scenes, for example: 0 represents driving task, 1 represents rest task, and 2 represents operating center console task. Each identifier corresponds to a set of posture requirements (such as hand position, line of sight angle), which is used for adaptive posture optimization.

[0027] S5: Based on the deformed three-dimensional human body grid and the optimal posture, the system coupling model is processed to generate a high-fidelity sitting posture digital twin.

[0028] Specifically, the physiological mechanism driven and data fusion method for sitting posture human body modeling realizes end-to-end automatic modeling through five core steps, each step undertakes a specific function in the technical process and connects with each other. The physiological mechanism driven model processes based on the input human macroscopic parameters to generate biomechanical attribute parameters including equivalent muscle activation threshold, spine segment bending stiffness coefficient and biomechanical posture type probability vector. These parameters extract the internal physiological characteristics from the easily obtained macroscopic features through nonlinear mapping, providing personalized driving basis for subsequent modules. The parameterized shape generation model processes based on the same human macroscopic parameters to generate a basic three-dimensional human body grid matched with the user's physiological characteristics. This grid is constructed from large-scale sitting posture data through statistical learning, ensuring that the initial shape accurately reflects individual anatomy. The soft tissue deformation simulation model processes based on the basic three-dimensional human body grid and human macroscopic parameters to generate a deformed three-dimensional human body grid integrating the nonlinear compression deformation characteristics of soft tissue under sitting posture. The real deformation behavior of the hips and thighs under load is simulated through physical simulation. The task adaptive posture prediction model processes based on the vehicle geometry parameters, task scene identifiers and biomechanical attribute parameters to obtain the optimal posture matched with the user's physiological characteristics, vehicle interior geometry and dynamic task scene. This step considers discrete task selection and continuous joint adjustment through optimization algorithm. The system coupling model processes based on the deformed three-dimensional human body grid and the optimal posture to generate a high-fidelity sitting posture digital twin, realizing dynamic fusion and visual output of shape and posture.

[0029] The synergy of these steps is reflected in data flow and functional dependence. The physiological mechanism-driven model directly inputs the biomechanical attribute parameters output by the model into the task-adaptive posture prediction model, ensuring that posture optimization is based on individual physiological constraints. At the same time, the parameterized morphology generation model and the soft tissue deformation simulation model successively process the morphology construction, adding physical reality to the base mesh. The optimal posture generated by the task-adaptive posture prediction model is combined with the deformed mesh through the system coupling model to form the final digital twin. This pipeline processing eliminates the problem of disconnection between morphology and posture in traditional methods, as all modules share the same set of physiological input data, and the outputs are dynamically integrated in the system coupling model. Technically, this method can automatically generate highly personalized sitting posture models, adapting to the needs of users of different ages and body types, and solving the limitation of insufficient population coverage. By integrating task scenarios for optimization, it supports diversified work simulation and improves adaptability. Automation of the entire process reduces human intervention and improves modeling efficiency and consistency. The physiological mechanism-driven model captures the internal relationship between macroscopic parameters and biomechanical attributes, providing a scientific basis for morphology and posture. The parameterized morphology generation model ensures the reasonableness of the morphology using statistical priors. The soft tissue deformation simulation model introduces physical precision. The task-adaptive posture prediction model balances comfort and task requirements through multi-objective optimization. The system coupling model maintains the unity of morphology and posture through dynamic feedback. Ultimately, this method can provide more accurate comfort evaluation and performance prediction in vehicle design and other applications, reducing reliance on physical prototypes.

[0030] This method achieves overall performance improvement through seamless connection between steps. The physiological mechanism-driven model lays the foundation for individualization for subsequent modules. The parameterized morphology generation model establishes an accurate initial morphology. The soft tissue deformation simulation model enhances physical reality. The task-adaptive posture prediction model optimizes dynamic posture. The system coupling model ensures the consistency of the final output. This integrated processing makes the digital twin not only visually realistic but also behaviorally consistent with biomechanical principles, providing a reliable simulation tool for industrial design. The technical effect is reflected in the end-to-end automated process. Users only need to input macroscopic parameters to obtain a high-fidelity model, greatly simplifying traditional multi-step manual modeling. At the same time, the model can dynamically adapt to different scenarios, improving practicality and scope of application. In summary, the method solves the core challenge of sitting posture human modeling through multi-module collaboration, providing an efficient solution for digital twin applications.

[0031] The building process of the above model is described in detail below.

[0032] 1. Building process of the physiological mechanism-driven model.

[0033] Physiomechanical-driven model is used to map human macroscopic parameters (e.g., gender, age, height, body mass index, sitting height to standing height ratio) to biomechanical attribute parameters (equivalent muscle activation threshold, spinal segment flexion stiffness coefficient, biomechanical posture type probability vector). The construction process is as follows:

[0034] Data preparation: Collect a large amount of labeled data, including human macroscopic parameters and corresponding biomechanical attribute parameters. Biomechanical attribute parameters can be obtained through biomechanical experiments, such as using electromyography (EMG) to measure muscle activation threshold, spinal flexion test to obtain stiffness coefficient, and posture analysis to obtain posture type probability.

[0035] Network structure design: Use deep feedforward neural network (DFFNN). The number of input layer nodes corresponds to the number of macroscopic parameters (e.g., 5 parameters), and standardization processing (e.g., Z-score normalization) is performed. The hidden layer includes at least two fully connected layers, the first fully connected layer uses ReLU activation function to introduce nonlinearity, followed by a dropout layer (Dropout) for regularization (dropout rate is set to 0.2-0.5), and the second fully connected layer further extracts features. The number of output layer nodes corresponds to the number of biomechanical attribute parameters (e.g., 3 parameters), and linear activation function is used for regression task.

[0036] Training process: Use backpropagation algorithm and optimizer (e.g., Adam), and use mean square error (MSE) as loss function. Training data is divided into training set and validation set, and hyperparameters (e.g., number of layers, number of nodes) are adjusted through cross-validation to prevent overfitting. After model training is completed, it can be deployed as a static module to receive macroscopic parameters and directly output biomechanical attribute parameters.

[0037] 2. Parameterized morphological generation model construction process.

