A fast geometry modeling method and device

Abstract

This invention discloses a rapid geometric modeling method and apparatus. The method includes extracting the three-dimensional symbolic distance field information of all frames of a human body template action sequence and processing it to obtain a data matrix; extracting the human body template torso skeleton of the first frame of human action; initializing the data matrix with parameters based on human body parts according to the human body template torso skeleton; inputting the parameters into a reconstruction model based on a quadratic kernel gate function to obtain the reconstructed symbolic distance fields of all frames; optimizing and updating the model parameters using the expectation-maximization algorithm; calculating the mean square error of the reconstructed symbolic distance fields of all frames and the original symbolic distance fields; reconstructing the symbolic distance field of each frame using the parameters corresponding to the minimum mean square error value; and reconstructing the three-dimensional mesh of the human body template from the reconstructed symbolic distance fields of all frames. This invention can effectively model the dynamic geometry of the three-dimensional human body, enabling the understanding of human actions to have spatiotemporal inductive generalization.

Description

Technical Field

[0001] This invention relates to the fields of computer graphics, 3D model reconstruction, motion capture, and digital human technology, and in particular to a rapid geometric modeling method and apparatus. Background Technology

[0002] Rapid geometric modeling is one of the important foundations for the development of digital humans and 3D reconstruction technology. It aims to build a rapid modeling system for dynamic 3D human data, laying the foundation for the characteristic expression of efficient digital human driving and multimodal control.

[0003] Dynamic human body modeling is crucial in 3D human body actuation research. Existing methods, such as local shape function representations based on Gaussian kernel implicit neural templates, are applied to human body template reconstruction; others are based on dynamic data from neural radiation fields. However, these algorithms cannot represent the characteristics of dynamic human body templates, thus hindering subsequent multimodal actuation research. Furthermore, existing dynamic geometric human body representations based on quadratic kernel models rely on neural network learning, exhibiting poor robustness. Traditional machine learning algorithms, on the other hand, struggle to handle the massive amounts of dynamic data. A robust yet independent approach that overcomes the speed barrier of deep learning remains elusive. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] To address this, the present invention proposes a rapid geometric modeling method to overcome the existing shortcomings.

[0006] Another object of the present invention is to provide a rapid geometric modeling device.

[0007] To achieve the above objectives, the present invention proposes a rapid geometric modeling method, comprising:

[0008] Extract the three-dimensional symbolic distance field information of all frames in the human body template action sequence and process it to obtain a data matrix;

[0009] Extract the human body template torso skeleton from the first frame of human motion;

[0010] Determine the number of modeling models;

[0011] The data matrix is ​​initialized with parameters based on human body parts according to the human body template to initialize the modeling symbolic distance field and obtain the original symbolic distance field.

[0012] The expected value maximization algorithm is used to optimize and update the model parameters;

[0013] Input the parameters into the reconstruction model based on the quadratic kernel gate function to obtain the symbolic distance fields of all frames reconstructed, and calculate the mean square error of the reconstructed symbolic distance fields of all frames and the original symbolic distance fields;

[0014] The model parameters are iteratively updated repeatedly, and the symbol distance fields of all frames are reconstructed using the parameters corresponding to the minimum mean square error value of the mean square error.

[0015] The 3D mesh of the human body template is reconstructed using the symbolic distance field reconstructed from all frames using the moving cube algorithm.

[0016] The rapid geometric modeling method of this invention may also have the following additional technical features:

[0017] In one embodiment of the present invention, the data matrix extracted frame by frame includes the symbolic distance field and corresponding position of T frames with resolution N0, arranged in a row number of N = N0. 3 ×T, a data matrix with 5 columns.

[0018] In one embodiment of the present invention, the human body template torso skeleton is obtained directly from the input human body template parameters and formed by connecting the vertices of the human body template in pairs.

[0019] In one embodiment of the present invention, if the number of models to be modeled is K, then the following condition must be met: the spatial symbolic distance field is uniformly divided into k1×k1×k1 cubes, and k1 3 =K, where k1 needs to be divisible by the resolution value N0.

[0020] In one embodiment of the present invention, the initialization of the modeling symbolic distance field includes:

[0021] 50 coordinate points are uniformly extracted from each segment of the extracted skeleton;

[0022] For each cube block, if any point on any line segment of the human skeleton is not within its range, then it is grouped into a group; for the remaining blocks, the sampling point on the nearest line segment of the skeleton is searched, and those that match the same segment are grouped together.

