Three-dimensional body surface shape generation method based on anthropometric parameter set

Abstract

The invention relates to the technical field of human body three-dimensional modeling, in particular to a human body measurement parameter set-based three-dimensional body surface shape generation method, which comprises the following steps of: S1, establishing a parameterized body surface model T; s2, determining a target anthropometric parameter group # imgabs0 #; S3, determining a transformation relation F based on the parameterized body surface model T and the target anthropometric parameter group # imgabs1 #; S4, constructing and verifying an optimized physical sign parameter beta *; and S5, substituting the optimized physical sign parameter beta * into any pose parameter theta, and changing the body surface geometric space pose of the optimized physical sign parameter beta *. By means of the method, finite element modeling of the human body in any form can be achieved, a statistical regression or condition generation model from one-dimensional measurement parameters to three-dimensional human body shapes does not need to be established, and various downstream human body modeling tasks oriented to sign diversity are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of human three-dimensional modeling, and specifically to a method for generating a three-dimensional body surface shape based on a set of anthropometric parameters. Background Art

[0002] The human finite element model is an important tool for biological injury simulation research. To reflect the diversity of the population body types, an algorithm for generating a human finite element model of a specific body type needs to be developed. Existing algorithms use limited and overall anthropometric parameters (height, weight, or BMI) as input conditions, generate a surface polygon mesh based on a statistical model, and then scale the reference model to a human finite element model corresponding to the body surface shape. Since anthropometric parameters exhibit characteristics of multiple postures (standing posture, sitting posture, etc.) and multiple modalities (length / dimension), a method needs to be designed to simultaneously match a set of multiple-posture and multi-modal measurement parameters in the three-dimensional shape parameter space.

[0003] Existing forward machine learning methods, such as multiple linear regression, need to preset a low-dimensional array of variables to be controlled, usually representative macroscopic dimensions such as height and weight, and more measurement parameters such as sitting height, hip circumference, and arm length, to achieve local shape changes. The reason for this problem is the lack of parametric association between a wide range of anthropometric parameters (such as 52 static dimensions covered in GB / T 10000-2023 "Human Dimensions of Chinese Adults") and the human outer epidermal geometry, which severely restricts automotive safety and biomechanics research for population physical sign diversity and urgently needs to be solved through technological upgrades. Summary of the Invention

[0004] The technical problem solved by the present invention is to provide a method for generating a three-dimensional body surface shape based on a set of anthropometric parameters, aiming to realize end-to-end shape parameter optimization for target attributes by constructing a measurement parameter representation in the form of a node coordinate function based on a human body surface statistical shape model, generate a three-dimensional body surface shape that meets the target measurement attributes, empower the finite element modeling of any shape of the human body, and there is no need to establish a statistical regression or conditional generation model from one-dimensional measurement parameters to three-dimensional human shapes.

[0005] The basic solution provided by the present invention:

[0006] A method for generating a three-dimensional body surface shape based on a set of anthropometric parameters, comprising:

[0007] S1. Establish a parametric body surface model T;

[0008] S2. Determine the target anthropometric parameter group

[0009] S3. Determine based on the parametric body surface model T and the target anthropometric parameter group Transformation relationship F:

[0010]

[0011] where β is a physical sign parameter, θ is a pose parameter, and N is the set of three-dimensional Cartesian coordinates of the nodes of the body surface mesh;

[0012] S4. Construct and verify the optimized physical sign parameter β*;

[0013] S5. According to the optimized physical sign parameter β*, any pose parameter, and the transformation relationship F, change the body surface geometric space posture of the optimized physical sign parameter β* to implement various downstream human body modeling tasks for physical sign diversity.

[0014] Basic principle: Based on the human body surface statistical shape model, construct a parameterized body surface model in the form of a node coordinate function, and establish a transformation relationship between the parameterized body surface model and the set of anthropometric parameters through a multi-objective optimization method. The multi-objective optimization method takes multiple pose parameters as inputs, optimizes multiple physical sign parameters, and then realizes end-to-end shape parameter optimization for target attributes by adjusting the measurement parameters, generates a three-dimensional body surface shape that meets the target measurement attributes, and empowers the finite element modeling of the human body in any form.