[0038] Parameterized morphological generation model generates a basic three-dimensional human mesh based on statistical shape model (SSM). The construction process is as follows:

[0039] Statistical shape model construction: Collect large-scale sitting posture human three-dimensional scanning data (e.g., from public data sets or custom scanning), and perform alignment and normalization processing on the mesh. Use principal component analysis (PCA) for dimensionality reduction to obtain the average mesh and principal components (eigenvectors), which represent the main directions of shape variation.

[0040] Multiple linear regression model training: Pre-train a multiple linear regression model, input human macroscopic parameters (e.g., gender, age, height, body mass index), output SSM principal component coefficient vector. Training data includes macroscopic parameters and corresponding principal component coefficients (obtained by projecting into SSM space). The regression model uses least squares method to fit, ensuring the accuracy of the coefficient prediction.

[0041] Mesh generation and post-processing: Reconstruct the personalized three-dimensional mesh using the predicted principal component coefficients and the linear combination of SSM. Subsequently, perform topology check and repair, detect and repair non-manifold edges, holes, and self-intersections using mesh processing tools (such as MeshLab or CGAL library), and ensure that the mesh meets the simulation requirements.

[0042] 3. Soft tissue deformation simulation model building process.

[0043] The soft tissue deformation simulation model is used to simulate the nonlinear compression deformation of the soft tissue in the hip and thigh area under sitting posture. Its building process is as follows:

[0044] ① Finite element discretization: Based on the basic three-dimensional human body mesh, use finite element mesh generation tools (such as Gmsh or Abaqus) to divide the hip and thigh area into tetrahedral elements. The grid density is adjusted according to the contact area to ensure the calculation accuracy.

[0045] ② Material parameter regression model building:

[0046] The first regression model: input body mass index and body fat rate, output shear modulus parameter of hyperelastic material model. The training data comes from biomechanical experiments, such as obtaining the shear modulus of soft tissue of different individuals through indentation test, and using linear regression to establish the mapping.

[0047] The second regression model: input body mass index and body fat rate, output hardening index parameter. Similarly, based on experimental data (such as uniaxial compression test), train the regression model to capture the nonlinear hardening behavior. The first regression model and the second regression model both use simple linear regression or polynomial regression, and the specific form is determined according to the data distribution.

[0048] ③ Nonlinear finite element solving: integrate personalized material parameters into finite element solver (such as FEBio or Abaqus), set hyperelastic material model (such as Neo-Hookean model), apply boundary conditions (such as seat contact surface fixed), perform quasi-static analysis to solve displacement field, and finally output the deformed three-dimensional mesh.

[0049] 4. Task-adaptive posture prediction model building process.

[0050] The task-adaptive posture prediction model is used to generate the optimal posture that matches the user's physiological characteristics, vehicle environment, and task scenario. Its building process is as follows:

[0051] Lookup table construction: based on biomechanics research literature or experimental data, predefine the mapping relationship between biomechanics posture types (such as "relaxed", "forward leaning", "backward leaning") and neutral posture joint angles. For example, collect joint angle data of different postures through motion capture system, and construct lookup table.

[0052] Mixed-integer programming model design: The objective function includes a pose deviation penalty term (quantifying the difference between the current pose and the neutral pose, e.g., joint angle Euclidean distance) and a muscle activation penalty term (estimating muscle effort based on biomechanical property parameters, e.g., using a muscle model). Constraints include joint range of motion, vehicle geometry limitations (e.g., seat position, steering wheel distance). The model is formalized using optimization modeling tools (e.g., Pyomo or Gurobi).

[0053] Optimization solver configuration: Branch-and-bound algorithm is used to handle integer variables (e.g., task selection), and parallel computing architecture (e.g., GPU acceleration) is used to evaluate candidate solutions. The optimal solution selection strategy is based on minimizing the objective function value, ensuring solution efficiency and global optimality.

[0054] 5. System coupling model building process.

[0055] The system coupling model is used to integrate the deformed three-dimensional human mesh and the optimal pose to generate a high-fidelity digital twin of the sitting posture. The building process is as follows:

[0056] Skeletal skinning technology implementation: Bind the skeletal system (e.g., use joint definition) on the deformed three-dimensional mesh, and use the linear blend skinning (LBS) algorithm to convert the joint rotation of the optimal pose to the mesh vertex displacement. The weight is determined by hand-drawing or automatic calculation (e.g., heat diffusion method).

[0057] Real-time rendering processing: Import the mesh adjusted by the pose using a physics engine (e.g., Unity or Unreal Engine), configure lighting, materials, and textures, and realize visual rendering. The engine supports real-time interaction and perspective switching.

[0058] Dynamic feedback mechanism: By listening to the pose change event, trigger the soft tissue deformation recalculation (e.g., call the finite element solver), and update the mesh shape. The feedback loop ensures that the digital twin maintains physical consistency when the pose changes.

[0059] In some implementations, S1 includes:

[0060] S1.1: Based on the human macroscopic parameters including gender, age, height, body mass index, and sitting height to height ratio, process through the input layer of the deep feedforward neural network model to generate standardized input features;

[0061] S1.2: Based on the standardized input features, perform nonlinear transformation processing through multiple hidden layers of the deep feedforward neural network model to generate high-level feature representations;

[0062] S1.3: Based on the high-level feature representation, process through the output layer of the deep feedforward neural network model to generate biomechanical property parameters.

[0063] In the physiological mechanism-driven and data fusion method, the S1 step is refined by a deep feedforward neural network model to achieve high-precision conversion from macro parameters to biomechanical properties. The input layer of the deep feedforward neural network model processes based on human macro parameters including gender, age, height, body mass index, and sitting height to height ratio to generate standardized input features. This process ensures data consistency through feature scaling and normalization, providing a stable foundation for subsequent network calculations. The multiple hidden layers of the deep feedforward neural network model process based on standardized input features to generate high-level feature representations. These hidden layers introduce non-linear relationships through activation functions, gradually abstracting complex interaction effects in input parameters to capture the effects of age and body composition on biomechanical properties. The output layer of the deep feedforward neural network model processes based on high-level feature representations to generate biomechanical property parameters, including equivalent muscle activation threshold, spinal segment bending stiffness coefficient, and biomechanical posture type probability vector. The output layer uses a linear activation function to directly regress target values, ensuring the continuity and physical meaning of the prediction results.