[0023] After the first frame is grouped, all other frames are grouped in the same way, and all frames that are in the same group are considered as a dynamic group.

[0024] After matching, the initial dynamic grouping is obtained as K groups. The mean vector, covariance matrix, and weight μ of the K groups of data are then calculated. j ,Σ j ,α j j = 1, 2, ..., K;

[0025] The parameter μ j ,Σj ,α j Substituting j = 1, 2, ..., K into the five-dimensional quadratic kernel gate function reconstruction formula Initialize the entire symbol distance field Where δ represents position, g(δ; α) j ,μ j ,Σ j ) represents the quadratic kernel gate function, m(δ; μ j ,Σ j ) represents the conditional mean of the second kernel.

[0026] In one embodiment of the present invention, the optimization and updating of the model parameters using the expected value maximization algorithm includes:

[0027] Update the second-order posterior probability Qt ij ,i=1,2,...,N,j=1,2,...,K;

[0028] Update parameters

[0029] In one embodiment of the present invention, the mean square error is the mean squared difference between the values ​​of the original and reconstructed symbol distance fields at the same location in each frame.

[0030] In one embodiment of the present invention, the model parameters are iteratively updated 50 times.

[0031] In one embodiment of the present invention, the method of reconstructing the three-dimensional mesh of the human body template using the symbolic distance field reconstructed in each frame using the moving cube algorithm includes:

[0032] For each frame of the reconstructed symbolic distance field, for each edge, if the symbols of the two connected vertices are different, interpolate along this edge, calculate the position where the interpolation value is 0, and place the vertex.

[0033] For each frame of the reconstructed symbolic distance field, construct triangular patches between the vertices.

[0034] To achieve the above objectives, another aspect of the present invention provides a rapid geometric modeling apparatus, comprising:

[0035] The data processing module is used to extract the three-dimensional symbolic distance field information of all frames of the human body template action sequence and process it to obtain a data matrix;

[0036] The skeleton extraction module is used to extract the human body template torso skeleton of the first frame of human motion.

[0037] The model number determination module is used to determine the number of modeling models.

[0038] The parameter initialization module is used to initialize the parameters of the data matrix based on the human body template torso skeleton, so as to initialize the modeling symbolic distance field and obtain the original symbolic distance field.

[0039] The model parameter update module is used to optimize and update the model parameters using the expected value maximization algorithm.

[0040] The mean square error calculation module is used to input parameters into the reconstruction model based on the quadratic kernel gate function to obtain the symbol distance field of each frame reconstruction, and to calculate the mean square error of the reconstructed symbol distance field and the original symbol distance field for all frames.

[0041] The distance field reconstruction module is used to iteratively update the model parameters and reconstruct the distance field of all frame symbols using the parameters corresponding to the minimum mean square error value of the mean square error.

[0042] The Human Body Mesh Modeling Module is used to reconstruct the 3D mesh of the human body template using the symbolic distance field reconstructed from all frames using the traveling cube algorithm.

[0043] The fast geometric modeling method and apparatus of this invention employ the expectation-maximization algorithm to learn the mean vector, covariance matrix, and weights of a five-dimensional quadratic kernel model, which can effectively model three-dimensional human dynamic geometry and enable the understanding of human movements to have spatiotemporal inductive generalization.

[0044] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0045] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0046] Figure 1 This is a flowchart of a rapid geometric modeling method according to an embodiment of the present invention;

[0047] Figure 2 This is a human body parameterized template motion capture vertex and skeleton diagram according to an embodiment of the present invention;

[0048] Figure 3 This is a structural diagram of a rapid geometric modeling apparatus according to an embodiment of the present invention. Detailed Implementation

[0049] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0050] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0051] The rapid geometric modeling method and apparatus according to embodiments of the present invention are described below with reference to the accompanying drawings.

[0052] Figure 1 This is a flowchart of a rapid geometric modeling method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes:

[0053] S1 extracts the three-dimensional symbolic distance field information of all frames of the human template action sequence and processes it to obtain a data matrix.

