Automated production line digital model reconstruction method and device

Abstract

The embodiment of the application provides a kind of automated production line digital model reconstruction method and device, extract geometric semantic features by layered section and depth convolution network, and innovatively build the implicit surface reconstruction framework with physical constraint.The system uses octree structure and parallel computing framework to realize efficient near neighbor point search, and combines principal component analysis and covariance matrix decomposition for feature extraction and noise filtering.Through the moving least square fitting of adaptive basis function and error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and are seamlessly integrated with the collision detection function of automated production line.The method breaks through the limitation of traditional model reconstruction, and provides an efficient and reliable solution for industrial digital twin.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to an automatic production line digital model reconstruction method and device. BACKGROUND

[0002] Traditional digital model reconstruction methods mainly rely on a single geometric reconstruction algorithm, which is difficult to handle model integrity and accuracy requirements in complex industrial scenarios. Existing technologies lack systematic methods in feature extraction and semantic understanding, and have low computational efficiency when processing large-scale point cloud data.

[0003] At the same time, the existing system has obvious shortcomings in surface reconstruction and topology preservation. Traditional methods often ignore physical constraints, resulting in deviations between reconstructed models and actual working conditions. The system is also relatively simple in surface splicing and error compensation, and cannot achieve high-quality model reconstruction.

[0004] In addition, existing technologies also have limitations in parallel computing optimization and industrial application integration. Lack of efficient nearest neighbor search strategy and adaptive surface fitting mechanism, and unable to fully consider the actual application requirements in industrial environment. The solution of these problems has important significance for improving the efficiency and accuracy of digital model reconstruction. SUMMARY

[0005] In view of the problems in the prior art, the present application provides an automatic production line digital model reconstruction method and device, which can break through the limitations of traditional model reconstruction and provide an efficient and reliable solution for industrial digital twin.

[0006] In order to solve at least one of the above problems, the present application provides the following technical scheme:

[0007] In a first aspect, the present application provides an automatic production line digital model reconstruction method, comprising:

[0008] Layered slicing of the automatic production line digital model along the main coordinate axis, extracting the feature point sequence of the slice contour line at different resolutions, geometric semantic annotation of the feature point sequence, construction of the annotated feature point sequence into a variational level set function, extraction of hierarchical feature mapping from the variational level set function using a deep convolutional network, establishment of an implicit surface reconstruction framework with physical constraint terms based on the hierarchical feature mapping, embedding the physical constraint terms as regularization conditions into the implicit surface reconstruction process;

[0009] Under the implicit surface reconstruction framework, an octree structure is adopted to spatially partition the feature point sequence, a near neighbor point search algorithm based on the spatial partition is constructed, the near neighbor point search algorithm is mapped into a parallel computing framework to establish an acceleration strategy, a local coordinate system is constructed according to the near neighbor point search result, an orthogonal basis vector of the feature point in the local coordinate system is calculated by using principal component analysis, an orthogonal basis space is established, and a feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix.

[0010] A local surface patch is constructed based on an eigenvector of the three-dimensional covariance matrix, an adaptive basis function is used for moving least squares fitting of the local surface patch, and in the fitting process, topological constraints and fairness constraints are added, the fitted surface patch is geometrically matched with the slice contour line, the surface patch is assembled and spliced according to the geometric matching result, the model after assembly and splicing is iteratively refined, an error compensation mechanism is established to eliminate the surface joint, and the compensated reconstruction model is imported into an automatic production line three-dimensional environment for collision detection.

[0011] Further, the automatic production line digital model is sliced along the main coordinate axis, and the feature point sequence of the slice contour line is extracted at different resolutions, the feature point sequence is geometrically annotated, and a variational level set function is constructed based on the annotated feature point sequence, including:

[0012] A layered scanning algorithm of a three-dimensional mesh model is used to calculate the slicing direction of the main coordinate axis, a slicing plane parallel to the coordinate axis is constructed based on the slicing direction, the slicing plane is intersected with the boundary surface of the digital model to generate a slice contour line, the slice contour line is meshed and divided to calculate the spatial coordinates of the feature points, and an adaptive resolution sampling mechanism is established to screen the feature points to obtain an ordered feature point sequence;

[0013] A curvature tensor analysis model is established for the feature point sequence, the geometric properties of the feature points are calculated based on the principal direction of the curvature tensor, the geometric properties are assigned to the feature points as label information, a variational target functional with a weight coefficient is constructed, the mathematical expression of the level set function is obtained by solving the variational target functional, and a mapping relationship between the feature point sequence and the level set function is established.

[0014] Further, the hierarchical feature mapping is extracted from the variational level set function by using a deep convolutional network, and an implicit surface reconstruction framework with a physical constraint term is established based on the hierarchical feature mapping, and the physical constraint term is embedded into the implicit surface reconstruction process as a regularization condition, including:

[0015] A multi-scale convolutional neural network structure is constructed, different sizes of convolution kernels are set in the convolutional layer to extract spatial features of the variational level set function, a pooling operation is performed on the feature map for dimension reduction compression, the dimension-reduced feature map is input into the deconvolutional layer for up-sampling, and the up-sampled result is spliced and fused with the feature map of the jump connection to generate a hierarchical feature mapping;

[0016] An implicit surface equation is constructed based on the hierarchical feature mapping, a physical constraint term of surface smoothness and shape preservation is added to the implicit surface equation, the physical constraint term is converted into a regularization condition in the form of a Lagrange multiplier, a target function with a regularization term is established, and a variational method is used to solve the target function to obtain an expression of implicit surface reconstruction.

[0017] Further, under the implicit surface reconstruction framework, an octree structure is used to spatially partition the sequence of feature points, a neighbor point search algorithm based on the spatial partition is constructed, and the neighbor point search algorithm is mapped to a parallel computing framework to establish an acceleration strategy, including:

[0018] A spatial bounding box of the sequence of feature points is calculated, the spatial bounding box is divided into eight subspaces according to the coordinate axis direction, the density distribution of the feature points in each subspace is calculated, the subspaces are recursively subdivided based on a density distribution threshold, an octree index structure is established to store the spatial position information of the feature points, and the spatial position information of the feature points is recorded in the corresponding leaf nodes;

[0019] A neighbor point search path is constructed in the octree index structure, the spatial distance between the feature point to be searched and the leaf nodes is calculated, the leaf nodes with a spatial distance less than a search radius are marked as candidate nodes, parallel search tasks are established for the candidate nodes, the parallel search tasks are distributed to the computing units for neighbor point search calculation, and the search results are summarized to obtain the neighborhood information of the feature points.

[0020] Further, a local coordinate system is constructed according to the neighbor point search results, a principal component analysis method is used to calculate the orthogonal basis vectors of the feature points in the local coordinate system, an orthogonal basis space is established, a feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix, and the three-dimensional covariance matrix is input into a noise filtering module for data enhancement, including:

[0021] The center of gravity coordinates of the feature point neighborhood are calculated using the neighbor point search results, the center of gravity coordinates are translated to the origin to establish the origin of the local coordinate system, the divergence matrix of the feature point neighborhood is calculated, the divergence matrix is eigenvalue decomposed to obtain a feature vector matrix, the feature vector matrix is sorted according to the eigenvalues to determine three basis vectors of the local coordinate system, and a rotation matrix of the local coordinate system is constructed based on the basis vectors;

[0022] Transform the feature points from the global coordinate system to the local coordinate system, calculate the coordinate components of the feature points in the local coordinate system, construct a projection matrix of the feature points, perform singular value decomposition on the projection matrix to obtain a three-dimensional covariance matrix, calculate a noise threshold based on the eigenvalue distribution of the three-dimensional covariance matrix, and input the noise threshold as a filtering parameter into an anisotropic diffusion filter.

[0023] Further, based on the eigenvectors of the three-dimensional covariance matrix, a local surface patch is constructed, an adaptive basis function is used to perform a moving least squares fitting on the local surface patch, and in the fitting process, a topological constraint and a fairness constraint are added. The fitted surface patch is geometrically matched with the slice contour line, including:

[0024] The eigenvectors of the three-dimensional covariance matrix are used as normal vectors to construct a parametric equation of the local surface patch, the principal curvatures and Gaussian curvatures of the local surface patch are calculated, the order of the adaptive basis function is selected based on the curvature information, a moving least squares equation with a weight function is constructed, the topological continuity constraint and the curvature fairness constraint of the surface are introduced into the moving least squares equation, and a weighted least squares objective function with constraint terms is established;

[0025] An analytical solution is constructed for the weighted least squares objective function, a numerical solution is obtained for the objective function using a domain decomposition method, the solution result is substituted into the parametric equation of the local surface patch to obtain a fitted surface, a distance field distribution of the fitted surface and the slice contour line is calculated, a geometric matching criterion is constructed based on the distance field, and the geometric matching criterion is used as an evaluation index to screen an optimal fitted surface.

[0026] Further, the surface patches are assembled and spliced according to the geometric matching results, the assembled and spliced model is iteratively refined, an error compensation mechanism is established to eliminate surface seams, and the compensated reconstructed model is imported into an automated production line three-dimensional environment for collision detection, including:

[0027] The splicing boundary of adjacent surface patches is calculated according to the geometric matching results, a transition region is constructed at the splicing boundary, the geometric parameters of the transition region are used as constraint conditions to establish a surface splicing equation, an iterative optimization algorithm is used to solve the surface splicing equation to obtain continuity parameters of the surface, and an error compensation function is constructed based on the continuity parameters to eliminate splicing gaps in the surface seam region;

[0028] A grid topology structure is established for the assembled reconstructed model, a geometric quality evaluation index of the grid topology structure is calculated, the area that needs to be refined is determined based on the geometric quality evaluation index, a grid densification criterion is constructed in the refined area, an adaptive grid densification algorithm is used to locally refine the reconstructed model, and the refined reconstructed model is converted into a three-dimensional solid model and imported into the automated production line environment.