[0064] This hierarchical processing enables the network to learn complex patterns from simple features. The input layer processes raw parameters to reduce noise, the hidden layer extracts key features through multiple transformations, and the output layer finally maps to biomechanical properties. Technically, this design improves the accuracy and generalization ability of the prediction. Standardized input features avoid training instability caused by data scale differences, multiple hidden layers capture deep relationships between macro parameters and biomechanical properties, and direct regression in the output layer ensures parameter interpretability. Standardization in the input layer ensures network training convergence, nonlinear activation functions in the hidden layer simulate complex relationships in biomechanics, and linear output in the output layer conforms to the numerical characteristics of biomechanical parameters. The entire network structure learns end-to-end, avoiding the shortcomings of manual feature engineering in traditional methods, allowing the model to automatically extract relevant rules from data. In addition, this processing enhances the model's adaptability to diverse populations, as input parameters cover key physiological dimensions. By learning the combined effects of these dimensions, the network can accurately predict the biomechanical properties of different individuals.

[0065] The application of the deep feedforward neural network model makes the physiological mechanism-driven more reliable. The input layer ensures data consistency, the hidden layer improves feature expression ability through multi-level processing, and the output layer directly generates practical parameters. This processing not only simplifies the data preprocessing process, but also improves the robustness of the model, so that it can still maintain stable prediction when facing noise or missing data. The technical effect is reflected in that the generation of biomechanical attribute parameters is more accurate, the equivalent muscle activation threshold can reflect the individual muscle function difference, the spinal segment bending stiffness coefficient quantifies the spinal mechanical properties, and the biomechanical posture type probability vector provides a classification basis for posture preference. In principle, the network learns from macro to micro through training, and uses the statistical law in big data, so it can also perform well on unknown data. Finally, this step provides a solid driving core for the entire method, ensuring that the subsequent morphology and posture modules are based on scientific and reasonable physiological inputs, improving the authenticity and practicality of the digital twin. Through this automated process, the method reduces the dependence on external biomechanical measurements, making the modeling process more efficient and scalable.

[0066] Reference Figure 2 , Figure 2 A deep feedforward neural network structure is shown, including an input layer, two hidden layers (i.e. fully connected layers), and an output layer. The input layer receives standardized human macro parameters (such as gender, age, height, body mass index, and sitting height to height ratio), and the number of nodes corresponds to the number of parameters (e.g. 5 nodes). The fully connected layer directly connected to the input layer processes the input features through a nonlinear activation function (such as ReLU) to generate preliminary feature representations; the fully connected layer directly connected to the output layer further refines the features to generate high-level feature representations. The output layer directly regresses biomechanical attribute parameters, including equivalent muscle activation threshold, spinal segment bending stiffness coefficient, and biomechanical posture type probability vector, and the number of nodes corresponds to the number of output parameters (e.g. 3 nodes). Figure 2 The arrow indicates the data flow from the input layer to the output layer through the fully connected layers, embodying the nonlinear transformation process of features, ensuring end-to-end mapping from human macro parameters to biomechanical attributes.

[0067] In some implementations, S1.2 includes:

[0068] S1.2.1: Based on the standardized input features, process through the first fully connected layer of the deep feedforward neural network model to generate first hidden layer features;

[0069] S1.2.2: Based on the first hidden layer features, perform regularization processing through the dropout layer to generate hidden layer features resistant to overfitting;

[0070] S1.2.3: Based on the anti-overfitting hidden layer features, the second fully connected layer of the deep feedforward neural network model is processed to generate high-level feature representation.

[0071] In the refinement of step S1, the deep feedforward neural network model further optimizes the feature extraction process through specific hidden layers processing, ensuring the stability and accuracy of biomechanical property prediction. The first fully connected layer of the deep feedforward neural network model processes based on standardized input features to generate first hidden layer features, which realizes initial feature transformation through multiple neurons and activation functions, capturing linear and nonlinear relationships between input parameters, providing a rich feature basis for subsequent processing. The dropout layer regularizes based on the first hidden layer features to generate anti-overfitting hidden layer features, which randomly ignores part of the neuron connections during training to reduce the model's over-reliance on training data, thereby improving the generalization ability and preventing performance degradation on unknown data. The second fully connected layer of the deep feedforward neural network model processes based on anti-overfitting hidden layer features to generate high-level feature representation, which further refines and combines features through nonlinear transformation to form more abstract representative vectors, ultimately serving the parameter generation of the output layer.

[0072] This sequential processing ensures the gradual deepening of feature learning, with the first fully connected layer performing basic feature extraction, the dropout layer introducing regularization control complexity, and the second fully connected layer completing high-level feature integration. Technically, this structure improves the robustness and generalization performance of the model, as the first fully connected layer widens the feature space, capturing more potential patterns; the dropout layer reduces the risk of overfitting through randomness, making the model more adaptable to diverse data distribution; and the second fully connected layer strengthens the interaction between features, ensuring that the output feature representation can fully reflect the biomechanical characteristics. The width design of the first fully connected layer allows the network to learn the combined effects of input parameters, the regularization mechanism of the dropout layer simulates the variability of biomechanical data, and the depth processing of the second fully connected layer corresponds to the high-level causal relationships in biomechanical properties. The entire process through hierarchical regularization and feature optimization enables the network to remain stable in complex data environments, avoiding the shortcomings of traditional single-level models.

[0073] The processing makes the biomechanical property prediction more reliable, the first fully connected layer ensures the initial feature diversity, the dropout layer maintains the model simplicity, and the second fully connected layer refines the key information. The technical effects are reflected in that the generated high-level feature representation can accurately drive the subsequent output, the predicted equivalent muscle activation threshold is more in line with the individual physiological state, the estimated spinal segment bending stiffness coefficient is more in line with the mechanical principle, and the classification of the biomechanical posture type probability vector is more accurate. In principle, the multi-level processing simulates the hierarchical structure of the human physiological system, thereby incorporating more biomechanical priors into the prediction. Ultimately, this refinement step enhances the efficiency of the entire physiological mechanism-driven module, providing a more consistent physiological basis for the digital twin of the sitting posture, and supporting more accurate comfort and performance evaluation in practical applications. By avoiding overfitting, the model can still maintain high performance when facing limited or noisy data, thereby expanding the applicability and practicality of the method.