[0015] Beneficial effects: Compared with the prior art, the target human body outer skin mesh generated by this solution can not only conform to the target height and weight, but also conform to more comprehensive and detailed anthropometric parameters (such as 52 parameters covered in GB / T 10000-2023).

[0016] The method of this solution can flexibly support the directional generation calculation of physical sign parameters for any pose state. By designing a multi-objective optimization based on the spatial coordinate transformation relationship, it can meet the generation requirements of specific measurement parameter signs under various complex postures. Users do not need to establish a statistical regression or conditional generation model from one-dimensional measurement parameters to three-dimensional human body shapes, but modify the input parameters according to needs, and through the multi-objective optimization of joint multi-objective poses, based on the statistical shape model, directly approximate the target measurement parameters and quickly obtain the output results. It can reduce the algorithm complexity and data dependence, and greatly enhance the practicability and adaptability of the system.

[0017] Further, the S1 includes:

[0018] S11. Collect human body surface scan data and establish a unified template mesh for the samples based on manual or algorithm registration

[0019] S12. Establish a parameterized body surface shape statistical model based on the unified template mesh ;

[0020] S13. Based on the template mesh Establish a node coordinate mapping relationship from the standard pose to an arbitrary pose;

[0021] S14. Combine the parametric body surface shape statistical model and the node coordinate mapping relationship to form a parametric body surface model T.

[0022] Furthermore, the S11 includes:

[0023] S111. Use a scanning device to obtain high-precision three-dimensional models of a large number of human bodies and extract the corresponding node coordinates (N);

[0024] S112. Divide the reference polygon template grid for the three-dimensional model by the Delaunay triangulation method;

[0025] S113. Map the template grid to each acquisition sample.

[0026] Furthermore, the S12 includes:

[0027] S121. Perform rotational alignment on the node coordinates;

[0028] S122. Perform a principal component analysis transformation to reduce the dimensionality of the physical sign data and extract the key features of the shape;

[0029] S123. Change the specific attributes of the shape by adjusting the principal component coefficients;

[0030] S124. Invoke the inverse principal component analysis process to remap the adjusted principal component coefficients back to the original node coordinate space to achieve shape generation.

[0031] Furthermore, the S13 includes:

[0032] S131. Define joint pose parameters;

[0033] S132. Construct a tree structure of the bone system;

[0034] S133. Invoke the linear skinning algorithm to transfer the motion of the bone system to the body surface grid nodes;

[0035] S134. Use the dual quaternion optimization method to improve the accuracy and efficiency of pose transformation;

[0036] S135. Construct a node coordinate mapping relationship from the standard pose to an arbitrary pose.

[0037] Furthermore, the coordinate mapping relationship T defined in the S14 step is as follows:

[0038]

[0039] where β is the physical sign parameter and θ is the pose parameter, It is a set of three-dimensional Cartesian coordinates of n nodes of the body surface grid in the current physical sign and current posture.

[0040] Furthermore, the target anthropometric parameter group includes a set of physical sign parameters and a set of target pose parameters θ (i) , where the set of physical sign parameters is the integration of target human physical sign parameters; the set of target pose parameters θ (i) is a set of pose parameters corresponding to various measurement postures.

[0041] Furthermore, the S4 includes:

[0042] S41. Construct optimized physical sign parameters;

[0043] S42. Verify the optimized physical sign parameters.

[0044] Furthermore, the S41 includes:

[0045] S411. Randomly generate initial physical sign parameters β0;

[0046] S412. Based on the preset pose parameters θ (i) , calculate the body surface node coordinates

[0047] S413. Calculate the current anthropometric parameters based on the body surface node coordinates ;

[0048] S414. Establish an anthropometric parameter error function L to measure the error between the target anthropometric parameters and the current anthropometric parameters ;

[0049] S415. Based on the anthropometric parameter error L, call an optimization algorithm to update the initial physical sign parameters:

[0050]

[0051] In the formula, β n+1 is the updated physical sign parameter, and △β(L) is the correction value obtained by the optimization algorithm;

[0052] S416. Iteratively execute S412 to S415 until the anthropometric parameter error L is less than the preset threshold L*, determine that the optimization convergence condition is reached, terminate the loop, and obtain the optimized physical sign parameter β*.