[0029] In a second aspect, the present application provides an automatic production line digital model reconstruction device, comprising:

[0030] a model slicing module, configured to slice an automatic production line digital model along a main coordinate axis, extract a feature point sequence of a slice contour line at different resolutions, perform geometric semantic labeling on the feature point sequence, construct a variational level set function from the labeled feature point sequence, extract a hierarchical feature mapping from the variational level set function using a deep convolutional network, and establish an implicit surface reconstruction framework with a physical constraint term based on the hierarchical feature mapping, and embed the physical constraint term as a regularization condition into an implicit surface reconstruction process;

[0031] a spatial partitioning module, configured to perform spatial partitioning on the feature point sequence using an octree structure under the implicit surface reconstruction framework, construct a nearest neighbor point search algorithm based on the spatial partitioning, map the nearest neighbor point search algorithm to a parallel computing framework to establish an acceleration strategy, construct a local coordinate system according to the nearest neighbor point search results, calculate an orthogonal basis vector of the feature points in the local coordinate system using a principal component analysis method, establish an orthogonal basis space, decompose a feature point projection matrix in the orthogonal basis space into a three-dimensional covariance matrix, and input the three-dimensional covariance matrix into a noise filtering module for data enhancement;

[0032] a model reconstruction module, configured to construct a local surface patch based on an eigenvector of the three-dimensional covariance matrix, perform a moving least squares fitting on the local surface patch using an adaptive basis function, add a topological constraint and a fairness constraint in the fitting process, perform geometric matching between the fitted surface patch and the slice contour line, assemble and splice the surface patches according to the geometric matching results, perform iterative refinement on the assembled and spliced model, establish an error compensation mechanism to eliminate surface seams, and import the compensated reconstruction model into an automatic production line three-dimensional environment for collision detection.

[0033] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the steps of the automatic production line digital model reconstruction method.

[0034] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the automatic production line digital model reconstruction method.

[0035] In a fifth aspect, the present application provides a computer program product, comprising a computer program / instruction, wherein the computer program / instruction is executable by a processor to implement the steps of the automatic production line digital model reconstruction method.

[0036] From the above technical solutions, the application provides an automatic production line digital model reconstruction method and device, geometric semantic features are extracted through layered slicing and deep convolutional network, and an implicit surface reconstruction framework with physical constraints is innovatively constructed. The system adopts an octree structure and a parallel computing framework to realize efficient neighbor point search, and combines principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through adaptive basis function moving least squares fitting and error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and the method is seamlessly integrated with the collision detection function of the automatic production line. The method breaks through the limitations of traditional model reconstruction and provides an efficient and reliable solution for industrial digital twinning. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 One of the flowcharts of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0039] Figure 2 The second flowchart of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0040] Figure 3 The third flowchart of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0041] Figure 4 The fourth flowchart of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0042] Figure 5 The fifth flowchart of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0043] Figure 6 The sixth flowchart of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0044] Figure 7 The seventh flowchart of the automatic production line digital model reconstruction method in the embodiments of the present application;

[0045] Figure 8 The structure diagram of the automatic production line digital model reconstruction device in the embodiments of the present application;

[0046] Figure 9A structural schematic diagram of an electronic device in an embodiment of the present application.

[0047] Reference signs:

[0048] Electronic device 9600, central processor 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage section 9142, data storage section 9143, driver program storage section 9144, antenna 9111, speaker 9131, microphone 9132. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0050] The acquisition, storage, use, processing, etc. of data in the technical solutions of the present application all comply with relevant provisions of national laws and regulations.

[0051] In view of the problems in the prior art, the present application provides an automatic production line digital model reconstruction method and device, which extracts geometric semantic features through hierarchical slicing and deep convolutional network, and innovatively constructs an implicit surface reconstruction framework with physical constraints. The system uses octree structure and parallel computing framework to realize efficient nearest neighbor point search, and combines principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through the moving least squares fitting of adaptive basis functions and the error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and the collision detection function of the automatic production line is seamlessly integrated. This method breaks through the limitations of traditional model reconstruction and provides an efficient and reliable solution for industrial digital twinning.

[0052] In order to break through the limitations of traditional model reconstruction and provide an efficient and reliable solution for industrial digital twinning, an embodiment of an automatic production line digital model reconstruction method is provided in the present application, as shown in Figure 1 , the automatic production line digital model reconstruction method specifically includes the following contents:

[0053] Step S101: slicing the digital model of the automated production line along the main coordinate axis, extracting a feature point sequence of a slice contour line at different resolutions, performing geometric semantic labeling on the feature point sequence, constructing a variational level set function from the labeled feature point sequence, extracting a hierarchical feature mapping from the variational level set function using a deep convolutional network, and establishing an implicit surface reconstruction framework with a physical constraint term based on the hierarchical feature mapping, and embedding the physical constraint term as a regularization condition into the implicit surface reconstruction process;

[0054] Optionally, in the digital model slicing process, an adaptive slicing strategy is adopted. For different types of equipment models in the automated production line, the optimal slicing direction is determined by a three-dimensional main direction analysis algorithm. First, the principal curvature distribution of the model surface grid is calculated, a curvature tensor field is constructed, and the local coordinate system is determined based on the characteristic vector of the curvature tensor. For the joint structure of the robot arm, the slicing direction parallel to the joint rotation axis is preferred; for the conveyor belt system, the slicing direction is determined according to the normal vector of the conveying plane; for regular equipment such as control cabinets, the slicing direction parallel to the main boundary surface is selected.

[0055] In the determination of the slice interval, a geometric feature-based interval adjustment algorithm is developed. By analyzing the local curvature rate and shape complexity of the model, an adaptive interval function is constructed. When processing precision components such as the end effector of a robot arm, the system automatically reduces the slice interval when the local curvature exceeds a preset threshold, improving the sampling accuracy; for gradual structures such as conveyor belts, the slice interval is appropriately increased to optimize the calculation efficiency.

[0056] In the slicing contour line extraction link, a multi-resolution feature sampling mechanism is realized. First, a quadtree structure is constructed to partition the slice plane in space, and the curvature and gradient information is calculated in each leaf node. For high-precision areas such as the end effector of a robot arm, the system automatically subdivides the grid until the preset precision requirement is met; for the planar area of the control cabinet, a larger grid size is maintained to reduce data redundancy.

[0057] In the geometric semantic labeling process, a local feature-based classification algorithm is developed. By constructing the local neighborhood of the feature points, geometric descriptors including principal curvature, Gaussian curvature, and average curvature are calculated. For the movement joints of the robot arm, the axial features and mating surface features are highlighted; for the guide structure of the conveyor belt, the boundary features and transition features are highlighted; for the mounting surface of the control cabinet, the flatness and perpendicularity features are focused on.

[0058] In the construction of the variational level set function, a target functional with weights is adopted. The target functional includes data terms and regularization terms. The data terms measure the distance between the reconstructed surface and the sampling points, and the regularization terms control the smoothness of the surface. When dealing with precision fitting parts such as the flange of a robot arm, the data term weight is increased to ensure the reconstruction accuracy. When dealing with the transition area of a conveyor belt, the regularization term weight is appropriately increased to ensure smooth transition of the surface.

[0059] In the design of the deep convolutional network, a multi-scale feature extraction mechanism is implemented. The network structure adopts the U-Net architecture, which includes 5 layers of encoders and 5 layers of decoders. The encoder gradually extracts features through 3x3 convolution layers and max pooling layers, and the decoder restores the spatial resolution through deconvolution layers. The input data is the normalized level set function value, the intermediate features are fused through jump connection to realize the fusion of different scale information, and the output is a 32-channel feature map, each channel corresponding to a geometric feature of different scales.

[0060] In the post-processing of the feature map, an attention mechanism enhanced feature fusion strategy is developed. By calculating the correlation matrix between different levels of features, the weight distribution of the features is adaptively adjusted. For complex structures such as robot bearing seats, the system can accurately identify and retain key geometric features; for continuous features such as conveyor guide grooves, the continuity of the surface is emphasized.

[0061] In the construction process of the physical constraint term, a constraint model based on material mechanics is adopted. The constraint conditions include the bending energy, stretching energy and torsional energy of the surface, and the physical reasonableness of the reconstruction result is ensured by minimizing the total energy. When dealing with load-bearing structures such as robot support seats, the stress distribution characteristics of the material are considered; when dealing with conveyor supports, the structure is ensured to meet the stiffness requirements.

[0062] In the embedding of the regularization condition, a weight adaptive mechanism based on geometric features is implemented. By analyzing the rate of change of local curvature and the complexity of features, a dynamic weight function is constructed. When dealing with smooth areas such as the round corner transition of a robot arm, the weight of the smoothing term is increased; when dealing with feature areas such as sensor installation grooves, the weight of the feature retention term is increased.

[0063] The above implementation scheme constructs a complete model reconstruction framework. This scheme has excellent performance in practical applications and can accurately reconstruct various types of automated production line equipment models, including robot arms, conveyor belts, control cabinets, etc. The reconstruction result not only maintains geometric accuracy, but also ensures the reasonableness of physical properties, providing a reliable digital model foundation for subsequent dynamics simulation and path planning.

[0064] In the production line digital upgrading project, the scheme of the embodiment is successfully applied to the equipment modeling of multiple automatic production lines. Through accurate geometric reconstruction and reasonable physical constraints, the reliability of the digital twin model is significantly improved, effectively supporting the optimization design and transformation and upgrading of the production line.

[0065] Step S102: Under the implicit surface reconstruction framework, the feature point sequence is spatially partitioned using an octree structure, a near neighbor point search algorithm based on the spatial partitioning is constructed, the near neighbor point search algorithm is mapped into a parallel computing framework to establish an acceleration strategy, a local coordinate system is constructed according to the near neighbor point search result, an orthogonal basis vector of the feature points in the local coordinate system is calculated using principal component analysis, an orthogonal basis space is established, and a feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix. The three-dimensional covariance matrix is input into a noise filtering module for data enhancement.