[0074] In some implementations, the parametric morphology generation model includes a statistical shape model, S2 includes:

[0075] S2.1: Based on the human macroscopic parameters, a pre-trained multiple linear regression model is used for processing to generate a principal component coefficient vector of the statistical shape model;

[0076] S2.2: Based on the principal component coefficient vector, a linear combination process is performed on the statistical shape model to generate a personalized basic three-dimensional human body mesh;

[0077] S2.3: Topological inspection and repair processing is performed on the personalized basic three-dimensional human body mesh to generate a basic three-dimensional human body mesh that meets the requirements of the manifold structure.

[0078] In the parametric morphology generation model, the application of the statistical shape model realizes efficient conversion from macroscopic parameters to three-dimensional meshes, ensuring that the generated basic human body mesh not only conforms to individual characteristics but also has anatomical rationality. The parametric morphology generation model includes a statistical shape model, which is constructed based on large-scale sitting posture scanning data. Through principal component analysis, the high-dimensional shape space is compressed into a low-dimensional representation, thereby capturing the morphological variation patterns in the population. In step S2, a pre-trained multiple linear regression model is used to process the human macroscopic parameters to generate a principal component coefficient vector of the statistical shape model. This process maps the macroscopic parameters to the weights of the shape space through regression analysis, so that individual physiological characteristics directly guide mesh generation. The statistical shape model performs linear combination processing based on the principal component coefficient vector to generate a personalized basic three-dimensional human body mesh. This mesh is synthesized by weighted average shape and principal component displacement field, ensuring that the output is consistent with the training data at the vertex level. Topological inspection and repair processing is performed on the personalized basic three-dimensional human body mesh to generate a basic three-dimensional human body mesh that meets the requirements of the manifold structure. This step detects and corrects mesh defects such as holes or self-intersections through algorithms, ensuring the integrity and usability of the mesh.

[0079] This process makes the shape generation automatic and accurate, the statistical shape model provides shape priors, the multiple linear regression realizes the conversion of parameters to coefficients, the linear combination generates the mesh, and the topology repair ensures the quality. Technically, this method can quickly generate a mesh that matches the user's physiological characteristics because the statistical shape model is based on real sitting posture data and accurately reflects the geometric characteristics in the support posture; the multiple linear regression model uses the statistical correlation between macro parameters and shape to simplify the mapping process; the linear combination process maintains the smoothness and continuity of the mesh; and the topology check and repair eliminates potential errors and improves the stability of the mesh in simulation. Principal component analysis of the statistical shape model reduces the data dimension, making the generation process efficient; the multiple linear regression model learns the linear relationship between macro parameters and shape coefficients, ensuring prediction accuracy; the linear combination uses orthogonal basis vectors to maintain shape reasonableness; and the topology processing guarantees mesh availability through computational geometry methods. The entire process combines statistical learning and geometric processing, achieving a balance between generation speed and quality.

[0080] The parametric shape generation model achieves high personalization through the statistical shape model, the pre-trained multiple linear regression model ensures reliable coefficient prediction, the linear combination process generates a visual mesh, and the topology repair enhances practicality. The technical effect is reflected in that the basic three-dimensional human mesh can accurately represent the individual sitting posture, including spinal curvature and limb proportion, providing a realistic basis for subsequent soft tissue deformation; at the same time, automated processing reduces manual modeling time and improves overall efficiency. In principle, this method uses the statistical rules of large-scale data to make the generated mesh not only realistic in appearance but also biologically constrained. Ultimately, this step provides a reliable shape input for the entire digital twin process, supporting more accurate seat interaction simulation and comfort analysis, and reducing the reliance on physical scanning in industrial design. By integrating topology checking, the mesh quality is guaranteed, avoiding computational errors in subsequent simulations, thereby improving the overall robustness and application value of the method.

[0081] The construction process of the multiple linear regression model is as follows.

[0082] The multiple linear regression model is a component of the parametric shape generation model and is used to predict the principal component coefficients of the SSM. Its construction process is as follows:

[0083] Data preparation: Collect human macro parameters and corresponding SSM principal component coefficients (obtained by projecting three-dimensional scan data into the SSM space).

[0084] Model training: Use the multiple linear regression algorithm, which is in the form of coefficients = W x human macro parameters + b, where W is the weight matrix and b is the bias vector. Use the least squares method or gradient descent to optimize the parameters.

[0085] Deployment: After model training, integrate into the system, receive macroscopic parameters directly output principal component coefficients.

[0086] In some implementations, S3 includes:

[0087] S3.1: Based on the basic three-dimensional human body grid, through finite element discretization processing, generate the tetrahedral element grid of the hip and thigh area;

[0088] S3.2: Based on the human macroscopic parameters, through the material parameter regression model for processing, determine the individualized material parameters of the hyperelastic material model;

[0089] S3.3: Based on the tetrahedral element grid and individualized material parameters, through the nonlinear finite element solver for quasi-static analysis processing, generate the deformed three-dimensional human body grid.

[0090] In the soft tissue deformation simulation model, S3 step simulates the physical response of soft tissue in sitting position through finite element analysis, ensuring that the deformed grid truly reflects the morphological changes of individuals under load. Finite element discretization processing is based on the basic three-dimensional human body grid to generate the tetrahedral element grid of the hip and thigh area. This process divides continuous geometry into discrete elements for numerical calculation, and refines the grid in the contact area of the sitting position to capture local deformation. The material parameter regression model processes based on human macroscopic parameters to determine the individualized material parameters of the hyperelastic material model. This model maps physiological characteristics such as body mass index and body fat percentage to material constants through regression analysis, allowing the mechanical properties to be dynamically adjusted according to individual differences. The nonlinear finite element solver performs quasi-static analysis processing based on the tetrahedral element grid and individualized material parameters to generate the deformed three-dimensional human body grid. The solver calculates the displacement field in the equilibrium state through an iterative algorithm and superimposes the results on the basic grid to output the final morphology integrating deformation characteristics.