[0053] Furthermore, the S42 includes:

[0054] Call the coordinate mapping relationship T to generate the node coordinate set of β*;

[0055] Substitute the set of node coordinates of β* into the template grid Generate an optimized body sign parameter body surface grid shape;

[0056] Calculate the measurement parameters X* of each dimension of the generated optimized body sign parameter body surface grid shape;

[0057] Calculate the deviation between X* and the target measurement parameters and compare it with a preset threshold. Description of the drawings

[0058] Figure 1 It is a flowchart of a three-dimensional body surface shape generation method based on a set of anthropometric parameters;

[0059] Figure 2 It is a schematic diagram of two measurement postures, standing / sitting postures;

[0060] Figure 3 It is a schematic diagram of the triangular grid contour segmentation method. Detailed implementation manners

[0061] The following is a further detailed description through specific implementation manners:

[0062] Example 1:

[0063] As Figure 1 shown, the three-dimensional body surface shape generation method based on a set of anthropometric parameters is as follows:

[0064] S1. Establish a parametric body surface model T, including the following steps:

[0065] S11. Collect human body surface scan data and establish a unified template grid for the samples based on manual or algorithm registration Specifically:

[0066] S111. Use a scanning device to obtain high-precision three-dimensional models of a large number of human bodies and extract the corresponding node coordinates (N);

[0067] S112. Divide a high-quality reference polygon template grid for the three-dimensional model;

[0068] This step can be completed by the standard triangular mesh of the human body surface obtained through Delaunay triangulation. First, a set of discrete points are extracted from the 3D model as the input for Delaunay triangulation. The extraction of discrete points should be based on the principles of uniform sampling and a moderate number; secondly, high-quality triangular meshes of the body surface are generated based on the sampled points, and constraint conditions are introduced to ensure that the generated triangular meshes conform to the geometric characteristics of the 3D model; the triangular meshes are optimized to eliminate bad shapes such as long and narrow triangles or degenerate triangles; finally, a standardized reference polygon template mesh is generated to ensure that the template mesh has a unified resolution and topological layout.

[0069] S113. Map the reference template mesh to each collected sample, extract the body surface shape features of each collected sample, and establish a human body surface shape dataset.

[0070] This step can adopt various methods such as shape non-rigid registration based on artificial shape descriptors such as HKS / SHOT, or based on deep learning features.

[0071] First, the HKS method is used to initially match the global features. HKS is a shape descriptor based on the Laplacian - Beltrami Operator, which can capture the local geometric characteristics of the shape, remain invariant under non-rigid deformations, and provides a robust way of shape description. Secondly, the SHOT method is used to further optimize the correspondence relationship and gradually refine the registration result. SHOT is a local shape descriptor that combines geometric and directional information, can effectively describe the local geometric structure, and remains invariant under rotation and translation transformations, and is widely used in point cloud registration and shape matching.

[0072] Use the HKS / SHOT features to find the initial corresponding points between the template mesh and the collected samples; based on these corresponding points, construct an energy function (such as distance error, smoothness constraint), and perform iterative registration through an optimization algorithm (such as ICP, Gauss - Newton); update the vertex positions of the template mesh to gradually approximate the shape of the collected samples.

[0073] Another method is to use a pre-trained deep learning model (such as PointNet, DGCNN) to extract the global or local features of the template mesh and the collected samples. Calculate the feature similarity (such as Euclidean distance, cosine similarity) between the template mesh and the collected samples to generate an initial pair of corresponding points. Use the optimal transport theory to optimize the point pair matching to ensure the global consistency of the matching result.

[0074] Based on the body surface shape features extracted from each collected sample, establish a human body surface shape dataset. Unify the mesh sample structures in the dataset to a preset template through registration in the form, where N is the number of nodes and F is the number of meshes. The mesh sample structure contains the same node - element connectivity relationship and is globally the same.