[0066] Optionally, in the octree structure construction process, an adaptive spatial partitioning strategy is adopted. First, the space bounding box of the feature point sequence is calculated, and the initial division scale is determined by analyzing the point cloud density distribution. For complex structure regions such as robot joints, when the local point density exceeds a predetermined threshold, the system automatically subdivides the space unit until a predetermined spatial resolution is reached; for feature simple regions such as conveyor bodies, a larger space unit size is maintained to improve computational efficiency.

[0067] In the construction of spatial index, a dynamic balanced tree structure optimization algorithm is developed. By calculating the spatial distribution characteristics of the feature points in each node, including the point density change rate and the direction distribution, the splitting and merging strategy of the node is adaptively adjusted. When dealing with precision components such as robot end effectors, the system dynamically adjusts the depth of the tree according to the complexity of the local geometric features; for regular structures such as control cabinet shells, a more uniform tree layer structure is maintained.

[0068] In the process of near neighbor point search, a multi-thread parallel search mechanism is realized. First, the feature points are divided into multiple spatially continuous data blocks, and each data block is assigned to an independent computing thread. For the moving joints of the robot, the topological structure of the local neighborhood is analyzed; for the guide structure of the conveyor belt, the continuous distribution characteristics of the feature points are focused on. Through the CUDA architecture, GPU acceleration is realized, which significantly improves the search efficiency.

[0069] In the construction of the local coordinate system, a coordinate system optimization strategy based on geometric features is adopted. By analyzing the spatial distribution of the near neighbor point set, the center of gravity position and the main direction of the point set are calculated. For the mating surface of the robot, the coordinate axes consistent with the assembly direction are preferentially selected; for the moving surface of the conveyor belt, the coordinate axes are aligned with the motion direction. The system ensures the stability of the coordinate system through iterative optimization.

[0070] In this embodiment, a robust feature extraction algorithm was developed for principal component analysis (PCA) calculations. The principal direction vector is obtained by constructing the divergence matrix of nearest neighbors and using SVD decomposition. When dealing with precision-fitting structures such as robotic arm flanges, the reliability of the principal direction is determined by analyzing the distribution of eigenvalues; for planar features such as sensor mounting surfaces, the accuracy of the normal vector is the primary focus.

[0071] In this embodiment, an iterative optimization orthogonalization method is used in the calculation of orthogonal basis vectors. Orthogonalization constraints are constructed by analyzing the angular relationships between principal direction vectors. For complex structures such as robotic arm bearing seats, the system ensures the orthogonality of the basis vectors through multiple iterative optimizations; for regular structures such as control cabinet mounting surfaces, the principal directions are directly used to construct the orthogonal basis.

[0072] In this embodiment, a feature-weight-based projection strategy is implemented when constructing the projection matrix. By analyzing the distribution characteristics of feature points in different directions, a weight coefficient is assigned to each projection direction. When dealing with smooth areas such as the rounded corners of the robotic arm, the weight of the tangential component is increased; when dealing with feature areas such as sensor mounting slots, the weight of the normal component is increased.

[0073] In this embodiment, an adaptive feature analysis algorithm was developed for the calculation of the three-dimensional covariance matrix. By analyzing the distribution pattern of eigenvalues, the geometric feature types of local regions are identified. For the irregular surface of the robotic arm, the principal curvature characteristics of the surface are determined by the proportion of eigenvalues; for the guide groove of the conveyor belt, the salience of the feature is determined by the magnitude of the eigenvalues.

[0074] In this embodiment, an adaptive filtering strategy based on geometric features is employed during noise filtering. An anisotropic filtering kernel function is constructed by analyzing the eigenvalue distribution of the covariance matrix. When processing precision parts of the robotic arm, a smaller filtering radius is used to preserve detailed features; when processing the conveyor belt body, a larger filtering radius is used to improve the smoothing effect.

[0075] This embodiment constructs a highly efficient and stable feature analysis framework through the above implementation scheme. This scheme demonstrates excellent performance in practical applications, accurately extracting and enhancing the geometric features of various automated production line equipment. The acceleration strategy based on parallel computing significantly improves processing efficiency, providing strong support for the digital reconstruction of large-scale production line equipment.

[0076] In the production line digital twin project, the solution in this embodiment was successfully applied to feature extraction and optimization in complex industrial scenarios. Through precise geometric analysis and efficient data processing, the quality of the reconstructed model was significantly improved, laying a solid foundation for subsequent simulation analysis and optimization design.

[0077] Step S103: Construct local surface patches based on the eigenvectors of the three-dimensional covariance matrix, and perform moving least squares fitting on the local surface patches using adaptive basis functions. During the fitting process, topological constraints and smoothness constraints are added. Geometric matching is performed between the fitted surface patches and the slice contour lines. The surface patches are assembled and spliced ​​according to the geometric matching results. The assembled and spliced ​​model is iteratively refined. An error compensation mechanism is established to eliminate surface seams. The compensated reconstructed model is imported into the three-dimensional environment of the automated production line for collision detection.

[0078] Optionally, in this embodiment, the principal direction of the surface is determined based on the eigenvectors of the three-dimensional covariance matrix during the construction of local surface patches. For different functional components in an automated production line, such as the joints of a robotic arm, the guide grooves of a conveyor belt, and the mounting surface of a control cabinet, the local parameterization method of the surface is determined by analyzing the directional distribution of the eigenvectors. For the spherical joints of the robotic arm, spherical coordinates are used for parameterization; for the cylindrical structure of the conveyor belt, cylindrical coordinates are used; and for the planar structure of the control cabinet, rectangular coordinates are used.

[0079] In this embodiment, a basis function construction strategy based on geometric features is developed for adaptive basis function selection. By analyzing the curvature distribution and topological features of local surfaces, the type and order of the basis functions are dynamically adjusted. When dealing with complex surfaces such as robotic arm end effectors, high-order B-spline basis functions are used to ensure fitting accuracy; for regular surfaces of conveyor belts, low-order polynomial basis functions are selected to improve computational efficiency.

[0080] This embodiment implements an adaptive weight allocation mechanism during the moving least squares fitting process. By constructing a distance-based weight function, different influence factors are assigned to sampling points in the neighborhood. When dealing with the precision mating surfaces of the robotic arm, the weight of nearest neighbors is increased to ensure local fitting accuracy; when dealing with the transition region of the conveyor belt, a larger influence radius is used to ensure a smooth transition of the surface.

[0081] In this embodiment, a boundary matching-based constraint model was developed during the topology constraint construction phase. Positional and tangential continuity constraints were established by analyzing the connection relationships between adjacent surface patches. For the complex assembly surfaces of the robotic arm, precise docking of adjacent surface patches was ensured; for the guide structure of the conveyor belt, continuous transition of the surfaces was guaranteed.

[0082] This embodiment employs a multi-objective optimization strategy in the design of smoothness constraints. A smoothness evaluation function is constructed by comprehensively considering the mean curvature, Gaussian curvature, and torsional energy of the surface. During the optimization process, the constraint weights are dynamically adjusted according to the functional requirements of different regions to ensure the geometric quality of the surface.

[0083] In this embodiment, a feature-based matching algorithm is implemented during the geometric matching process. By analyzing the geometric features of the surface patches and slice contours, feature descriptors and similarity metrics are constructed. For precision parts of the robotic arm, strict matching criteria are adopted; for conventional components of the conveyor belt, the matching conditions are appropriately relaxed to improve splicing efficiency.

[0084] In this embodiment, an adaptive transition region generation algorithm was developed for the surface patch splicing process. By analyzing the geometric features of adjacent surface patches, the range and shape of the transition region are dynamically determined. At critical joints of the robotic arm, a narrower transition region is constructed to ensure accuracy; at general transitions of the conveyor belt, a wider transition region is used to ensure smoothness.

[0085] In this embodiment, a mesh optimization strategy based on error analysis is adopted during the model iteration and refinement process. By calculating the distance field between the reconstructed model and the original data, the areas that need refinement are identified. For the precision parts of the robotic arm, local mesh refinement is performed; for the flat areas of the conveyor belt, a moderate mesh density is maintained.

[0086] In this embodiment, an adaptive compensation algorithm is implemented in the error compensation mechanism design. A local deformation field is constructed by analyzing the geometric discontinuities at the surface joints. During the compensation process, material properties and mechanical constraints are considered to ensure the physical rationality of the compensation results.

[0087] This embodiment establishes a complete framework for surface reconstruction and optimization through the above implementation scheme. This scheme can accurately reconstruct various functional components in automated production lines and ensure that the reconstruction results meet practical application requirements. In the collision detection stage, by establishing an efficient spatial index structure, rapid interference analysis is achieved, providing a reliable geometric basis for production line layout optimization and path planning.

[0088] In practical applications, the solution in this embodiment has successfully handled a variety of complex production line equipment reconstruction problems, such as the irregular joints of robotic arms and the guiding structure of conveyor belts, significantly improving the accuracy and usability of digital models and effectively supporting the intelligent upgrading and transformation of production lines.

[0089] As described above, the automated production line digital model reconstruction method provided in this application can extract geometric semantic features through hierarchical slicing and deep convolutional networks, innovatively constructing an implicit surface reconstruction framework with physical constraints. The system employs an octree structure and a parallel computing framework to achieve efficient nearest neighbor search, combining principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through adaptive basis function moving least squares fitting and an error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and it is seamlessly integrated with the collision detection function of the automated production line. This method overcomes the limitations of traditional model reconstruction, providing an efficient and reliable solution for industrial digital twins.

[0090] In one embodiment of the automated production line digital model reconstruction method of this application, see [link to relevant documentation]. Figure 2 It can also specifically include the following:

[0091] Step S201: The layered scanning algorithm of the three-dimensional mesh model is used to calculate the slicing direction of the main coordinate axis. Based on the slicing direction, a slicing plane parallel to the coordinate axis is constructed. The slicing plane is intersected with the boundary surface of the digital model to generate a slicing contour line. The slicing contour line is meshed to calculate the spatial coordinates of the feature points. An adaptive resolution sampling mechanism is established to filter the feature points to obtain an ordered sequence of feature points.