[0091] This process realizes the physically accurate simulation of soft tissue deformation. Finite element discretization provides a computational framework, material parameter regression individualizes mechanical properties, and nonlinear solver handles large deformation behavior. Technically, this method can accurately predict the compression deformation of the hip and thigh in sitting position, as finite element discretization captures geometric details, material parameter regression ensures that the mechanical properties are consistent with individual physiology, and the nonlinear solver handles the hyperelastic response of soft tissue. Finite element discretization converts the continuous problem into a discrete system through grid division, making the simulation feasible; material parameter regression uses the correlation between physiological data and mechanical tests to make the model parameters real and reliable; the nonlinear solver solves the equilibrium equation to simulate the deformation process under quasi-static load. The entire process combines computational mechanics and physiological data, thereby incorporating individualized characteristics in the simulation.

[0092] The soft tissue deformation simulation model enhances the authenticity of the digital twin through finite element analysis. The finite element discretization ensures that the mesh is suitable for simulation. The material parameter regression model introduces personalized material properties. The nonlinear finite element solver calculates the deformation results. The technical effects are embodied in the following aspects. Firstly, the three-dimensional human mesh after deformation can accurately simulate the contact pressure distribution between the seat and the human body, providing a direct basis for comfort evaluation. Secondly, the personalized material parameters take into account the influence of body fat and body weight, making the deformation more consistent with the principles of biomechanics. In principle, this method describes the nonlinear behavior of soft tissue based on the hyperelastic constitutive relationship, thereby reproducing real physical phenomena in simulation. Finally, this step provides high-fidelity shape output for the system, supports more reliable engineering design decisions, and reduces the need for experimental measurements in applications. By integrating quasi-static analysis, the deformation simulation controls the computational cost while ensuring accuracy, making the method suitable for real-time or near-real-time scenarios, thereby improving overall practicality and efficiency.

[0093] In some implementations, S3.2 includes:

[0094] S3.2.1: Based on the body mass index and body fat rate in the human macroscopic parameters, a first regression model is used to process and generate the shear modulus parameter in the hyperelastic material model parameters;

[0095] S3.2.2: Based on the body mass index and body fat rate in the human macroscopic parameters, a second regression model is used to process and generate the hardening index parameter in the hyperelastic material model parameters;

[0096] S3.2.3: The shear modulus parameter and the hardening index parameter are integrated to generate personalized material parameters.

[0097] In the process of determining the personalized material parameters of the hyperelastic material model, S3.2 uses two independent regression models to process key material constants, ensuring that the mechanical properties of soft tissue are accurately related to individual physiological characteristics. The first regression model processes the body mass index and body fat rate in the human macroscopic parameters to generate the shear modulus parameter in the hyperelastic material model parameters. This model establishes a mathematical relationship between physiological indicators and material shear modulus through multiple linear regression, allowing the initial stiffness of soft tissue to adaptively adjust with changes in body composition. The second regression model processes the body mass index and body fat rate in the human macroscopic parameters to generate the hardening index parameter in the hyperelastic material model parameters. This model also uses regression analysis to capture the influence of body fat distribution on the nonlinear hardening behavior of materials, thereby accurately representing the stress response characteristics of soft tissue under large deformation. Integrating the shear modulus parameter and the hardening index parameter generates personalized material parameters, which combines the outputs of the two regression models into a complete set of material parameters for subsequent constitutive relationship definition in finite element analysis.

[0098] This separation approach allows specialized modeling for different mechanical properties of the material, with the first regression model focusing on the shear modulus that controls the small deformation behavior, and the second regression model responsible for describing the hardening characteristics in the large deformation stage, and the integration process ensures the integrity and consistency of the parameter set. Technically, this method can more accurately simulate the deformation behavior of soft tissue in individuals of different body types, because the shear modulus parameter is directly related to the initial stiffness of the soft tissue, and the hardening index parameter determines the stiffness change rule in the deformation process, and the combination of the two makes the hyperelastic model fully reflect the mechanical response from light compression to large deformation. The first regression model uses the known physiological correlation between body mass index and body fat rate and tissue stiffness, and the second regression model is based on the influence mechanism of fat tissue distribution on strain hardening effect, and the integration step ensures the physical coordination between parameters. Through this hierarchical regression strategy, the material parameters are not only associated with individual characteristics, but also maintain the internal physical consistency of the hyperelastic constitutive relationship.

[0099] The generation of personalized material parameters makes the soft tissue deformation simulation more realistic and reliable, the first regression model accurately predicts the basic stiffness characteristics, the second regression model captures the nonlinear hardening behavior, and the parameter integration ensures the integrity of the constitutive model. This processing significantly improves the accuracy of seated posture pressure distribution prediction, because the material parameters are completely based on individual physiological characteristics rather than group averages, making the simulation able to distinguish the mechanical differences brought by different fat distribution. In principle, this method converts macroscopic human measurement data into microscopic mechanical properties by establishing a direct mapping between physiological parameters and continuum mechanics parameters, realizing the connection of cross-scale modeling. Finally, this step provides highly personalized material input for finite element analysis, enabling the deformation of the three-dimensional human mesh to truly reflect the soft tissue response of a specific individual in a seated position, providing a more reliable simulation basis for seat comfort evaluation and human engineering design. Through this detailed processing, the simulation results are significantly improved in terms of consistency with real human behavior, effectively supporting decision optimization in the product design process.

[0100] In some implementations, S4 includes:

[0101] S4.1: Based on the biomechanical posture type probability vector in the biomechanical attribute parameters, processing through a predefined lookup table to determine the individualized neutral posture joint angles;

[0102] S4.2: Based on the individualized neutral posture joint angles, vehicle geometry parameters, and task scenario identifiers, processing through a mixed integer programming model to establish a target function that includes posture deviation penalty terms and muscle activation penalty terms;

[0103] S4.3: Based on the target function and the constraint conditions, processing through an optimization solver to obtain the optimal posture.