[0054] Specifically, the symbolic distance field of each frame of the human body template is extracted using the TSDF method;

[0055] Specifically, the motion template consists of T frames, each frame being a three-dimensional symbolic distance field with a spatial voxel grid of resolution N0. Each voxel stores a symbolic distance value, representing the distance from the center of the voxel to the surface of the object.

[0056] The 3D symbol range field position information and symbol range value are extracted frame by frame. For each voxel, its 3D position coordinates, time frame number, and symbol range value are extracted. The extracted data are arranged vertically into N = N0 rows. 3 ×T, a data matrix with 5 columns.

[0057] S2, extract the human body template torso skeleton of the first frame of human motion.

[0058] Specifically, the human skeleton is obtained directly from the input human template parameters, including the three-dimensional coordinates of 23 vertices, such as... Figure 2 As shown, the vertices are connected pairwise to form a human skeleton, and the skeleton segments are as follows. Figure 2 There are 23 segments shown.

[0059] S3, determine the number of modeling models.

[0060] If the number of models to be modeled is K, then the following condition must be met: the spatial symbolic distance field is uniformly divided into k1×k1×k1 cubes, and k1 3 =K. Where k1 needs to be divisible by the resolution value N0.

[0061] S4. The data matrix is ​​initialized with parameters based on human body parts according to the human body template to initialize the modeling symbolic distance field and obtain the original symbolic distance field.

[0062] Specifically, the rapid initialization of the modeled symbolic distance field in step S4 includes the following steps:

[0063] 50 coordinate points are uniformly extracted from each segment of the extracted skeleton;

[0064] For each cube block, if any point on any line segment of the human skeleton is outside its range, it is grouped into a separate group. For the remaining blocks, the sampling point on the nearest line segment of the skeleton is searched, and blocks matching the same segment are grouped together. Specifically, this includes: for each cube block, traversing each of the 23 segments of the skeleton, determining whether the 50 sampling points of each segment are within the block, i.e., whether the coordinates of the sampling points are all distributed within the three-dimensional coordinate range of the cube; if any point on any line segment of the human skeleton is outside its range, it is grouped into a separate group; if a cube block matches multiple sampling points, then the segment corresponding to the nearest sampling point is matched by calculating the mean square error of the distance; cube blocks matching the same segment are grouped together.

[0065] After the first frame is grouped, all other frames are grouped in the same way, and all frames that are in the same group are considered as a dynamic group.

[0066] After matching, the initial dynamic grouping is obtained as K groups. The mean vector, covariance matrix, and weight μ of the K groups of data are then calculated. j ,Σ j ,α j j = 1, 2, ..., K;

[0067] The parameter μ j ,Σ j ,α j Substituting j = 1, 2, ..., K into the five-dimensional quadratic kernel gate function reconstruction formula Initialize the entire symbol distance field Where δ represents position, g(δ; α) j ,μ j ,Σ j ) represents the quadratic kernel gate function, m(δ; μ j ,Σ j ) represents the conditional mean of the second kernel.

[0068] S5 uses the expected value maximization algorithm to optimize and update the model parameters.

[0069] E-step: Update the second-order posterior probability Qt ij ,i=1,2,...,N,j=1,2,...,K;

[0070] M-step: Update parameters

[0071] S6. Input the parameters into the reconstruction model based on the quadratic kernel gate function to obtain the symbolic distance fields of all frames reconstructed, and calculate the mean square error of the reconstructed symbolic distance fields of all frames and the original symbolic distance fields.

[0072] Specifically, the parameter μ of the r-th iteration rj ,Σ rj ,α rj j = 1, 2, ..., K is input into the gate function reconstruction formula. In this process, the model will reconstruct the symbolic distance field for each frame based on these parameters and kernel functions. The symbolic distance field (SDF) is a function that assigns a value to each point in space, representing the distance from that point to the nearest surface, and the value has a positive or negative sign to distinguish whether the point is inside or outside the surface.

[0073] The mean square error is the average of the squared differences between the original symbol range field and the reconstructed symbol range field at the same location in each frame. Specifically, the mean square error for the r-th iteration is defined as follows:

[0074]

[0075] S7, repeatedly iterate and update the model parameters, and use the parameters corresponding to the minimum mean square error value of the mean square error to reconstruct the symbol distance field of all frames.

[0076] In this embodiment, the parameter update iterations are performed 50 times.