[0075] S12. Establish a parametric body surface shape statistical model based on a unified template mesh including the following steps:

[0076] S121. Rotate and align the node coordinates;

[0077] Calculate the centroid of the set of node coordinates for each sample; translate all node coordinates to a new coordinate system with the centroid as the origin; calculate the inertia tensor matrix of the sample and extract its eigenvectors to determine the main axis direction of the sample; then use the rotation matrix to rotate the sample to a standard spatial orientation. Rotating and aligning the node coordinates can ensure that each sample has a unified center point position and spatial orientation, eliminating the translational and rotational differences between samples, enabling subsequent analysis to focus on the characteristics of the shape itself rather than the changes in posture.

[0078] S122. Perform principal component analysis transformation to achieve data dimensionality reduction and compression, and extract the key features of the shape;

[0079] S123. Change specific attributes of the shape by adjusting the principal component coefficients;

[0080] Adjusting the values of the principal component coefficients in the principal component space can flexibly control the range of shape changes and generate new shape instances. To ensure that the generated shapes are reasonable, certain constraints can be imposed on the principal component coefficients, such as restricting them within the statistical range of the training data.

[0081] S124. Invoke the inverse principal component analysis process to remap the adjusted principal component coefficients back to the original node coordinate space to achieve shape generation.

[0082] S13. Establish a node coordinate mapping relationship from the default pose to any pose based on the template mesh The default pose can be any preset pose, which is preset as the standard standing pose in this embodiment.

[0083] Specifically, it includes the following steps:

[0084] S131. Define joint pose parameters;

[0085] Joint pose parameters are used to describe the changes in body surface postures, including joint rotation angles, rotation matrices, and quaternions, etc.

[0086] Joint rotation angle: Represent the rotation angle of the joint through Euler angles (such as rotation around the X - axis, Y - axis, and Z - axis).

[0087] Rotation matrix: Use a 3×3 rotation matrix to represent the rotation of the joint.

[0088] Quaternion: Represent rotations using unit quaternions to avoid gimbal lock problems and simplify interpolation calculations.

[0089] S132. Construct a tree structure of the skeletal system;

[0090] Establish a skeletal system with a tree structure, taking each joint as a node, and associating each joint with a local coordinate system to describe its pose relative to the parent node. The skeletal hierarchy provides a framework for subsequent motion transfer and skinning transformation.

[0091] S133. Invoke the linear skinning algorithm to transfer the motion of the skeletal system to the surface mesh nodes;

[0092] S134. Use the dual quaternion optimization method to improve the accuracy and efficiency of pose transformation;

[0093] S135. Construct a mapping relationship of node coordinates from the standard pose to any pose;

[0094] Record the coordinates of all nodes on the surface mesh in the standard pose. Update the transformation matrix of the skeletal system according to the joint pose parameters, and calculate the new coordinates of the surface mesh through linear skinning or dual quaternions.

[0095] S14. Combine the parametric surface shape statistical model and the node coordinate mapping relationship to form a parametric surface model.

[0096] The parametric surface model reflects the mapping relationship from physical sign parameters and pose parameters to node coordinates. By controlling the physical sign parameters and pose parameters, a three-dimensional surface mesh is output to characterize the surface shape of the human body. Geometric models of various human body shapes and sizes can be generated by adjusting a small number of parameters.

[0097] Among them, the physical sign parameters are used to describe individual geometric characteristics, including body type parameters (such as height, weight, BMI), dimension parameters (chest circumference, waist circumference, hip circumference, shoulder width, etc.), proportion parameters (head-to-body ratio, upper and lower limb ratio), bone parameters, etc. Different physical sign parameters have different value ranges.

[0098] The pose parameters are the parameters describing the pose changes of the model, including joint angles (Euler angles), rotation matrices, quaternions or dual quaternions, which are used to control the deformation of the model in different poses to achieve dynamic pose simulation. The pose parameters can be aligned with the national standard GB / T 5703-2023 "Basic Anthropometric Items for Technical Design", or can be obtained through on-site measurement of volunteer measurement data.