[0092] Step S202: Establish a curvature tensor analysis model for the feature point sequence, calculate the geometric properties of the feature points based on the principal direction of the curvature tensor, assign the geometric properties as label information to the feature points, construct a variational objective functional with weight coefficients, obtain the mathematical expression of the level set function by solving the variational objective functional, and establish the mapping relationship between the feature point sequence and the level set function.

[0093] Optionally, in this embodiment, a hierarchical scanning strategy is used to determine the optimal slicing direction during the 3D mesh model analysis. For robotic arms in automated production lines, their joint motion characteristics are first analyzed, and the main motion axes are identified by calculating the normal vector distribution of local surfaces. For conveyor belt systems, the slicing direction is determined based on the conveying direction and the mounting reference plane. For equipment such as control cabinets, a suitable slicing direction is determined by analyzing the boundary features of their main structure.

[0094] In this embodiment, an adaptive slicing algorithm based on geometric features was developed during the construction of the slicing plane. By analyzing the geometric complexity of the model in different directions, the slice spacing is dynamically adjusted. When processing precision components such as robotic arm end effectors, a smaller slice spacing is used to ensure sampling accuracy; for regular structures such as conveyor belts, a larger slice spacing is used to improve computational efficiency.

[0095] In this embodiment, a precise intersection calculation mechanism is implemented in the slice contour line generation stage. Intersection detection is accelerated by constructing a spatial bounding box, and a precise numerical algorithm is used to calculate the intersection line between the slice plane and the model boundary surface. For the irregular surfaces of the robotic arm, special attention is paid to the contour features of the curved surface; for the guide structure of the conveyor belt, key geometric features are preserved.

[0096] In this embodiment, an adaptive mesh generation strategy is employed during the meshing process. By analyzing the curvature changes of the contour lines, mesh nodes are densified in high-curvature regions, while mesh density is appropriately reduced in flat regions. For the precision mating surfaces of the robotic arm, high-density meshing is used to ensure reconstruction accuracy; for the planar structure of the control cabinet, coarser meshing is used to improve efficiency.

[0097] In this embodiment, a multi-criteria evaluation mechanism was developed for the feature point selection stage. By comprehensively considering the spatial distribution, curvature characteristics, and geometric importance of the points, a priority ranking of the feature points was established. When dealing with complex curved surfaces at the joints of the robotic arm, more feature points were retained; for the regular structure of the conveyor belt, the number of sampling points was appropriately reduced.

[0098] This embodiment implements a robust curvature estimation algorithm during curvature tensor analysis. It calculates the principal curvature and principal direction by constructing a neighborhood of feature points fitted to a local quadratic surface. For the spherical joints of the robotic arm, spherical features are accurately extracted; for the cylindrical structure of the conveyor belt, the generatrix direction is precisely calculated.

[0099] In this embodiment, a multi-scale feature extraction strategy is employed in the geometric attribute calculation stage. By analyzing the variation patterns of the curvature tensor at different scales, a hierarchical geometric descriptor is constructed. For the precision parts of the robotic arm, detailed-level geometric features are extracted; for the conventional components of the conveyor belt, macroscopic geometric features are focused on.

[0100] In this embodiment, a weighting scheme based on physical meaning was developed when constructing the variational objective functional. By analyzing the importance of different geometric features, reasonable weight coefficients were assigned to each term of the objective functional. When dealing with the functional surfaces of the robotic arm, the weight of the conformal terms was increased; when dealing with the transition region of the conveyor belt, the weight of the smoothing terms was increased.

[0101] This embodiment implements an efficient numerical computation method in solving the level set function. By constructing a multi-grid solver, adaptive time step size is used to control the convergence of the solution. For the complex structure of the robotic arm, a fine solution mesh is used; for the simple structure of the conveyor belt, a coarser solution mesh is used.

[0102] In this embodiment, a nonlinear mapping strategy is employed in the feature mapping establishment stage. By analyzing the distribution characteristics of feature points, a highly adaptable mapping function is constructed. When dealing with the irregular surfaces of the robotic arm, a locally adaptive mapping method is used; for the regular surfaces of the conveyor belt, a globally uniform mapping method is used.

[0103] This embodiment constructs a complete geometric feature extraction and analysis framework through the above implementation scheme. This scheme can accurately capture various geometric features of automated production line equipment and establish reasonable mathematical expressions. In practical applications, this scheme has successfully processed various equipment models, including robotic arms, conveyor belts, and control cabinets, providing reliable feature data support for the subsequent reconstruction process.

[0104] In the production line digital upgrade project, the solution in this embodiment significantly improved the accuracy and efficiency of geometric feature extraction, laying a solid foundation for the accurate modeling of production line equipment and effectively supporting subsequent simulation analysis and optimization design work.

[0105] In one embodiment of the automated production line digital model reconstruction method of this application, see [link to relevant documentation]. Figure 3 It can also specifically include the following:

[0106] Step S301: Construct a multi-scale convolutional neural network structure, set convolutional kernels of different sizes in the convolutional layer to extract the spatial features of the variational level set function, perform dimensionality reduction and compression on the feature map through pooling operation, input the dimensionality-reduced feature map into the deconvolutional layer for upsampling, and concatenate and fuse the upsampling result with the feature map of the skip connection to generate a hierarchical feature map.

[0107] Step S302: Construct an implicit surface equation based on the hierarchical feature mapping, add physical constraints for surface smoothness and shape preservation to the implicit surface equation, transform the physical constraints into regularization conditions in the form of Lagrange multipliers, establish an objective function with regularization terms, and use variational method to solve the objective function to obtain the expression for implicit surface reconstruction.

[0108] Optionally, this embodiment employs a hierarchical feature extraction strategy during the construction of the multi-scale convolutional neural network. A U-Net structure comprising a five-layer encoder and a five-layer decoder is designed to address the geometric features of different types of equipment in automated production lines. In the encoder, combinations of convolutional kernels of three sizes—3×3, 5×5, and 7×7—capture the local details of precision parts in the robotic arm, the medium-scale features of the conveyor belt guide structure, and the large-scale geometric features of the control cabinet shell, respectively.

[0109] In this embodiment, an adaptive feature extraction mechanism is implemented in the convolutional layer design. By setting deformable convolutional modules at different levels, the network's adaptability to complex geometries is enhanced. When processing complex curved surfaces at the joints of a robotic arm, deformable convolution can automatically adjust the shape of the receptive field to accurately capture local features; for the regular structure of a conveyor belt, a standard rectangular receptive field is maintained to improve computational efficiency.

[0110] In this embodiment, an adaptive pooling strategy based on geometric features was developed in the pooling operation design. By analyzing the distribution of geometric information in the feature map, different pooling strategies are adopted in different regions. For the precision mating surfaces of the robotic arm, a smaller pooling window is used to retain detailed information; for the planar structure of the control cabinet, a larger pooling window is used to reduce computational complexity.

[0111] This embodiment employs a multi-stage feature recovery mechanism in the deconvolution layer implementation. By designing a learnable upsampling kernel, geometric details are preserved during feature recovery. A refined upsampling strategy is used when reconstructing precision components such as robotic arm end effectors; for conventional structures such as conveyor belts, standard bilinear interpolation is used to improve efficiency.

[0112] In this embodiment, an attention-enhanced feature integration strategy is implemented in the feature fusion stage. By calculating the correlation matrix between features at different levels, the weight distribution of features is adaptively adjusted. For the irregular surface of the robotic arm, the focus is on local geometric features; for the continuous curved surface of the conveyor belt, the focus is on preserving global shape features.

[0113] In this embodiment, a constraint model based on physical laws was developed when constructing the implicit surface equations. By introducing surface energy functionals, including bending energy, tensile energy, and torsional energy, the physical rationality of the reconstruction results is ensured. When dealing with load-bearing structures such as robotic arm support seats, the mechanical properties of the materials are given special consideration; when dealing with conveyor belt supports, the structure is ensured to meet stiffness requirements.

[0114] In this embodiment, an adaptive constraint weight adjustment strategy is adopted during the physical constraint transformation process. By analyzing the complexity of local geometric features, the weight coefficients of constraint terms are dynamically adjusted. For the functional surfaces of the robotic arm, the weight of the shape preservation term is increased; for the transition region of the conveyor belt, the weight of the smoothness constraint is increased.

[0115] In this embodiment, a multi-objective optimization constraint framework is implemented during the regularization condition construction stage. By constructing constraints in the form of Lagrange multipliers, the reconstruction accuracy and surface quality are balanced. When processing precision parts of the robotic arm, stronger regularization constraints are used to ensure surface quality; for conventional parts of the conveyor belt, the constraints are appropriately relaxed to improve reconstruction efficiency.

[0116] In this embodiment, an iterative optimization strategy was developed during the objective function solution process. A gradient descent optimizer was constructed, and an adaptive step size control algorithm was employed to ensure solution convergence. For the complex structure of the robotic arm, a smaller optimization step size was used to ensure accuracy; for the simple structure of the conveyor belt, a larger step size was used to accelerate convergence.

[0117] This embodiment constructs a complete deep learning reconstruction framework through the above implementation scheme. This scheme can accurately reconstruct various equipment models in automated production lines and ensure that the reconstruction results meet engineering application requirements. In practical applications, this scheme has successfully handled the reconstruction tasks of various complex equipment, including robotic arms, conveyor belts, and control cabinets, significantly improving the accuracy and reliability of digital models.

[0118] In the production line digital upgrade project, the solution in this embodiment effectively improves the quality and efficiency of model reconstruction by combining deep learning and physical constraints, providing a reliable geometric basis for the intelligent transformation of the production line and supporting subsequent simulation optimization and intelligent control work.

[0119] In one embodiment of the automated production line digital model reconstruction method of this application, see [link to relevant documentation]. Figure 4 It can also specifically include the following:

[0120] Step S401: Calculate the spatial bounding box of the feature point sequence, divide the spatial bounding box into eight subspaces according to the coordinate axis direction, calculate the density distribution of feature points in each subspace, recursively subdivide the subspaces based on the density distribution threshold, establish an octree index structure to store the spatial location information of the feature points, and record the spatial location information of the feature points in the corresponding leaf nodes.

