Automated production line digital model simplification method and apparatus

Abstract

The embodiment of the application provides a kind of automatic production line digital model simplification method and device, pass through salient feature extraction module identification and save key feature information, establish four levels topological structure tree.Adaptively designed feature weight computing mechanism, combined with multi-resolution wavelet analysis realizes the intelligent simplification of grid.System is extracted and matched topological structure feature and local geometric feature under different scales by multi-scale feature fusion module, realizes the accurate maintenance of key feature in model simplification process.The method breaks through the limitation of traditional model simplification, and provides intelligent and automated solution for efficient processing of large-scale digital model.

Description

TECHNICAL FIELD

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

[0002] Traditional digital model simplification methods mainly use uniform mesh compression algorithms, which are difficult to preserve key features of the model and have low computational efficiency when dealing with complex automatic production line models. Existing technologies lack adaptive mechanisms in feature extraction and weight distribution, and cannot simplify different regions according to their importance.

[0003] At the same time, existing systems have obvious shortcomings in multi-scale feature processing and topological structure preservation. Traditional methods often ignore the hierarchical relationship of the model, resulting in distortion of the simplified model or loss of key structures. The system is also relatively simple in feature fusion and node optimization, and fails to achieve the optimal balance between model simplification and feature preservation.

[0004] In addition, existing technologies also have limitations in mesh reconstruction and quality control. There is a lack of adaptive simplification strategies based on salient features, and the advantages of multi-resolution analysis are not fully utilized. Solving these problems is of great significance to improving the processing efficiency and visual quality of large-scale digital models. SUMMARY

[0005] To solve the problems in the prior art, the present application provides an automatic production line digital model simplification method and device, which can break through the limitations of traditional model simplification and provide intelligent and automated solutions for efficient processing of large-scale digital models.

[0006] To solve at least one of the above problems, the present application provides the following technical solutions:

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

[0008] A salient feature extraction module is constructed, which calculates saliency scores for point, line, surface and volume data in the automatic production line digital model, extracts feature points with saliency scores exceeding a first threshold as a set of salient feature points, and stores morphological parameters and position parameters corresponding to the set of salient feature points in a salient feature database. A four-level topological structure tree is established based on the set of salient feature points;

[0009] The four-level topology tree is input into an adaptive feature weight calculation module. The adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the set of salient feature points. The weight coefficient is multiplied by feature data of the corresponding level to obtain weighted feature data. The weighted feature data is input into an adaptive octree grid construction module. The adaptive octree grid construction module constructs a grid node based on the weighted feature data. A multi-resolution wavelet basis function is embedded in the grid node. The wavelet basis function coefficients are processed by hierarchical thresholding. The thresholded wavelet coefficients are reconstructed into an initial simplified grid.

[0010] The initial simplified grid is input into a multi-scale feature fusion module. The multi-scale feature fusion module extracts a topology structure feature of the initial simplified grid at a coarse scale and extracts a local geometric feature of the initial simplified grid at a fine scale. A feature correspondence graph is constructed based on the topology structure feature and the local geometric feature. The feature correspondence graph is matched with a shape parameter and a position parameter in a salient feature database. The initial simplified grid is adjusted based on a matching result. A final simplified model is generated.

[0011] Further, the saliency feature extraction module is constructed. The saliency feature extraction module calculates a saliency score for point, line, surface, and volume data in the automated production line digital model. Feature points with a saliency score exceeding a first threshold value are extracted as a set of salient feature points.

[0012] The automated production line digital model is discretized to obtain point cloud data. A three-dimensional grid structure is established based on the point cloud data. A curvature value, a normal vector, and a neighborhood density of each vertex in the three-dimensional grid structure are calculated. The curvature value, the normal vector, and the neighborhood density are input into a saliency calculation formula to obtain a saliency coefficient. A saliency feature extraction module is constructed based on the saliency coefficient.

[0013] The saliency coefficient in the saliency feature extraction module is compared with a first threshold value. Vertices with a saliency coefficient greater than the first threshold value are extracted as feature points. A set of salient feature points is constructed based on a spatial position relationship of the feature points. Geometric attribute parameters and topological attribute parameters of the set of salient feature points are stored as a shape parameter and a position parameter, respectively.