[0099] The node coordinates are defined as the three-dimensional coordinates of each vertex in the surface mesh, representing the specific geometric shape of the model in the corresponding pose.

[0100] The coordinate mapping relationship T defined by the parametric body surface model is as follows:

[0101]

[0102] where β is the physical sign parameter and θ is the pose parameter, is the set of three-dimensional Cartesian coordinates of n nodes of the body surface mesh under the current physical signs and current pose.

[0103] S2. Determine the target anthropometric parameter group

[0104] Target anthropometric parameter group is a set of measurement values related to the human body shape, including the set of physical sign parameters and the set of target pose parameters θ (i) . The set of physical sign parameters is the integration of human body physical sign parameters of different samples, including body type parameters, size parameters, proportion parameters, bone parameters, etc.; the set of target pose parameters θ(i) is the set of pose parameters corresponding to various measurement postures, such as the joint rotation angles and rotation matrices corresponding to the two measurement postures of standing / sitting postures defined in GB / T 10000-2023, as Figure 2 shown. It is also possible to find volunteers to reproduce the required poses and measure the corresponding pose parameters.

[0105] S3. Based on the parametric body surface model T and the target anthropometric parameter group Use the multi-objective optimization method to determine the transformation relationship F:

[0106]

[0107] where β is the set of physical sign parameters, θ is the set of target pose parameters, and N is the set of three-dimensional Cartesian coordinates of the body surface mesh nodes.

[0108] Determine the transformation relationship F as a multi-objective optimization problem, define multiple objective functions, and each objective function corresponds to a physical sign parameter. Add constraint conditions to ensure that the generated body surface model conforms to the real human body situation, synthesize multiple objective functions into a multi-objective optimization problem, select the Pareto optimal solution method or the multi-objective evolutionary algorithm for solution, perform iterative optimization and judge whether the convergence condition is reached. Use the optimized control parameters to establish the transformation relationship F.

[0109] When calculating length variables such as height, that is, the difference between the maximum and minimum values of the vertical direction coordinates, and at this time θ (i) is the pose parameter corresponding to the standard standing pose;

[0110] When calculating dimensional variables such as chest circumference, a triangular grid contour segmentation method is constructed, that is, the surface of the three-dimensional model is represented by a triangular grid. By segmenting the contour of the target area (such as the chest), segmented line segments are extracted and their total length is calculated to obtain the value of the target dimension (such as chest circumference), as Figure 3 shown.

[0111] It should be noted that the transformation function relationship can be flexibly designed according to specific transformation requirements. For example, for simple morphological changes, a linear model can be used to reduce the amount of calculation; for more complex pose changes, a non-linear model or even a combination of multiple types of functions can be adopted. To better capture details, the transformation relationship can be defined at different scales. When designing the change relationship, the actual physical limitations and biological principles of the human body can also be considered, such as the angular range of joints, the maximum contraction force of muscles, etc., and appropriate constraint conditions can be set to ensure that the transformation relationship is reasonable and practical.

[0112] S4. Generate and verify optimized physical sign parameters, including the following steps:

[0113] S41. Generate optimized physical sign parameter β*. Specifically, it includes:

[0114] S411. Generate initial physical sign parameter β0;

[0115] Select a suitable random generation method to generate the initial physical sign parameter, including based on the probability distribution model, based on constraint conditions, multi-parameter joint generation, etc. Check the generated initial physical sign parameter, including whether the parameter values are reasonable and the correlation relationship between parameters (for example, the chest circumference of an overweight individual is too small).

[0116] S412. Based on the preset pose parameter θ (i) , calculate the body surface node coordinates

[0117] Select the target pose from the set of target pose parameters, obtain the preset pose parameter, and calculate the body surface node coordinates according to the transformation relationship.

[0118] S413. Calculate the current anthropometric parameter based on the body surface node coordinates

[0119] S414. Establish an anthropometric parameter error function L to measure the error between the target anthropometric parameter and the current anthropometric parameter .