[0121] Step S402: Construct a nearest neighbor search path in the octree index structure, calculate the spatial distance between the feature point to be searched and the leaf node, mark the leaf node whose spatial distance is less than the search radius as a candidate node, establish a parallel search task for the candidate node, allocate the parallel search task to the computing unit to perform the nearest neighbor search calculation, and summarize the search results to obtain the neighborhood information of the feature point.

[0122] Optionally, this embodiment employs an adaptive boundary analysis strategy during the spatial bounding box calculation. For different types of equipment models in an automated production line, the spatial range of the feature point set is first calculated. For complex structures such as robotic arms, the bounding box boundary is determined by analyzing the joint motion space; for conveyor belt systems, the spatial range is determined based on the conveying path and guiding structure; for regular equipment such as control cabinets, the boundary is directly determined using the main body contour.

[0123] This embodiment implements an adaptive partitioning mechanism based on geometric features during subspace partitioning. By analyzing the distribution characteristics of feature points within each region, the partitioning strategy is dynamically adjusted. When processing precision components such as robotic arm end effectors, a smaller partitioning scale is used to accurately capture local features; for regular structures such as conveyor belts, a larger partitioning scale is used to improve efficiency.

[0124] In this embodiment, a multi-scale density estimation algorithm was developed for the density distribution calculation stage. By constructing a kernel function density estimator, the degree of point clustering is analyzed at different spatial scales. For the complex curved surface of the robotic arm, a smaller kernel function radius is used to improve the local accuracy of the density estimation; for the planar structure of the control cabinet, a larger kernel function radius is used to reduce computational overhead.

[0125] This embodiment implements an adaptive termination criterion during the recursive subdivision process. The subdivision depth is dynamically adjusted by analyzing the geometric complexity of the subspace. When dealing with the precision mating surfaces of a robotic arm, deeper recursion levels are allowed; when dealing with the regular structure of a conveyor belt, the recursion depth is appropriately controlled to balance accuracy and efficiency.

[0126] In this embodiment, a hierarchical node management strategy is adopted in the octree construction stage. By designing an efficient parent-child node link structure, the storage and access efficiency of the tree is optimized. For the irregular surface of the robotic arm, local feature information is mainly saved; for the continuous curved surface of the conveyor belt, the overall shape features are mainly recorded.

[0127] In this embodiment, a heuristic search strategy was developed for constructing the nearest neighbor search path. By analyzing the spatial relationship between the target point and the node center, a priority queue was constructed to guide the search order. When dealing with complex structures at the joints of the robotic arm, regions with high geometric similarity are searched first; for the regular structure of the conveyor belt, a simple distance-first criterion is adopted.

[0128] This embodiment implements an adaptive measurement mechanism in the spatial distance calculation process. By considering the geometric characteristics of local curved surfaces, an anisotropic distance metric is constructed. For the spherical joints of the robotic arm, a spherical distance metric is used; for the cylindrical structure of the conveyor belt, a cylindrical distance metric is used.

[0129] In this embodiment, a dynamic load balancing strategy is adopted in the parallel task allocation stage. By analyzing the computational complexity of candidate nodes, computing resources are allocated rationally. When processing the precision parts of the robotic arm, more computing units are allocated to high-complexity areas; for the conventional parts of the conveyor belt, a uniform allocation strategy is adopted.

[0130] In this embodiment, a feature similarity-based filtering mechanism was developed during the search result aggregation process. By analyzing the geometric features of neighboring points, unreasonable search results are eliminated. For the functional surfaces of the robotic arm, the geometric consistency of neighboring points is strictly controlled; for the transition areas of the conveyor belt, the filtering conditions are appropriately relaxed.

[0131] This embodiment constructs an efficient and stable spatial index structure and nearest neighbor search framework through the above implementation scheme. This scheme can quickly and accurately identify local feature associations in the automated production line equipment model, providing reliable neighborhood information support for subsequent geometric reconstruction.

[0132] In the production line digital upgrade project, the solution in this embodiment significantly improves the efficiency and accuracy of feature analysis. By combining parallel computing and intelligent indexing, it effectively supports the digital reconstruction of large-scale production line equipment, providing important technical support for the intelligent transformation of production lines.

[0133] In one embodiment of the automated production line digital model reconstruction method of this application, see [link to relevant documentation]. Figure 5 It can also specifically include the following:

[0134] Step S501: Calculate the centroid coordinates of the neighborhood of the feature point using the nearest neighbor search results, translate the centroid coordinates to the origin to establish the origin of the local coordinate system, calculate the divergence matrix of the neighborhood of the feature point, perform eigenvalue decomposition on the divergence matrix to obtain the eigenvector matrix, sort the eigenvector matrix according to the size of the eigenvalues ​​to determine the three basis vectors of the local coordinate system, and construct the rotation matrix of the local coordinate system based on the basis vectors.

[0135] Step S502: Transform the feature points from the global coordinate system to the local coordinate system, calculate the coordinate components of the feature points in the local coordinate system, construct the projection matrix of the feature points, perform singular value decomposition on the projection matrix to obtain the three-dimensional covariance matrix, calculate the noise threshold based on the eigenvalue distribution of the three-dimensional covariance matrix, and input the noise threshold as a filtering parameter into the anisotropic diffusion filter.

[0136] Optionally, this embodiment employs a weighted average calculation strategy in the calculation of the center of gravity coordinates. For the feature point distribution of different types of equipment in the automated production line, weight coefficients are determined by analyzing the geometric importance of neighboring points. For the precision mating surfaces of the robotic arm, different weights are assigned based on curvature changes; for the guiding structure of the conveyor belt, weights are determined based on the spatial distribution characteristics of the points; and for the planar structure of the control cabinet, uniform weighting is used to simplify the calculation.

[0137] This embodiment implements a coordinate system optimization mechanism based on geometric features when establishing the local coordinate system. By analyzing the distribution characteristics of neighboring points, it ensures that the coordinate axis directions correspond to the physical features. When dealing with the spherical surface at the joint of the robotic arm, the main axis is aligned with the direction of the sphere's center; for the cylindrical structure of the conveyor belt, the main axis is aligned with the direction of the generatrix.

[0138] In this embodiment, a robust feature extraction algorithm was developed for the divergence matrix calculation stage. By constructing a high-order matrix to describe the spatial distribution of neighborhood points, local geometric features are accurately captured. For the irregular surface of the robotic arm, high-order divergence features are calculated; for the regular curved surface of the conveyor belt, low-order divergence is used for description.

[0139] This embodiment implements an adaptive numerical calculation strategy during eigenvalue decomposition. A suitable decomposition algorithm is selected by analyzing the condition number of the matrix. A high-precision iterative algorithm is used when processing the precision parts of the robotic arm; a fast direct solution method is used for the conventional structure of the conveyor belt.

[0140] In this embodiment, a multi-criteria evaluation mechanism is employed in the basis vector sorting process. The priority of basis vectors is determined by comprehensively considering the magnitude and geometric meaning of eigenvalues. For the functional surfaces of the robotic arm, particular attention is paid to the accuracy of the normal vector; for the guide structure of the conveyor belt, the focus is on ensuring the direction of the tangent vector.

[0141] In this embodiment, a physical constraint-based optimization strategy was developed when constructing the rotation matrix. By introducing orthogonality and right-handed constraints, the rationality of coordinate transformation is ensured. When processing the assembly surface of the robotic arm, the coordinate axes are ensured to be consistent with the assembly direction; for the motion surface of the conveyor belt, the coordinate axes are made to correspond to the motion direction.

[0142] This embodiment implements a high-precision numerical calculation mechanism during coordinate transformation. A stable transformation algorithm is designed to reduce the accumulation of numerical errors. Double-precision calculations are used for the precision components of the robotic arm, while single-precision calculations are used to improve efficiency for the conventional structure of the conveyor belt.

[0143] In this embodiment, an adaptive projection strategy is adopted in the projection matrix construction stage. The optimal projection direction is determined by analyzing the geometric characteristics of the local curved surface. Spherical projection is used for the spherical joints of the robotic arm, while cylindrical projection is used for the cylindrical structure of the conveyor belt.

[0144] In this embodiment, a robust computational method was developed for singular value decomposition. An iterative optimization strategy was designed to ensure the accuracy of the decomposition results. An accurate solution algorithm was used when dealing with the complex curved surfaces of the robotic arm; a fast approximation algorithm was used for the simple curved surfaces of the conveyor belt.

[0145] This embodiment implements an adaptive parameter estimation mechanism when calculating the noise threshold. The filtering strength is dynamically determined by analyzing the eigenvalue distribution of the covariance matrix. For the precision parts of the robotic arm, a smaller filtering threshold is used to preserve details; for the conventional parts of the conveyor belt, a larger threshold is used to improve the smoothing effect.

[0146] This embodiment constructs a complete local feature analysis framework through the above implementation scheme. This scheme can accurately extract the local geometric features of automated production line equipment models and effectively suppress noise influence through anisotropic filtering. In practical applications, this scheme has successfully processed various complex equipment models, including robotic arms, conveyor belts, and control cabinets, significantly improving the quality of geometric feature extraction.

[0147] In the production line digital upgrade project, the solution in this embodiment provides reliable feature support for accurate modeling of production line equipment through precise geometric analysis and effective noise suppression, thereby promoting the intelligent transformation process of the production line.

[0148] In one embodiment of the automated production line digital model reconstruction method of this application, see [link to relevant documentation]. Figure 6 It can also specifically include the following:

[0149] Step S601: Construct the parametric equation of the local surface patch by using the eigenvector of the three-dimensional covariance matrix as the normal vector, calculate the principal curvature and Gaussian curvature of the local surface patch, select the order of the adaptive basis function based on the curvature information, construct the moving least squares equation with weight function, introduce the topological continuity constraint and curvature compliance constraint of the surface into the moving least squares equation, and establish the weighted least squares objective function with constraint terms.