[0014] Further, the shape parameter and the position parameter corresponding to the set of salient feature points are stored in a salient feature database. A four-level topology tree is constructed based on the set of salient feature points.

[0015] The shape parameters and the position parameters of the set of significant feature points are input into a feature encoder, the feature encoder performs geometric feature encoding on the shape parameters to obtain a geometric feature vector, performs spatial position encoding on the position parameters to obtain a position feature vector, combines the geometric feature vector and the position feature vector to form feature data, and writes the feature data into a significant feature database according to a hierarchical structure.

[0016] A four-level topological structure tree is constructed based on the set of significant feature points, point data in the set of significant feature points is taken as a first-level node, a connection relationship between adjacent points is taken as a second-level node, a face patch enclosed by a closed line segment is taken as a third-level node, and a body element formed by a closed face patch is taken as a fourth-level node, and corresponding feature data indexes are stored in each node of the four-level topological structure tree.

[0017] Further, the four-level topological structure tree is input into an adaptive feature weight calculation module, the adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the set of significant feature points, multiplies the weight coefficient by feature data of the corresponding level to obtain weighted feature data, and the method comprises the following steps:

[0018] A spatial density distribution map is constructed based on the four-level topological structure tree, a number of significant feature points in a neighborhood range of each node in the spatial density distribution map is calculated, the number is divided by a neighborhood volume to obtain a local density value, the local density value is normalized to obtain a density distribution coefficient, and the density distribution coefficient is input into a weight mapping function to generate a corresponding weight coefficient of each level;

[0019] The weight coefficient and feature data of each level node in the four-level topological structure tree are subjected to matrix multiplication to generate a weighted feature matrix, the weighted feature matrix is subjected to standardization processing to obtain weighted feature data, and the weighted feature data is reorganized into a tree structure according to a hierarchical relationship.

[0020] Further, the weighted feature data is input into an adaptive octree grid construction module, the adaptive octree grid construction module constructs a grid node based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid node, performs hierarchical thresholding processing on a wavelet basis function coefficient, and reconstructs the thresholded wavelet coefficient into an initial simplified grid, and the method comprises the following steps:

[0021] determining a spatial bounding box based on the weighted feature data, recursively partitioning the spatial bounding box into an octree structure, calculating a local variance of the weighted feature data at each node of the octree structure, continuing to subdivide the node when the local variance of the node is greater than a second threshold value until a preset number of subdivision levels is reached or the local variance is less than the second threshold value, constructing an adaptive mesh on the octree node where the subdivision is completed, and mapping a Haar wavelet basis function to a node coordinate of the adaptive mesh;

[0022] performing coefficient decomposition on the Haar wavelet basis function on the adaptive mesh node, calculating an energy value of each layer of wavelet coefficients, setting an adaptive threshold value according to the energy value, setting wavelet coefficients less than the adaptive threshold value to zero, and performing a reconstruction operation on the remaining wavelet coefficients to generate an initial simplified mesh.

[0023] Further, the initial simplified mesh is input into a multi-scale feature fusion module, the multi-scale feature fusion module extracts topological structure features of the initial simplified mesh at a coarse scale and local geometric features of the initial simplified mesh at a fine scale, including:

[0024] performing multi-level decomposition on the initial simplified mesh to obtain mesh representations of different scales, constructing a mesh connectivity graph at a coarse scale level, calculating a degree distribution and connectivity property of each node in the mesh connectivity graph, constructing a mesh skeleton based on the degree distribution and connectivity property, and extracting a branch structure and connection relationship of the mesh skeleton as topological structure features;

[0025] calculating a local curvature tensor of the initial simplified mesh node at a fine scale level, calculating a principal curvature value and a principal curvature direction based on the curvature tensor, combining the principal curvature value and the principal curvature direction to form a curvature descriptor, performing clustering analysis on the curvature descriptor to obtain a local geometric feature, and establishing a mapping relationship between the local geometric feature and the mesh node.