[0120] Among them, the anthropometric parameter error L is a measure of the set of all relative to , rather than being limited to a single dimension only.

[0121] The error function of the surveying parameters can use common methods such as weighted mean square error and cosine similarity. In this embodiment, the cosine similarity method is adopted to establish the error function of the surveying parameters.

[0122] S415. Based on the surveying parameter error L, call the optimization algorithm to update the initial physical sign parameter β0:

[0123]

[0124] where β n+1 is the updated physical sign parameter, and △β(L) is the correction value obtained by the optimization algorithm.

[0125] An optimization algorithm adopted in this embodiment is to use a deep learning framework such as PyTorch that supports automatic differentiation / backpropagation, and write the calculation expressions such as the principal component analysis result in step S12, the linear skinning calculation result in step S13, the surveying parameters in S3, and the surveying parameter error L in S414 into a function form that supports automatic differentiation, and call the framework to achieve end-to-end real-time update of the physical sign parameter β based on the surveying parameter error L.

[0126] S416. Iteratively execute S412 - S415 until the surveying parameter error L is less than the preset threshold L*, determine that the optimization convergence condition is reached, terminate the loop, and obtain the optimized physical sign parameter β*.

[0127] Since the user can input any set of surveying parameters, such as (height, sitting height, upper arm length, lower limb length), and the corresponding optimized physical sign parameter set can be generated according to the method of S41, the diversity of the input ensures that this method can be applied to diverse physical sign modeling analysis.

[0128] S42. Verify the optimized physical sign parameter β*, including the following steps:

[0129] Call the coordinate mapping relationship T to generate the node coordinate set of β*;

[0130] Substitute the node coordinate set of β* into the template grid Generate the surface grid shape of the optimized physical sign parameter;

[0131] Calculate the surveying parameters X* of each dimension of the generated surface grid shape of the optimized physical sign parameter;

[0132] Calculate the deviation between X* and the target surveying parameter and it should not exceed the preset threshold.

[0133] S5. Substitute the optimized physical sign parameter β* into any pose parameter θ, change the surface geometric space posture of the optimized physical sign parameter β*, and realize various downstream human body modeling tasks facing physical sign diversity.

[0134] According to the foregoing steps, any pose parameter includes but is not limited to the national standard standing / sitting posture. The target pose parameter set can be flexibly expanded by soliciting volunteers to measure the position parameters. Taking vehicle collision tests as an example, the pose can be expanded from the driver's seat to various poses of other seat members, and even the three-dimensional body surface shape analysis of zero-gravity members can be realized. Moreover, only the pose parameters of the members need to be measured during the analysis process, without the need for scanning and modeling.

[0135] The above are only embodiments of the present invention. Common knowledge such as the specific structures and characteristics known in the art are not described in detail herein. Those of ordinary skill in the art know all the common general technical knowledge in the technical field to which the invention pertains before the application date or the priority date, can know all the existing technologies in this field, and have the ability to apply the conventional experimental means before this date. Those of ordinary skill in the art can, with the inspiration given in this application and in combination with their own abilities, improve and implement this solution. Some typical well-known structures or well-known methods should not be an obstacle for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope required by this application should be based on the content of its claims, and the specific implementation manners described in the specification can be used to interpret the content of the claims.

Claims

1. A method for generating a three-dimensional body surface shape based on a set of anthropometric parameters, characterized in that: It includes the following steps: S1. Establish a parametric body surface model T; S2. Determine the target anthropometric parameter group S3. Determine the transformation relationship F based on the parametric body surface model T and the target anthropometric parameter group : where β is a physical sign parameter, θ is a pose parameter, and N is the set of three-dimensional Cartesian coordinates of the body surface mesh nodes; S4. Construct and verify the optimized physical sign parameter β*; S5. According to the optimized physical sign parameter β*, any pose parameter, and the transformation relationship F, change the body surface geometric space posture of the optimized physical sign parameter β*.