[0104] In the task-adaptive posture prediction process, the S4 step combines biomechanical properties with task requirements through a systematic approach to achieve highly personalized posture optimization. The biomechanical posture type probability vector is processed based on a pre-defined lookup table to determine the individualized neutral posture joint angles. The lookup table establishes a correspondence between posture types and ideal angles of each joint through biomechanical research, so that the dominant type in the probability vector can be directly mapped to a specific joint angle configuration, providing a personalized comfort benchmark for posture optimization. The mixed integer programming model is constructed based on the individualized neutral posture joint angles, vehicle geometry parameters, and task scenario identifiers. The objective function includes a posture deviation penalty term and a muscle activation penalty term. The posture deviation penalty term quantifies the difference between the current posture and the individualized neutral posture, and the muscle activation penalty term estimates the muscle effort required to maintain the posture based on biomechanical property parameters. The two terms are balanced by a weight coefficient to balance comfort and energy consumption requirements.

[0105] This construction process formalizes the posture optimization problem as a structured mathematical programming problem. The lookup table processing converts discrete biomechanical classifications into continuous joint angle references, and the mixed integer programming model unifies continuous posture variables and discrete task selection within the same optimization framework. Technically, this method can generate an optimal posture that meets individual physiological characteristics and adapts to specific task requirements. The individualized neutral posture joint angles provide a comfort benchmark based on biomechanical principles, the posture deviation penalty term drives the optimization result close to this benchmark, and the muscle activation penalty term minimizes muscle load. The comprehensive consideration of the objective function ensures the balance between comfort, functionality, and efficiency. The lookup table is based on a large amount of biomechanical experimental data, ensuring the scientificity of the neutral posture angles. The mixed integer programming model transforms the complex multi-objective optimization problem into a solvable standard form through mathematical formalization, enabling the algorithm to systematically explore the solution space.

[0106] Task-adaptive posture prediction realizes precise personalized posture generation through this structured method, the lookup table provides individualized benchmarks, the mixed integer programming model constructs a complete optimization problem, and the objective function balances multiple optimization goals. This processing significantly improves the practicality of digital twins in diversified scenarios, because the optimal posture obtained by optimization not only respects the individual's physiological preferences, but also meets the functional requirements of specific tasks, while considering the space constraints in the vehicle. In principle, this method encodes biomechanical knowledge as a component of the optimization problem, making the algorithm decision-making process have clear physical meaning and physiological basis. Finally, this step provides a highly adaptive posture output for the seated posture digital twin, supporting accurate comfort prediction and work performance evaluation, effectively simulating the natural posture response of different users in various operation scenarios during product design. Through this systematic optimization framework, posture prediction is no longer dependent on empirical rules, but is based on scientific optimization principles, significantly improving the reliability and consistency of the prediction results.

[0107] In some implementations, S4.3 includes:

[0108] S4.3.1: Based on the objective function and the constraint conditions, processing is performed through a branch and bound algorithm to generate a candidate solution space;

[0109] S4.3.2: Based on the candidate solution space, processing is performed through a parallel computing architecture to accelerate the evaluation of multiple possible solutions;

[0110] S4.3.3: Based on the accelerated evaluation results, an optimal solution selection strategy is adopted to obtain the optimal posture.

[0111] In solving the mixed integer programming problem, the S4.3 step realizes efficient calculation through advanced algorithm combination, ensuring rapid positioning of the optimal posture in complex solution space. The branch and bound algorithm processes based on the objective function and the constraint conditions to generate a candidate solution space. This algorithm systematically enumerates possible integer solution combinations and calculates their boundary values, gradually excluding search areas that are unlikely to achieve the optimal solution, thereby decomposing complex mixed integer programming problems into a series of manageable sub-problems. The parallel computing architecture processes based on the candidate solution space to accelerate the evaluation of multiple possible solutions. This architecture utilizes the many-core computing power of the graphics processing unit to simultaneously process a large number of solution candidates, and through parallel computation of objective function values and constraint satisfaction, significantly shortens the calculation time of each iteration. The optimal solution selection strategy obtains the optimal posture based on the accelerated evaluation results. This strategy compares the objective function values of all evaluated solutions, selects the posture configuration with the best overall performance as the final output, and ensures the quality and feasibility of the solution.

[0112] This hierarchical solving strategy takes full advantage of different computing techniques, the branch-and-bound algorithm provides a rigorous mathematical programming framework, the parallel computing architecture solves the computational bottleneck, and the optimal solution selection strategy guarantees the quality of the solution. Technically, this method can solve high-dimensional pose optimization problems within a reasonable time, because the branch-and-bound algorithm reduces the number of solutions to be evaluated through intelligent pruning, the parallel computing architecture significantly improves the evaluation speed of the remaining solutions, and the optimal solution selection strategy ensures the global optimality of the final result. The branch-and-bound algorithm uses the mathematical properties of integer programming to build a search tree and gradually converges to the optimal solution; the parallel computing architecture distributes the computational load to multiple processing units through data parallelism; and the optimal solution selection strategy makes decisions based on the optimality conditions in optimization theory. The entire solving process effectively controls the computational complexity while ensuring the quality of the solution.

[0113] The efficient solving mechanism enables task-adaptive pose prediction to meet the real-time requirements of practical applications, the branch-and-bound algorithm ensures the completeness of the search, the parallel computing architecture provides necessary computational acceleration, and the optimal solution selection strategy guarantees the optimality of the solution. This processing makes complex multi-objective pose optimization feasible in engineering practice, because the algorithm can complete the calculation within seconds, supporting interactive applications and rapid iteration in the design process. In principle, this method balances the quality of the solution and the cost of computation by combining the high reliability of precise algorithms with the high efficiency of parallel computing. Ultimately, this step provides fast and reliable pose decision-making capabilities for the entire digital twin system, enabling the system to respond promptly to different task scenarios and user needs, and playing an important role in practical engineering environments. Through the implementation of this optimization solving, the pose prediction module is no longer a theoretical concept, but a practical engineering tool with significant value, significantly improving the overall performance and usability of the digital twin system.