[0026] Further, a feature correspondence graph is constructed based on the topological structure features and the local geometric features, the feature correspondence graph is matched with a shape parameter and a position parameter in the salient feature database, and a final simplified model is generated by adjusting the nodes of the initial simplified mesh according to the matching result, including:

[0027] input the topological structure features and the local geometric features into a feature fusion network, the feature fusion network generates a feature correspondence graph based on graph convolution operation, calculates a similarity matrix between nodes in the feature correspondence graph, performs matching operation on the similarity matrix and morphological parameters and position parameters stored in a significant feature database, generates a feature matching degree measurement value, and constructs a node optimization objective function based on the feature matching degree measurement value;

[0028] iteratively solves the node optimization objective function, adjusts node positions and connection relationships of the initial simplified grid, stops iteration when a target function value is less than a third threshold value or a maximum iteration number is reached, reconstructs final grid node coordinates and connection relationships into a three-dimensional model, and performs grid smoothing processing on the reconstructed three-dimensional model to obtain a final simplified model.

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

[0030] A feature extraction module is configured to construct a saliency feature extraction module, calculate saliency scores of point, line, surface and volume data in an automatic production line digital model, extract feature points with saliency scores exceeding a first threshold value as a set of significant feature points, store morphological parameters and position parameters corresponding to the set of significant feature points into a significant feature database, and establish a four-level topological structure tree based on the set of significant feature points.

[0031] An architecture building module is configured to input the four-level topological structure tree into an adaptive feature weight calculation module, calculate weight coefficients of each level based on spatial distribution density of the set of significant feature points, multiply the weight coefficients by feature data of the corresponding level to obtain weighted feature data, and input the weighted feature data into an adaptive octree grid construction module, which constructs grid nodes based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid nodes, performs hierarchical thresholding processing on coefficients of the wavelet basis function, and reconstructs the thresholded wavelet coefficients into an initial simplified grid.

[0032] A model simplification module is configured to input the initial simplified grid into a multi-scale feature fusion module, extract topological structure features of the initial simplified grid at a coarse scale and local geometric features of the initial simplified grid at a fine scale, construct a feature correspondence graph based on the topological structure features and the local geometric features, perform matching on the feature correspondence graph and morphological parameters and position parameters in the significant feature database, adjust nodes of the initial simplified grid according to the matching result, and generate a final simplified model.

[0033] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the automatic production line digital model simplification method when executing the program.

[0034] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein the computer program is executable on a processor to implement the steps of the automatic production line digital model simplification method.

[0035] In a fifth aspect, the present application provides a computer program product, comprising computer programs / instructions, wherein the computer programs / instructions are executable on a processor to implement the steps of the automatic production line digital model simplification method.

[0036] From the above technical solutions, the present application provides an automatic production line digital model simplification method and device, which identifies and saves key feature information through a saliency feature extraction module, and establishes a four-level topological structure tree. An adaptive feature weight calculation mechanism is innovatively designed, and intelligent simplification of the grid is realized in combination with multi-resolution wavelet analysis. The system extracts and matches topological structure features and local geometric features at different scales through a multi-scale feature fusion module, and realizes accurate preservation of key features in the model simplification process. The method breaks through the limitations of traditional model simplification, and provides an intelligent and automatic solution for efficient processing of large-scale digital models. 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 in the embodiments or prior art description will be briefly introduced below. 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 Figure 1 is a flowchart of an automatic production line digital model simplification method according to an embodiment of the present application;

[0039] Figure 2 Figure 2 is a flowchart of an automatic production line digital model simplification method according to another embodiment of the present application;

[0040] Figure 3 Figure 3 is a flowchart of an automatic production line digital model simplification method according to another embodiment of the present application;

[0041] Figure 4 Figure 4 is a flowchart of an automatic production line digital model simplification method according to another embodiment of the present application;

[0042] Figure 5Fig. 5 is a flowchart of a seventh embodiment of the method for simplifying a digital model of an automated production line according to the present application;

[0043] Figure 6 Fig. 6 is a flowchart of a sixth embodiment of the method for simplifying a digital model of an automated production line according to the present application;

[0044] Figure 7 Fig. 7 is a flowchart of a seventh embodiment of the method for simplifying a digital model of an automated production line according to the present application;

[0045] Figure 8 Fig. 8 is a block diagram of a device for simplifying a digital model of an automated production line according to the present application;

[0046] Figure 9 Fig. 9 is a block diagram of an electronic device according to 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 following will be combined with the accompanying drawings for the embodiments of the present application to make a clear and complete description of the technical solutions 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, and the like of data in the technical solutions of the present application comply with relevant provisions in national laws and regulations.