2. The three-dimensional body surface shape generation method based on the set of anthropometric parameters according to claim 1, wherein: The S1 includes: S11. Collect the body surface scan data of a human body and establish a unified template grid of a sample based on manual or algorithm registration S12. Establish a parametric statistical model of body surface shape based on a unified template grid ; S13. Based on a unified template grid Establish a node coordinate mapping relationship from the standard pose to an arbitrary pose; S14. Combine the parametric body surface shape statistical model and the node coordinate mapping relationship to form the parametric body surface model T.

3. The three-dimensional body surface shape generation method based on the set of anthropometric parameters according to claim 2, characterized in that: The S11 includes: S111. Use a scanning device to obtain high-precision three-dimensional models of a large number of human bodies and extract the corresponding node coordinates; S112. Divide the reference polygon template mesh for the three-dimensional model by the Delaunay triangulation method; S113. Map the template mesh to each collected sample.

4. The three-dimensional body surface shape generation method based on the set of anthropometric parameters according to claim 2, characterized in that: The S12 includes: S121. Perform rotational alignment on the node coordinates; S122. Perform principal component analysis transformation to reduce the dimension of the physical sign data and extract the key features of the shape; S123. Change the specific attributes of the shape by adjusting the principal component coefficients; S124. Invoke the inverse principal component analysis process to remap the adjusted principal component coefficients back to the original node coordinate space to achieve shape generation.

5. The method for generating a three-dimensional body surface shape based on a set of anthropometric parameters according to claim 2, wherein: The S13 includes: S131. Define the joint pose parameters; S132. Construct the tree structure of the bone system; S133. Invoke the linear skinning algorithm to transfer the motion of the bone system to the body surface mesh nodes; S134. Use the dual quaternion optimization method to improve the accuracy and efficiency of the pose transformation; S135. Construct the node coordinate mapping relationship from the standard pose to any pose.

6. The method for generating a three-dimensional body surface shape based on a set of anthropometric parameters according to claim 2, wherein: The coordinate mapping relationship T defined in the S14 step is as follows: where β is a physical sign parameter and θ is a pose parameter, which is a set of three-dimensional Cartesian coordinates of n nodes of the body surface mesh under the current physical signs and current pose.

7. The three-dimensional body surface shape generation method based on the anthropometric parameter set according to claim 6, characterized in that: The target anthropometric parameter set includes a set of physical sign parameters and a set of target pose parameters θ (i) , where the set of physical sign parameters is an integration of target human physical sign parameters; the set of target pose parameters θ (i) is a set of pose parameters corresponding to various measurement postures.

8. The three-dimensional body surface shape generation method based on the anthropometric parameter set according to claim 1, characterized in that: The S4 includes: S41. Construct the optimized physical sign parameter; S42. Verify the optimized physical sign parameter.

9. The three-dimensional body surface shape generation method based on the set of anthropometric parameters according to claim 8, characterized in that: The S41 includes: S411. Generate the initial physical sign parameter β0; S412. Calculate the body surface node coordinates based on the preset attitude parameter θ (i) ​ S413. Calculate the current measurement parameters based on the body surface node coordinates of the current measurement parameters S414. Establish a measurement parameter error function L to measure the target measurement parameter and the current measurement parameter error; S415. Based on the measurement parameter error L, invoke the optimization algorithm to update the initial physical sign parameter: where βn+1 is the updated physical sign parameter, and △β(L) is the correction value obtained by the optimization algorithm; S416. Iteratively execute S412 - S415 until the measurement parameter error L is less than the preset threshold L*, determine that the optimization convergence condition is reached, terminate the loop, and obtain the optimized physical sign parameter β*.

10. The three-dimensional body surface shape generation method based on a set of anthropometric parameters according to claim 8, characterized in that: The S42 includes: Invoke the coordinate mapping relationship T to generate the set of node coordinates of β*; Substitute the node coordinate set of β* into the unified template grid Generate an optimized body surface grid shape for the physical sign parameters; Calculate the measurement parameters X* of each dimension for the body surface mesh shape of the generated optimized physical sign parameter. Calculate the deviation of X* from the target metrological parameter and compare it with a preset threshold value.

Images