[0114] In some implementations, S5 includes:

[0115] S5.1: Based on the deformed three-dimensional human mesh and the optimal pose, a three-dimensional human mesh after pose adjustment is generated through bone skinning technology;

[0116] S5.2: Based on the three-dimensional human mesh after pose adjustment, a visual sitting posture digital model is generated through real-time rendering processing by a physics engine;

[0117] S5.3: Based on the visual sitting posture digital model, a high-fidelity sitting posture digital twin is generated through a dynamic feedback mechanism to reflect the deformation state of soft tissues after pose change.

[0118] In the system coupling process, the S5 step realizes the deep fusion of morphology and posture through multi-level technical integration, and completes the final construction from components to complete digital twin. The skinning technology is based on the deformation of three-dimensional human grid and the optimal posture, and generates a three-dimensional human grid after posture adjustment. This technology transmits the joint transformation of the skeletal system to the soft tissue grid vertices attached to it, so that the basic grid can deform naturally following the skeletal movement while maintaining the existing soft tissue deformation characteristics. The physics engine performs real-time rendering processing based on the three-dimensional human grid after posture adjustment, generates a visual sitting posture digital model, and through lighting calculation, material rendering and scene management technology, converts the geometric model into a visually realistic three-dimensional image, and provides interactive observation and analysis capabilities. The dynamic feedback mechanism processes based on the visual sitting posture digital model, generates a high-fidelity sitting posture digital twin reflecting the soft tissue deformation state after posture change, and through monitoring the influence of posture change on the contact state, triggers necessary recalculation to update the soft tissue deformation, ensuring that the mechanical response is consistent with the current posture.

[0119] This sequential processing establishes a complete process from data to visualization to feedback, the skinning technology realizes posture adaptation at the geometric level, the physics engine provides visual presentation, and the dynamic feedback mechanism maintains physical authenticity. Technically, this method can output a digital twin that is both visually realistic and physically accurate, because the skinning technology maintains the continuity and natural appearance of the grid, the physics engine provides high-quality visual performance, and the dynamic feedback mechanism ensures real-time consistency of mechanical behavior. The skinning technology converts joint rotation into vertex displacement through linear blend skinning algorithm, the physics engine uses computer graphics principles to achieve realistic rendering, and the dynamic feedback mechanism establishes the causal relationship between posture and deformation based on physical simulation principles. The entire processing flow forms a complete modeling closed loop, enabling the digital twin to dynamically respond to internal state changes.

[0120] The system coupling realizes the final presentation of the digital twin through this comprehensive processing, the skeleton skinning technology completes the geometric deformation, the physical engine realizes the visualization, and the dynamic feedback mechanism guarantees the physical accuracy. This integration makes the generated sitting posture digital twin not only have static authenticity, but also have dynamic response capability, because the skeleton skinning ensures the natural performance of posture changes, the physical engine provides an intuitive visual interface, and the dynamic feedback maintains the physical consistency of the model in different states. In principle, this method combines computer graphics, computational mechanics, and real-time simulation technology to create a digital counterpart that faithfully reflects the behavior of a real human body. Finally, this step provides a usable output result for the entire modeling process, allowing designers and engineers to intuitively evaluate the effects of different design schemes and support simulation-based design decisions and optimization processes. Through this end-to-end coupling process, the digital twin is no longer a collection of isolated components, but an organic whole system that can deliver greater value in practical applications.

[0121] Embodiment 2

[0122] As Figure 3 shown, in a second aspect, the present application proposes a physiological mechanism driven and data fusion system for sitting posture human body modeling, the system adopts the method provided by any of the above embodiments, and the system comprises:

[0123] A physiological mechanism driven module is configured to generate biomechanical attribute parameters including equivalent muscle activation threshold, spine segment bending stiffness coefficient, and biomechanical posture type probability vector based on input human macroscopic parameters.

[0124] A parameterized morphology generation module is configured to generate a basic three-dimensional human body mesh matching the physiological characteristics of a user based on human macroscopic parameters.

[0125] A soft tissue deformation simulation module is configured to generate a deformed three-dimensional human body mesh integrating the nonlinear compression deformation characteristics of soft tissue in a sitting posture based on the basic three-dimensional human body mesh and human macroscopic parameters.

[0126] A task-adaptive posture prediction module is configured to obtain an optimal posture matching the physiological characteristics of a user, the geometry of a vehicle interior, and a dynamic task scenario based on vehicle geometry parameters, task scenario identifiers, and biomechanical attribute parameters.

[0127] A system coupling module is configured to generate a high-fidelity sitting posture digital twin based on the deformed three-dimensional human body mesh and the optimal posture.

[0128] The system corresponds to the method provided by the above embodiment 1, and will not be described here one by one.

[0129] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can still be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present application.

Claims

1. Physiomechanics-driven and data-fused method for seated human body modeling, characterized in that, The method comprises the following steps: Based on the input human macroscopic parameters, the physiological mechanism driven model is processed to generate biomechanical attribute parameters including equivalent muscle activation threshold, spinal segment bending stiffness coefficient and biomechanical posture type probability vector; Based on the human macroscopic parameters, the parameterized shape generation model is processed to generate a basic three-dimensional human body grid matching the physiological characteristics of the user; Based on the basic three-dimensional human body grid and the human macroscopic parameters, the soft tissue deformation simulation model is processed to generate a deformed three-dimensional human body grid integrating the nonlinear compression deformation characteristics of the soft tissue in the sitting posture; Based on the vehicle geometric parameters, the task scene identifier and the biomechanical attribute parameters, the task adaptive posture prediction model is processed to obtain the optimal posture matching the physiological characteristics of the user, the vehicle interior geometry and the dynamic task scene; Based on the deformed three-dimensional human body grid and the optimal posture, the system coupling model is processed to generate a high-fidelity sitting posture digital twin; Based on the vehicle geometric parameters, the task scene identifier and the biomechanical attribute parameters, the task adaptive posture prediction model is processed to obtain the optimal posture matching the physiological characteristics of the user, the vehicle interior geometry and the dynamic task scene, including: Based on the biomechanical posture type probability vector in the biomechanical attribute parameters, the predefined lookup table is processed to determine the individualized neutral posture joint angle; Based on the individualized neutral posture joint angle, the vehicle geometric parameters and the task scene identifier, the mixed integer programming model is constructed and processed to establish a target function containing posture deviation penalty term and muscle activation penalty term; Based on the target function and the constraint conditions, the optimization solver is processed to solve and obtain the optimal posture; Based on the target function and the constraint conditions, the optimization solver is processed to solve and obtain the optimal posture, including: Based on the target function and the constraint conditions, the branch and bound algorithm is processed to generate a candidate solution space; Based on the candidate solution space, the parallel computing architecture is processed to accelerate the evaluation of multiple possible solutions; Based on the accelerated evaluation results, the optimal solution selection strategy is adopted to solve and obtain the optimal posture.