[0150] Step S602: Construct an analytical solution for the weighted least squares objective function, use the domain decomposition method to numerically solve the objective function, substitute the solution result into the parametric equation of the local surface patch to obtain the fitted surface, calculate the distance field distribution between the fitted surface and the slice contour line, construct a geometric matching criterion based on the distance field, and use the geometric matching criterion as an evaluation index to select the optimal fitted surface.

[0151] Optionally, in this embodiment, during the construction of local surface patches, the principal direction of the surface is determined based on the eigenvectors of the three-dimensional covariance matrix. For the robotic arm joints in automated production lines, the eigenvector corresponding to the largest eigenvalue is selected as the normal vector to ensure that the surface adapts to the motion characteristics; for the guide structure of the conveyor belt, the tangential direction is determined based on the eigenvector corresponding to the second largest eigenvalue to ensure consistency with the conveying direction.

[0152] In this embodiment, a multi-scale curvature estimation strategy is employed in the curvature calculation stage. By calculating the principal curvature and Gaussian curvature within different neighborhood ranges, local geometric features are accurately captured. For the spherical joint of the robotic arm, the uniformity of curvature is the primary focus; for the transition region of the conveyor belt, the trend of curvature variation is analyzed.

[0153] This embodiment implements an adaptive order adjustment mechanism when selecting basis functions. By analyzing the complexity of local surfaces, the type and order of the basis functions are dynamically determined. When dealing with the precision mating surfaces of the robotic arm, high-order B-spline basis functions are used to ensure fitting accuracy; for the planar structure of the control cabinet, low-order polynomial basis functions are used to improve efficiency.

[0154] In this embodiment, a weight allocation strategy based on geometric features was developed during the construction of the weight function. By analyzing the spatial distribution and geometric importance of points, reasonable weight coefficients were determined. For the irregular surface of the robotic arm, the weight of key feature points was increased; for the regular curved surface of the conveyor belt, a distance-attenuated weight allocation was adopted.

[0155] In this embodiment, multi-level continuity requirements are adopted in the topological constraint design. By establishing positional continuity and tangential continuity constraints, smooth transitions between adjacent surface patches are ensured. When dealing with the assembly surfaces of the robotic arm, topological constraints are strengthened to ensure assembly accuracy; for the connection points of the conveyor belt, the constraints are appropriately relaxed to improve modeling efficiency.

[0156] In this embodiment, a physically-based constraint mechanism is implemented when constructing the smoothness constraint. The smoothness of the surface is controlled by introducing a surface energy functional. When dealing with the functional surfaces of the robotic arm, the smoothness constraint is increased to ensure surface quality; for conventional components of the conveyor belt, the constraint strength is appropriately reduced to improve efficiency.

[0157] This embodiment employs a block-based iterative computation strategy in solving the objective function. Domain decomposition reduces the problem size and improves solution efficiency. For the complex structure of the robotic arm, a detailed block-based scheme is used; for the regular structure of the conveyor belt, a coarser block-based strategy is employed.

[0158] In this embodiment, a multi-criteria evaluation mechanism was developed for the fitted surface evaluation stage. By calculating the distance field distribution, both fitting accuracy and surface quality are comprehensively considered. For precision parts of the robotic arm, the focus is on local fitting errors; for conventional components of the conveyor belt, the emphasis is on maintaining the overall shape.

[0159] This embodiment implements an adaptive matching mechanism when constructing geometric matching criteria. By analyzing the characteristics of the slice contour lines, the matching threshold is dynamically adjusted. Strict matching conditions are applied at the critical joints of the robotic arm; for the transition area of ​​the conveyor belt, the matching requirements are appropriately relaxed.

[0160] This embodiment constructs a complete framework for surface fitting and optimization through the above implementation scheme. This scheme can accurately reconstruct the complex geometry of automated production line equipment and ensure that the reconstruction results meet engineering application requirements. In practical applications, this scheme has successfully handled various equipment models, including robotic arms, conveyor belts, and control cabinets, significantly improving the quality and efficiency of geometric reconstruction.

[0161] In the production line digital upgrade project, the solution in this embodiment provides reliable technical support for accurate modeling of production line equipment through precise surface fitting and effective optimization strategies, thus promoting the intelligent transformation of the production line. The optimized surface reconstruction results effectively support subsequent simulation analysis and intelligent control, providing crucial assurance for improving production line efficiency and quality.

[0162] In one embodiment of the automated production line digital model reconstruction method of this application, see [link to relevant documentation]. Figure 7 It can also specifically include the following:

[0163] Step S701: Calculate the splicing boundary of adjacent surface patches based on the geometric matching results, construct a transition region at the splicing boundary, establish a surface splicing equation using the geometric parameters of the transition region as constraints, solve the surface splicing equation using an iterative optimization algorithm to obtain the continuity parameters of the surface, construct an error compensation function based on the continuity parameters, and apply the error compensation function to the surface joint area to eliminate splicing gaps.

[0164] Step S702: Establish a mesh topology for the assembled reconstructed model, calculate the geometric quality evaluation index of the mesh topology, determine the areas that need to be refined based on the geometric quality evaluation index, construct mesh densification criteria in the refined areas, use an adaptive mesh densification algorithm to locally refine the reconstructed model, and convert the refined reconstructed model into a three-dimensional solid model for import into the automated production line environment.

[0165] Optionally, this embodiment employs an adaptive boundary recognition strategy during the splicing boundary calculation process. The boundary position is determined by analyzing the curvature continuity of the surface, taking into account the geometric characteristics of different equipment in the automated production line. For the joint connections of the robotic arm, the transition region range is determined based on motion characteristics; for the splicing points of the conveyor belt, the boundary width is determined based on force analysis.

[0166] In this embodiment, a shape control mechanism based on physical constraints is implemented during the construction of the transition region. By establishing curvature continuity and tangential continuity constraints, the smoothness of the transition region is ensured. When dealing with the precision mating surfaces of the robotic arm, high-order continuity constraints are used; for the ordinary mating surfaces of the control cabinet, low-order continuity constraints are used to simplify the calculation.

[0167] In this embodiment, a multi-objective optimization solution framework was developed for the surface splicing equation construction stage. By comprehensively considering geometric errors and surface quality, reasonable optimization objectives were established. For the functional surfaces of the robotic arm, the focus was on shape preservation; for the transition region of the conveyor belt, the emphasis was on smoothness optimization.

[0168] In this embodiment, a progressive optimization strategy is adopted when calculating continuous parameters. An adaptive step size control algorithm is designed to ensure the convergence of the optimization process. When dealing with complex curved surfaces of the robotic arm, a smaller optimization step size is used; for simple curved surfaces of the conveyor belt, a larger step size is used to accelerate convergence.

[0169] This embodiment implements a locally adaptive compensation mechanism in the error compensation function design. A reasonable compensation strategy is constructed by analyzing the geometric characteristics of the seam area. Precise error compensation is used when processing the precision parts of the robotic arm; a simplified compensation scheme is used for the conventional components of the conveyor belt.

[0170] In this embodiment, a hierarchical mesh generation strategy is adopted during the mesh topology construction process. The mesh density is dynamically adjusted by analyzing the geometric complexity of the model. For the irregular surfaces of the robotic arm, a fine mesh is used; for the planar structure of the control cabinet, a coarser mesh is used to improve efficiency.

[0171] In this embodiment, a multi-dimensional evaluation system was developed when designing the quality evaluation indicators. The reconstruction effect is comprehensively evaluated by calculating the shape quality, dimensional uniformity, and topological correctness of the mesh. For critical parts of the robotic arm, the focus is on local mesh quality; for regular areas of the conveyor belt, the focus is on overall uniformity.

[0172] In this embodiment, a feature-based adaptive mechanism is implemented in the grid encryption criterion construction stage. By analyzing the complexity of geometric features, a reasonable encryption threshold is determined. Strict encryption criteria are applied when dealing with the precision mating surfaces of the robotic arm; however, the encryption requirements are appropriately relaxed for the transition areas of the conveyor belt.

[0173] In this embodiment, a feature-preserving conversion strategy is adopted during the model conversion process. By maintaining the continuity and integrity of geometric features, the conversion quality is ensured. For the functional surfaces of the robotic arm, feature information is accurately preserved; for the regular structure of the control cabinet, a simplified conversion scheme is used to improve efficiency.

[0174] This embodiment constructs a complete model reconstruction and optimization framework through the above implementation scheme. This scheme can accurately reconstruct complex equipment models in automated production lines and ensure that the reconstruction results meet engineering application requirements. In practical applications, this scheme has successfully handled the reconstruction tasks of various equipment, including robotic arms, conveyor belts, and control cabinets, significantly improving the quality and reliability of digital models.

[0175] In the production line digital upgrade project, the solution in this embodiment provides reliable technical support for accurate modeling of production line equipment through precise surface splicing and effective mesh optimization. The optimized 3D solid model can accurately reflect the geometric features of the equipment, providing a high-quality geometric foundation for subsequent assembly analysis, collision detection, and motion simulation, effectively supporting the intelligent transformation and efficiency improvement of the production line.

[0176] To overcome the limitations of traditional model reconstruction and provide an efficient and reliable solution for industrial digital twins, this application provides an embodiment of an automated production line digital model reconstruction device for implementing all or part of the aforementioned automated production line digital model reconstruction method. See [link to embodiment]. Figure 8 The automated production line digital model reconstruction device specifically includes the following components:

[0177] The model slicing module 10 is used to slice the digital model of the automated production line along the principal coordinate axis, extract the feature point sequence of the slice outline at different resolutions, perform geometric semantic annotation on the feature point sequence, construct a variational level set function from the annotated feature point sequence, extract hierarchical feature mapping from the variational level set function using a deep convolutional network, establish an implicit surface reconstruction framework with physical constraints based on the hierarchical feature mapping, and embed the physical constraints as regularization conditions into the implicit surface reconstruction process.