[0077] The steps for reconstructing the symbol range field for each frame using the parameters corresponding to the minimum mean square error value include:

[0078] The iteration with the smallest mean square error is r. m ,Right now Its corresponding parameters are Therefore, the reconstructed symbolic distance field is...

[0079]

[0080] S8 uses the moving cube algorithm to reconstruct the 3D mesh of the human body template from the symbolic distance field reconstructed from all frames.

[0081] For each frame of the reconstructed symbolic distance field, for each edge, if the symbols of the two connected vertices are different, interpolate along this edge, calculate the position where the interpolation value is 0, and place the vertex.

[0082] For each frame of the reconstructed symbolic distance field, construct triangular patches between the vertices.

[0083] The rapid geometric modeling method according to embodiments of the present invention employs the expectation-maximization algorithm to learn the mean vector, covariance matrix, and weights of a five-dimensional quadratic kernel model, effectively modeling three-dimensional human dynamic geometry and enabling spatiotemporal inductive generalization in understanding human movements. Data classification initialization based on the human motion capture skeleton solves the difficulty of iterative computation of high-dimensional data and does not rely on neural networks, exhibiting a certain degree of robustness.

[0084] To achieve the above embodiments, such as Figure 3 As shown, this embodiment also provides a rapid geometric modeling device 10, including:

[0085] The data processing module 100 is used to extract the three-dimensional symbolic distance field information of all frames of the human body template action sequence and process it to obtain a data matrix;

[0086] The skeleton extraction module 200 is used to extract the human body template torso skeleton of the first frame of human motion.

[0087] The model number determination module 300 is used to determine the number of modeling models.

[0088] The parameter initialization module 400 is used to initialize the parameters of the data matrix based on the human body template torso skeleton to initialize the modeling symbolic distance field and obtain the original symbolic distance field.

[0089] The model parameter update module 500 is used to optimize and update the model parameters using the expected value maximization algorithm.

[0090] The mean square error calculation module 600 is used to input parameters into the reconstruction model based on the quadratic kernel gate function to obtain the symbol distance field of all frames reconstructed, and to calculate the mean square error of the reconstructed symbol distance field of all frames and the original symbol distance field.

[0091] The distance field reconstruction module 700 is used to iteratively update the model parameters and reconstruct the distance field of all frame symbols using the parameters corresponding to the minimum mean square error value of the mean square error.

[0092] The Human Mesh Modeling Module 800 is used to reconstruct the 3D mesh of the human body template using the symbolic distance field reconstructed from all frames using the traveling cube algorithm.

[0093] The rapid geometric modeling device according to embodiments of the present invention employs the expectation-maximization algorithm to learn the mean vector, covariance matrix, and weights of a five-dimensional quadratic kernel model, effectively modeling three-dimensional human dynamic geometry and enabling spatiotemporal inductive generalization in understanding human movements. Data classification initialization based on the human motion capture skeleton solves the difficulty of iterative computation of high-dimensional data and does not rely on neural networks, exhibiting a certain degree of robustness.

[0094] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0095] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A rapid geometric modeling method, characterized in that, include: Extract the three-dimensional symbolic distance field information of all frames in the human body template action sequence and process it to obtain a data matrix; Extract the human body template torso skeleton from the first frame of human motion; Determine the number of modeling models; Based on the topological structure of the human body template torso skeleton, the data points in the data matrix are dynamically grouped according to their spatial positions and the correspondence between the skeleton structure to obtain the initial signed distance field for modeling. Specifically, this includes: uniformly extracting 50 coordinate points on each segment of the extracted skeleton; traversing each cube block, if points on all segments of the human skeleton are not within its range, it is grouped into a group; for the remaining blocks, the sampling points on the nearest skeleton segment are found, and those matching the same segment are grouped together; after the position grouping of the first frame is completed, all other frames are grouped in the same way, and those in the same group in all frames are considered a dynamic group; after matching, the initial dynamic groups are obtained as K groups, and the mean vector, covariance matrix, and weights of the K groups of data are calculated. , , , j = 1,2,...,K; The parameters are... , , , j Substituting 1,2,...,K into the five-dimensional quadratic kernel gate function reconstruction formula Initialize the entire symbol distance field ,in Represents position, Represents a quadratic kernel gate function, This represents the conditional mean of the second-order kernel; The parameters of the quadratic kernel hybrid model are optimized and updated using the expected value maximization algorithm; The optimized and updated parameters are input into the reconstruction model based on the quadratic kernel gate function to obtain the symbolic distance fields of all frames reconstructed, and the mean square error of the symbolic distance fields of all frames reconstructed and the symbolic distance fields of the initial model are calculated. The model parameters are iteratively updated repeatedly, and the symbolic distance fields of all frames are reconstructed using the parameters corresponding to the minimum mean square error value of the mean square error. The 3D mesh of the human body template is reconstructed using the symbolic distance fields of all frames reconstructed by the traveling cube algorithm.