[0051] In view of the problems in the prior art, the present application provides a method and device for simplifying a digital model of an automated production line. A salient feature extraction module is used to identify and save key feature information, and a four-level topological structure tree is established. An adaptive feature weight calculation mechanism is innovatively designed, and a multi-resolution wavelet analysis is combined to realize intelligent simplification of a grid. A multi-scale feature fusion module is used to extract and match topological structure features and local geometric features under different scales, so that the key features are accurately maintained during the model simplification process. The method breaks through the limitations of traditional model simplification and provides an intelligent and automated solution for efficient processing of large-scale digital models.

[0052] In order to break through the limitations of traditional model simplification and provide an intelligent and automatic solution for efficient processing of large-scale digital models, an embodiment of an automatic production line digital model simplification method is provided, as shown in Figure 1 The automatic production line digital model simplification method specifically includes the following contents:

[0053] Step S101: Construct a saliency feature extraction module, which calculates saliency scores for point, line, surface, and volume data in the automatic production line digital model. Feature points with saliency scores exceeding a first threshold value are extracted as a set of salient feature points. The morphological parameters and position parameters corresponding to the set of salient feature points are stored in a salient feature database. A four-level topological structure tree is established based on the set of salient feature points.

[0054] Optionally, in the embodiment, a multi-level feature extraction framework is developed for the complex spatial structure features of the automatic production line. First, an improved octree spatial partitioning algorithm is used to divide the entire production line digital model into spatial sub-regions of different sizes. In the spatial partitioning process, an adaptive subdivision strategy based on local feature density is introduced. For regions containing key connection points, valves, sensors, and other important components, more detailed partitioning is used, with a partitioning accuracy of millimeters. For regions with relatively simple features such as conveyor belts and pipelines, coarser partitioning is used. This differentiated spatial partitioning strategy effectively balances computational efficiency and feature extraction accuracy.

[0055] In the point cloud sampling stage, an importance-based adaptive sampling algorithm is implemented. For each spatial sub-region, first calculate the local curvature variation rate and shape complexity, and input these indicators into the importance evaluation function. The importance evaluation function uses a weighted combination form, with a curvature variation rate weight of 0.4, a shape complexity weight of 0.3, and a spatial position importance weight of 0.3. Based on the evaluation results, dynamically adjust the sampling density. In areas with high importance, use denser sampling point distribution, with a sampling interval of 0.1mm. In areas with low importance, the sampling interval can be relaxed to 10mm, ensuring the integrity of key features while significantly reducing data redundancy.

[0056] In the aspect of feature point saliency calculation, a multi-dimensional feature fusion calculation framework is developed. Firstly, local geometric features are extracted, including principal curvatures k1, k2, Gaussian curvature K, average curvature H, shape index S and curvature variation rate C. At the same time, local shape distribution features of each point are calculated, a shape descriptor D=(λ1-λ2) / (λ1-λ3) is constructed by analyzing eigenvalues λ1, λ2, λ3 of the point cloud covariance matrix, and is used to represent the shape characteristics of the region where the point is located. For line structure features, continuity index and curvature variation cumulative value of tangent vector are calculated. For face structure features, variation gradient of normal vector and boundary curvature distribution of face sheet are analyzed. For body structure features, local volume density and shape complexity index are calculated.

[0057] In the feature fusion process, an adaptive weight allocation mechanism is realized. By analyzing the discriminability and stability of different types of features, a feature importance evaluation model is established. The model considers factors such as scale invariance, noise robustness and discriminability of features. For pipeline systems, the weight of curvature feature is set to 0.4, the weight of shape feature is 0.3, and the weight of position feature is 0.3; for box structures, the weight of volume feature is 0.4, the weight of boundary feature is 0.3, and the weight of topological feature is 0.3. The comprehensive saliency score S of each point is obtained by weighted fusion:

[0058] S=Σ(wi*fi), where wi is the weight of the ith feature, and fi is the normalized feature value.