[0178] The spatial partitioning module 20 is used to spatially partition the feature point sequence using an octree structure under the implicit surface reconstruction framework, construct a nearest neighbor search algorithm based on the spatial partitioning, map the nearest neighbor search algorithm to a parallel computing framework to establish an acceleration strategy, construct a local coordinate system based on the nearest neighbor search results, calculate the orthogonal basis vectors of the feature points in the local coordinate system using principal component analysis, establish an orthogonal basis space, decompose the feature point projection matrix in the orthogonal basis space into a three-dimensional covariance matrix, and input the three-dimensional covariance matrix into a noise filtering module for data augmentation.

[0179] The model reconstruction module 30 is used to construct local surface patches based on the eigenvectors of the three-dimensional covariance matrix, perform moving least squares fitting on the local surface patches using adaptive basis functions, add topological constraints and smoothness constraints during the fitting process, perform geometric matching between the fitted surface patches and the slice contour lines, assemble and splice the surface patches according to the geometric matching results, iteratively refine the assembled and spliced ​​model, establish an error compensation mechanism to eliminate surface seams, and import the compensated reconstructed model into the three-dimensional environment of the automated production line for collision detection.

[0180] As described above, the automated production line digital model reconstruction device provided in this application can extract geometric semantic features through hierarchical slicing and deep convolutional networks, innovatively constructing an implicit surface reconstruction framework with physical constraints. The system employs an octree structure and a parallel computing framework to achieve efficient nearest neighbor search, combining principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through adaptive basis function moving least squares fitting and an error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and it is seamlessly integrated with the collision detection function of the automated production line. This method overcomes the limitations of traditional model reconstruction, providing an efficient and reliable solution for industrial digital twins.

[0181] From a hardware perspective, in order to overcome the limitations of traditional model reconstruction and provide an efficient and reliable solution for industrial digital twins, this application provides an embodiment of an electronic device for implementing all or part of the aforementioned automated production line digital model reconstruction method. The electronic device specifically includes the following components:

[0182] The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the automated production line digital model reconstruction device and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the automated production line digital model reconstruction method and the automated production line digital model reconstruction device in the embodiments, the content of which is incorporated herein, and repeated details will not be described again.

[0183] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0184] In practical applications, parts of the digital model reconstruction method for automated production lines can be executed on the electronic device side as described above, or all operations can be completed in the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0185] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0186] Figure 9 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 9 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 9 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0187] In one embodiment, the automated production line digital model reconstruction method can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0188] Step S101: The digital model of the automated production line is sliced ​​along the principal coordinate axis. Feature point sequences of the slice contour lines are extracted at different resolutions. Geometric semantic annotations are performed on the feature point sequences. A variational level set function is constructed from the annotated feature point sequences. A deep convolutional network is used to extract hierarchical feature maps from the variational level set function. An implicit surface reconstruction framework with physical constraints is established based on the hierarchical feature maps. The physical constraints are embedded as regularization conditions in the implicit surface reconstruction process.

[0189] Step S102: Under the implicit surface reconstruction framework, an octree structure is used to spatially partition the feature point sequence, and a nearest neighbor search algorithm based on the spatial partition is constructed. The nearest neighbor search algorithm is mapped to a parallel computing framework to establish an acceleration strategy. A local coordinate system is constructed based on the nearest neighbor search results. The principal component analysis method is used to calculate the orthogonal basis vectors of the feature points in the local coordinate system to establish an orthogonal basis space. The feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix. The three-dimensional covariance matrix is ​​input into the noise filtering module for data augmentation.

[0190] Step S103: Construct local surface patches based on the eigenvectors of the three-dimensional covariance matrix, and perform moving least squares fitting on the local surface patches using adaptive basis functions. During the fitting process, topological constraints and smoothness constraints are added. Geometric matching is performed between the fitted surface patches and the slice contour lines. The surface patches are assembled and spliced ​​according to the geometric matching results. The assembled and spliced ​​model is iteratively refined. An error compensation mechanism is established to eliminate surface seams. The compensated reconstructed model is imported into the three-dimensional environment of the automated production line for collision detection.

[0191] As described above, the electronic device provided in this application innovatively constructs a physically constrained implicit surface reconstruction framework by extracting geometric semantic features through hierarchical slicing and deep convolutional networks. The system employs an octree structure and a parallel computing framework to achieve efficient nearest neighbor search, and combines principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through adaptive basis function moving least squares fitting and an error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and it is seamlessly integrated with the collision detection function of automated production lines. This method overcomes the limitations of traditional model reconstruction and provides an efficient and reliable solution for industrial digital twins.

[0192] In another embodiment, the automated production line digital model reconstruction device can be configured separately from the central processing unit 9100. For example, the automated production line digital model reconstruction device can be configured as a chip connected to the central processing unit 9100, and the automated production line digital model reconstruction method function can be realized through the control of the central processing unit.

[0193] like Figure 9 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 9 All components shown; in addition, the electronic device 9600 may also include Figure 9 For components not shown, please refer to existing technologies.

[0194] like Figure 9 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0195] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0196] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0197] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0198] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0199] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0200] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module 9110 (transmitter / receiver) is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored audio via the speaker 9131.

[0201] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the automated production line digital model reconstruction method with a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the automated production line digital model reconstruction method with a server or client as the execution subject in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0202] Step S101: The digital model of the automated production line is sliced ​​along the principal coordinate axis. Feature point sequences of the slice contour lines are extracted at different resolutions. Geometric semantic annotations are performed on the feature point sequences. A variational level set function is constructed from the annotated feature point sequences. A deep convolutional network is used to extract hierarchical feature maps from the variational level set function. An implicit surface reconstruction framework with physical constraints is established based on the hierarchical feature maps. The physical constraints are embedded as regularization conditions in the implicit surface reconstruction process.

[0203] Step S102: Under the implicit surface reconstruction framework, an octree structure is used to spatially partition the feature point sequence, and a nearest neighbor search algorithm based on the spatial partition is constructed. The nearest neighbor search algorithm is mapped to a parallel computing framework to establish an acceleration strategy. A local coordinate system is constructed based on the nearest neighbor search results. The principal component analysis method is used to calculate the orthogonal basis vectors of the feature points in the local coordinate system to establish an orthogonal basis space. The feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix. The three-dimensional covariance matrix is ​​input into the noise filtering module for data augmentation.

[0204] Step S103: Construct local surface patches based on the eigenvectors of the three-dimensional covariance matrix, and perform moving least squares fitting on the local surface patches using adaptive basis functions. During the fitting process, topological constraints and smoothness constraints are added. Geometric matching is performed between the fitted surface patches and the slice contour lines. The surface patches are assembled and spliced ​​according to the geometric matching results. The assembled and spliced ​​model is iteratively refined. An error compensation mechanism is established to eliminate surface seams. The compensated reconstructed model is imported into the three-dimensional environment of the automated production line for collision detection.

[0205] As described above, the computer-readable storage medium provided in this application innovatively constructs a physically constrained implicit surface reconstruction framework by extracting geometric semantic features through hierarchical slicing and deep convolutional networks. The system employs an octree structure and a parallel computing framework to achieve efficient nearest neighbor search, and combines principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through adaptive basis function moving least squares fitting and an error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and it is seamlessly integrated with the collision detection function of automated production lines. This method overcomes the limitations of traditional model reconstruction and provides an efficient and reliable solution for industrial digital twins.

[0206] Embodiments of this application also provide a computer program product capable of implementing all steps in the automated production line digital model reconstruction method with a server or client as the execution subject in the above embodiments. When executed by a processor, this computer program / instruction implements the steps of the automated production line digital model reconstruction method. For example, the computer program / instruction implements the following steps:

[0207] Step S101: The digital model of the automated production line is sliced ​​along the principal coordinate axis. Feature point sequences of the slice contour lines are extracted at different resolutions. Geometric semantic annotations are performed on the feature point sequences. A variational level set function is constructed from the annotated feature point sequences. A deep convolutional network is used to extract hierarchical feature maps from the variational level set function. An implicit surface reconstruction framework with physical constraints is established based on the hierarchical feature maps. The physical constraints are embedded as regularization conditions in the implicit surface reconstruction process.

[0208] Step S102: Under the implicit surface reconstruction framework, an octree structure is used to spatially partition the feature point sequence, and a nearest neighbor search algorithm based on the spatial partition is constructed. The nearest neighbor search algorithm is mapped to a parallel computing framework to establish an acceleration strategy. A local coordinate system is constructed based on the nearest neighbor search results. The principal component analysis method is used to calculate the orthogonal basis vectors of the feature points in the local coordinate system to establish an orthogonal basis space. The feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix. The three-dimensional covariance matrix is ​​input into the noise filtering module for data augmentation.

[0209] Step S103: Construct local surface patches based on the eigenvectors of the three-dimensional covariance matrix, and perform moving least squares fitting on the local surface patches using adaptive basis functions. During the fitting process, topological constraints and smoothness constraints are added. Geometric matching is performed between the fitted surface patches and the slice contour lines. The surface patches are assembled and spliced ​​according to the geometric matching results. The assembled and spliced ​​model is iteratively refined. An error compensation mechanism is established to eliminate surface seams. The compensated reconstructed model is imported into the three-dimensional environment of the automated production line for collision detection.

[0210] As described above, the computer program product provided in this application innovatively constructs a physically constrained implicit surface reconstruction framework by extracting geometric semantic features through hierarchical slicing and deep convolutional networks. The system employs an octree structure and a parallel computing framework to achieve efficient nearest neighbor search, and combines principal component analysis and covariance matrix decomposition for feature extraction and noise filtering. Through adaptive basis function moving least squares fitting and an error compensation mechanism, the accuracy and continuity of surface reconstruction are ensured, and it is seamlessly integrated with the collision detection function of automated production lines. This method overcomes the limitations of traditional model reconstruction and provides an efficient and reliable solution for industrial digital twins.

[0211] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0212] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0213] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0214] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 Figure 1 The steps of the function specified in one or more boxes.