2. The method according to claim 1, characterized in that, The data matrix extracted frame by frame contains a symbol distance field with a T-frame resolution of N0 and corresponding positions, arranged into a data matrix with N=N0 rows and 5 columns. 3 ×T, 5 columns.

3. The method according to claim 1, characterized in that, The human body template torso skeleton is obtained directly from the input human body template parameters and formed by connecting the vertices of the human body template in pairs.

4. The method according to claim 2, characterized in that, If the number of models to be modeled is K, the following condition needs to be met: divide the spatial signed distance field uniformly into k1xk1xk1 cubes, and k1 3 = K, where k1 needs to be an integer that divides the resolution value N0.

5. The method according to claim 1, characterized in that, The optimization and updating of model parameters using the expected value maximization algorithm includes: Update the second-order post-validation probability ; Update parameters , .

6. The method according to claim 1, characterized in that, The mean square error is the squared mean of the difference between the original and reconstructed symbol distance fields at the same location in each frame.

7. The method according to claim 1, characterized in that, The model parameters are iterated and updated 50 times.

8. The method according to claim 1, characterized in that, The method of reconstructing the 3D mesh of the human body template using the symbolic distance field reconstructed from all frames using the moving cube algorithm includes: For each frame of the reconstructed symbolic distance field, for each edge, if the symbols of the two connected vertices are different, interpolate along this edge, calculate the position where the interpolation value is 0, and place the vertex. For each frame of the reconstructed symbolic distance field, construct triangular patches between the vertices.

9. A rapid geometric modeling device, characterized in that, include: The data processing module is used to extract the three-dimensional symbolic distance field information of all frames of the human body template action sequence and process it to obtain a data matrix; The skeleton extraction module is used to extract the human body template torso skeleton of the first frame of human motion. The model number determination module is used to determine the number of modeling models. The parameter initialization module is used to dynamically group data points in the data matrix according to the correspondence between their spatial positions and the skeleton structure, based on the topological structure of the human body template to obtain the initial signed distance field for modeling. Specifically, this includes: uniformly extracting 50 coordinate points on each segment of the extracted skeleton; traversing each cube block, if points on all line segments of the human skeleton are not within its range, it is grouped into a group; for the remaining blocks, it searches for sampling points on the nearest skeleton line segment, and those matching the same segment are grouped together; after the position grouping of the first frame is completed, all other frames are grouped in the same way, and those in the same group in all frames are considered a dynamic group; after matching, the initial dynamic groups are obtained as K groups, and the mean vector, covariance matrix, and weights of the K groups of data are calculated. , , , j = 1,2,...,K; The parameters are... , , , j Substituting 1,2,...,K into the five-dimensional quadratic kernel gate function reconstruction formula Initialize the entire symbol distance field ,in Represents position, Represents a quadratic kernel gate function, This represents the conditional mean of the second-order kernel; The model parameter update module is used to optimize and update the parameters of the quadratic kernel hybrid model using the expected value maximization algorithm; The mean square error calculation module is used to input the optimized and updated parameters into the reconstruction model based on the quadratic kernel gate function to obtain the symbolic distance fields of all reconstructed frames, and to calculate the mean square error of the symbolic distance fields of all reconstructed frames and the symbolic distance fields of the initial model. The distance field reconstruction module is used to iteratively update the model parameters and reconstruct the distance field of all frame symbols using the parameters corresponding to the minimum mean square error value of the mean square error. The Human Body Mesh Modeling Module is used to reconstruct the 3D mesh of the human body template using the symbolic distance field reconstructed from all frames using the traveling cube algorithm.

Images