2. The physiological mechanism driven and data fusion method for sitting posture human body modeling according to claim 1, characterized in that, Based on the input human macroscopic parameters, the physiological mechanism driven model is processed to generate biomechanical attribute parameters including equivalent muscle activation threshold, spinal segment bending stiffness coefficient and biomechanical posture type probability vector, including: Based on the human macroscopic parameters including gender, age, height, body mass index and sitting height to height ratio, the input layer of the deep feedforward neural network model is processed to generate standardized input features; Based on the standardized input features, the multiple hidden layers of the deep feedforward neural network model are processed for nonlinear transformation to generate high-level feature representation; Based on the high-level feature representation, the output layer of the deep feedforward neural network model is processed to generate biomechanical attribute parameters.

3. The physiological mechanism driven and data fusion method for sitting posture human body modeling according to claim 2, characterized in that, Based on the standardized input features, the multiple hidden layers of the deep feedforward neural network model are processed for nonlinear transformation to generate high-level feature representation, including: Based on the standardized input features, the first fully connected layer of the deep feedforward neural network model is processed to generate the first hidden layer features; Based on the first hidden layer features, the regularization processing is performed through the dropout layer to generate the anti-overfitting hidden layer features; Based on the anti-overfitting hidden layer features, the second full connection layer of the deep feedforward neural network model is processed to generate high-level feature representation.

4. The physiological mechanism driven and data fusion method for sitting posture human body modeling according to claim 1, characterized in that, The parameterized shape generation model includes a statistical shape model, based on the human macroscopic parameters, the parameterized shape generation model is processed to generate a basic three-dimensional human mesh that matches the user's physiological characteristics, including: Based on the human macroscopic parameters, the principal component coefficient vector of the statistical shape model is generated by processing the pre-trained multiple linear regression model; Based on the principal component coefficient vector, the personalized basic three-dimensional human mesh is generated by linear combination processing through the statistical shape model; The topological inspection and repair processing is performed on the personalized basic three-dimensional human mesh to generate a basic three-dimensional human mesh that meets the requirements of the manifold structure.

5. The physiological mechanism driven and data fusion method for sitting posture human body modeling according to claim 1, characterized in that, Based on the basic three-dimensional human mesh and the human macroscopic parameters, the soft tissue deformation simulation model is processed to generate a deformed three-dimensional human mesh that integrates the nonlinear compression deformation characteristics of soft tissue in a sitting posture, including: Based on the basic three-dimensional human mesh, the tetrahedral element mesh of the hip and thigh region is generated by finite element discretization processing; Based on the human macroscopic parameters, the personalized material parameters of the hyperelastic material model are determined by processing the material parameter regression model; Based on the tetrahedral element mesh and the personalized material parameters, the deformed three-dimensional human mesh is generated by quasi-static analysis processing through the nonlinear finite element solver.

6. The physiological mechanism driven and data fusion method for sitting posture human body modeling according to claim 5, characterized in that, Based on the human macroscopic parameters, the personalized material parameters of the hyperelastic material model are determined by processing the material parameter regression model, including: Based on the body mass index and body fat rate in the human macroscopic parameters, the shear modulus parameter in the hyperelastic material model parameter is generated by processing the first regression model; Based on the body mass index and body fat rate in the human macroscopic parameters, the hardening index parameter in the hyperelastic material model parameter is generated by processing the second regression model; The shear modulus parameter and the hardening index parameter are integrated to generate personalized material parameters.

7. The physiology mechanism driven and data fusion method for sitting-posture human body modeling according to claim 1, characterized in that, Based on the deformed three-dimensional human mesh and the optimal posture, the high-fidelity sitting posture digital twin is generated by processing the system coupling model, including: Based on the deformed three-dimensional human mesh and the optimal posture, the three-dimensional human mesh after posture adjustment is generated by processing the skinning technology; Based on the three-dimensional human mesh after posture adjustment, the visualized sitting posture digital model is generated by real-time rendering processing through the physics engine; Based on the visualized sitting posture digital model, the high-fidelity sitting posture digital twin reflecting the deformation state of soft tissue after posture change is generated by processing the dynamic feedback mechanism.

8. A physiomechanics-driven and data-fused system for modeling a seated human body, characterized by, The system adopts the method of any one of claims 1 to 7, and the system comprises: A physiological mechanism driving module for generating biomechanical attribute parameters including equivalent muscle activation threshold, spinal segment bending stiffness coefficient and biomechanical posture type probability vector based on input human macroscopic parameters; A parameterized shape generation module for generating a basic three-dimensional human mesh that matches the user's physiological characteristics based on human macroscopic parameters; The soft tissue deformation simulation module is configured to generate a deformed three-dimensional human body mesh integrating the nonlinear compression deformation characteristics of soft tissue in a sitting posture based on a basic three-dimensional human body mesh and human body macroscopic parameters. The task-adaptive posture prediction module is configured to obtain an optimal posture matched with physiological characteristics of a user, geometry of a vehicle interior, and a dynamic task scene based on vehicle geometry parameters, a task scene identifier, and biomechanical attribute parameters. The system coupling module is configured to generate a high-fidelity sitting posture digital twin based on the deformed three-dimensional human body mesh and the optimal posture.

Images