[0215] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for reconstructing a digital model of an automated production line, characterized in that, The method includes: The digital model of the automated production line is sliced ​​along the principal coordinate axis. Feature point sequences of the slice contour lines are extracted at different resolutions. Geometric semantic annotations are performed on the feature point sequences. Variational level set functions are constructed from the annotated feature point sequences. Hierarchical feature maps are extracted from the variational level set functions using a deep convolutional network. An implicit surface reconstruction framework with physical constraints is established based on the hierarchical feature maps. The physical constraints are embedded as regularization conditions in the implicit surface reconstruction process. Within the implicit surface reconstruction framework, an octree structure is used to spatially partition the feature point sequence, and a nearest neighbor search algorithm based on the spatial partition is constructed. The nearest neighbor search algorithm is mapped to a parallel computing framework to establish an acceleration strategy. A local coordinate system is constructed based on the nearest neighbor search results, and the orthogonal basis vectors of the feature points in the local coordinate system are calculated using principal component analysis to establish an orthogonal basis space. The feature point projection matrix in the orthogonal basis space is decomposed into a three-dimensional covariance matrix, and the three-dimensional covariance matrix is ​​input into a noise filtering module for data augmentation. Local surface patches are constructed based on the eigenvectors of the three-dimensional covariance matrix. Adaptive basis functions are used to perform moving least squares fitting on the local surface patches. Topological constraints and smoothness constraints are added during the fitting process. The fitted surface patches are geometrically matched with the slice contour lines. The surface patches are assembled and spliced ​​according to the geometric matching results. The assembled and spliced ​​model is iteratively refined. An error compensation mechanism is established to eliminate surface seams. The compensated reconstructed model is imported into the three-dimensional environment of the automated production line for collision detection.

2. The method for reconstructing a digital model of an automated production line according to claim 1, characterized in that, The process of slicing the automated production line digital model along the principal coordinate axis, extracting feature point sequences of the slice contours at different resolutions, performing geometric semantic annotation on the feature point sequences, and constructing a variational level set function from the annotated feature point sequences includes: A layered scanning algorithm based on a 3D mesh model is used to calculate the slicing direction of the principal coordinate axis. Based on the slicing direction, a slicing plane parallel to the coordinate axis is constructed. The slicing plane is intersected with the boundary surface of the digital model to generate a slicing contour line. The slicing contour line is meshed to calculate the spatial coordinates of the feature points. An adaptive resolution sampling mechanism is established to filter the feature points to obtain an ordered sequence of feature points. A curvature tensor analysis model is established for the feature point sequence. The geometric properties of the feature points are calculated based on the principal direction of the curvature tensor. The geometric properties are assigned as label information to the feature points. A variational objective functional with weight coefficients is constructed. The mathematical expression of the level set function is obtained by solving the variational objective functional. The mapping relationship between the feature point sequence and the level set function is established.

3. The method for reconstructing a digital model of an automated production line according to claim 1, characterized in that, The process involves using a deep convolutional network to extract hierarchical feature maps from the variational level set function, establishing an implicit surface reconstruction framework with physical constraints based on these hierarchical feature maps, and embedding these physical constraints as regularization conditions into the implicit surface reconstruction process. This includes: A multi-scale convolutional neural network structure is constructed. Convolutional kernels of different sizes are set in the convolutional layer to extract the spatial features of the variational level set function. The feature map is reduced in dimension by pooling operation. The reduced feature map is input into the deconvolutional layer for upsampling. The upsampling result is concatenated and fused with the feature map of the skip connection to generate a hierarchical feature map. An implicit surface equation is constructed based on the hierarchical feature mapping. Physical constraints for surface smoothness and shape preservation are added to the implicit surface equation. The physical constraints are transformed into regularization conditions in the form of Lagrange multipliers. An objective function with regularization terms is established. The expression for implicit surface reconstruction is obtained by solving the objective function using the variational method.

4. The method for reconstructing a digital model of an automated production line according to claim 1, characterized in that, Within the implicit surface reconstruction framework, an octree structure is used to spatially partition the feature point sequence, and a nearest neighbor search algorithm based on this spatial partitioning is constructed. An acceleration strategy is then established by mapping this nearest neighbor search algorithm to a parallel computing framework, including: Calculate the spatial bounding box of the feature point sequence, divide the spatial bounding box into eight subspaces according to the coordinate axis direction, calculate the density distribution of feature points in each subspace, recursively subdivide the subspaces based on the density distribution threshold, establish an octree index structure to store the spatial location information of the feature points, and record the spatial location information of the feature points in the corresponding leaf nodes. In the octree index structure, a nearest neighbor search path is constructed, the spatial distance between the feature point to be searched and the leaf node is calculated, the leaf node whose spatial distance is less than the search radius is marked as a candidate node, a parallel search task is established for the candidate node, the parallel search task is assigned to the computing unit to perform the nearest neighbor search calculation, and the search results are summarized to obtain the neighborhood information of the feature point.

5. The method for reconstructing a digital model of an automated production line according to claim 1, characterized in that, The process involves constructing a local coordinate system based on the nearest neighbor search results, calculating the orthogonal basis vectors of feature points in the local coordinate system using principal component analysis, establishing an orthogonal basis space, decomposing the feature point projection matrix in the orthogonal basis space into a three-dimensional covariance matrix, and inputting the three-dimensional covariance matrix into a noise filtering module for data augmentation, including: The centroid coordinates of the neighborhood of a feature point are calculated using the nearest neighbor search results. The centroid coordinates are then translated to the origin to establish the origin of the local coordinate system. The divergence matrix of the neighborhood of the feature point is calculated. The eigenvalue decomposition of the divergence matrix is ​​performed to obtain the eigenvector matrix. The eigenvector matrix is ​​sorted according to the size of the eigenvalues ​​to determine the three basis vectors of the local coordinate system. The rotation matrix of the local coordinate system is then constructed based on the basis vectors. The feature points are transformed from the global coordinate system to the local coordinate system, the coordinate components of the feature points in the local coordinate system are calculated, the projection matrix of the feature points is constructed, the singular value decomposition of the projection matrix is ​​performed to obtain the three-dimensional covariance matrix, the noise threshold is calculated based on the eigenvalue distribution of the three-dimensional covariance matrix, and the noise threshold is used as a filtering parameter input to the anisotropic diffusion filter.

6. The method for reconstructing a digital model of an automated production line according to claim 1, characterized in that, The process involves constructing local surface patches based on the eigenvectors of the three-dimensional covariance matrix, performing moving least squares fitting on the local surface patches using adaptive basis functions, incorporating topological and smoothness constraints during the fitting process, and geometrically matching the fitted surface patches with the slice contour lines, including: The eigenvectors of the three-dimensional covariance matrix are used as normal vectors to construct the parametric equations of the local surface patch. The principal curvature and Gaussian curvature of the local surface patch are calculated. The order of the adaptive basis function is selected based on the curvature information. A moving least squares equation with a weight function is constructed. The topological continuity constraint and curvature compliance constraint of the surface are introduced into the moving least squares equation. A weighted least squares objective function with constraint terms is established. An analytical solution is constructed for the weighted least squares objective function. The objective function is numerically solved using the domain decomposition method. The solution is substituted into the parametric equation of the local surface patch to obtain the fitted surface. The distance field distribution between the fitted surface and the slice contour is calculated. A geometric matching criterion is constructed based on the distance field. The geometric matching criterion is used as an evaluation index to select the optimal fitted surface.

7. The method for reconstructing a digital model of an automated production line according to claim 1, characterized in that, The process of assembling and splicing the surface patches according to the geometric matching results, iteratively refining the assembled model, establishing an error compensation mechanism to eliminate surface seams, and importing the compensated reconstructed model into the 3D environment of an automated production line for collision detection includes: The splicing boundary of adjacent surface patches is calculated based on the geometric matching results. A transition region is constructed at the splicing boundary. The geometric parameters of the transition region are used as constraints to establish the surface splicing equation. The surface splicing equation is solved by an iterative optimization algorithm to obtain the continuity parameters of the surface. An error compensation function is constructed based on the continuity parameters. The error compensation function is applied to the surface joint area to eliminate the splicing gap. A mesh topology is established for the assembled reconstructed model, and the geometric quality evaluation index of the mesh topology is calculated. Based on the geometric quality evaluation index, the areas that need to be refined are determined. Mesh densification criteria are constructed in the refined areas, and an adaptive mesh densification algorithm is used to locally refine the reconstructed model. The refined reconstructed model is then converted into a three-dimensional solid model and imported into an automated production line environment.

8. An automated production line digital model reconstruction device, characterized in that, The device includes: The model slicing module is used to slice the digital model of the automated production line along the principal coordinate axis, extract the feature point sequence of the slice outline at different resolutions, perform geometric semantic annotation on the feature point sequence, construct a variational level set function from the annotated feature point sequence, extract hierarchical feature maps from the variational level set function using a deep convolutional network, establish an implicit surface reconstruction framework with physical constraints based on the hierarchical feature maps, and embed the physical constraints as regularization conditions into the implicit surface reconstruction process. The spatial partitioning module is used to spatially partition the feature point sequence using an octree structure within the implicit surface reconstruction framework, construct a nearest neighbor search algorithm based on the spatial partitioning, map the nearest neighbor search algorithm to a parallel computing framework to establish an acceleration strategy, construct a local coordinate system based on the nearest neighbor search results, calculate the orthogonal basis vectors of the feature points in the local coordinate system using principal component analysis, establish an orthogonal basis space, decompose the feature point projection matrix in the orthogonal basis space into a three-dimensional covariance matrix, and input the three-dimensional covariance matrix into a noise filtering module for data augmentation. The model reconstruction module is used to construct local surface patches based on the eigenvectors of the three-dimensional covariance matrix, perform moving least squares fitting on the local surface patches using adaptive basis functions, add topological constraints and smoothness constraints during the fitting process, perform geometric matching between the fitted surface patches and the slice contour lines, assemble and stitch the surface patches according to the geometric matching results, iteratively refine the assembled and stitched model, establish an error compensation mechanism to eliminate surface seams, and import the compensated reconstructed model into the three-dimensional environment of the automated production line for collision detection.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the automated production line digital model reconstruction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the automated production line digital model reconstruction method according to any one of claims 1 to 7.

Images