[0059] In the feature point extraction link, an adaptive clustering algorithm based on region growing is developed. Firstly, the point with the highest saliency score in each spatial sub-region is selected as the seed point, and then the region is gradually expanded based on spatial proximity and feature similarity. A dynamic threshold mechanism is introduced in the expansion process, and the initial threshold is set to the mean value of the saliency score, which is adjusted gradually as the region expands. When the feature variance of the points in the region exceeds 1.5 times of the threshold, the growth is stopped. This adaptive stopping criterion effectively avoids the problem of over-clustering.

[0060] In the morphological parameter extraction process, a multi-scale feature description strategy is adopted. At the local scale (1-5mm), the density distribution and normal vector consistency of the point set are calculated; at the medium scale (5-50mm), boundary curvature and surface roughness are extracted; at the global scale (>50mm), overall shape features and topological properties are analyzed. Through this multi-scale feature description method, the geometric characteristics of the feature point set are fully described.

[0061] In the position parameter calculation link, the embodiment realizes the position description mechanism of multiple reference systems. First, a global coordinate system of the production line is established, the origin is set at the entrance of the production line, the z-axis is vertically upward, and the x-axis is along the main conveying direction of the production line. Then, a local coordinate system is established for each main device, and the conversion relationship with the global coordinate system is recorded through the coordinate transformation matrix. The position of the feature point is represented by the global coordinates (X, Y, Z) and the local coordinates (x, y, z) relative to the nearest device. This representation method is convenient for feature matching and model simplification.

[0062] In the construction of the four-level topological structure tree, a bottom-up hierarchical construction algorithm is developed. The first level stores the three-dimensional coordinates and the saliency score of the original feature points; the second level constructs edge features by analyzing the distance relationship and feature similarity between points, and records the length, direction and curvature of the edge; the third level constructs face features based on the boundary closure analysis, and stores the normal vector, area and boundary feature of the face; the fourth level constructs body features through the closure analysis of the face, and records the volume, barycenter and main direction.

[0063] In terms of data storage, the embodiment adopts a hierarchical database structure. A feature index table is established as the main table, which contains feature ID, type and level information. The shape parameter table stores geometric feature data, and the position parameter table records spatial position information. Efficient query and update of data are realized through foreign key association. At the same time, a data compression mechanism based on compressed sensing is realized, which can store repetitive feature data in a merged manner, and the data compression rate can reach more than 60%.

[0064] Through the above implementation scheme, the embodiment establishes an accurate saliency feature extraction system. This scheme is particularly suitable for digital model simplification of complex automated production lines, and can accurately identify and retain key features in the model. Through multi-dimensional feature analysis and hierarchical data organization, a reliable feature basis is provided for subsequent model simplification. Practical application shows that this scheme can effectively extract key structural features in the production line, providing strong support for equipment monitoring, fault diagnosis and other applications.

[0065] Step S102: input the four-level topological structure tree into an adaptive feature weight calculation module. The adaptive feature weight calculation module calculates the weight coefficient of each level based on the spatial distribution density of the salient feature point set, multiplies the weight coefficient by the feature data of the corresponding level to obtain weighted feature data, and inputs the weighted feature data into an adaptive octree grid construction module. The adaptive octree grid construction module constructs grid nodes based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid nodes, performs hierarchical thresholding processing on the wavelet basis function coefficients, and reconstructs the thresholded wavelet coefficients into an initial simplified grid.

[0066] Optionally, in the adaptive feature weight calculation process of the embodiment, first, depth-first traversal is performed on the four-level topology structure tree, and a local feature statistical window is established at each level node. Different sizes of feature statistical windows are set for different functional areas in the automated production line, such as the conveyor belt system, the mechanical arm workstation, the detection station, and the like, to ensure the locality and accuracy of feature analysis.

[0067] In the embodiment, the local density value is calculated by analyzing the distribution rule of the feature points in the three-dimensional space. For each feature statistical window, the kernel density estimation method is adopted, and the adaptive bandwidth Gaussian kernel function is selected for density calculation. At the key nodes of the production line, such as the device connection and the turning device, a higher feature point density is often shown, and the feature retention of these areas is crucial for model simplification.

[0068] In the weight coefficient calculation process of the embodiment, an adaptive weight distribution mechanism based on the density gradient is realized. For the point features of the first level, the weight coefficient is positively correlated with the local density value; for the line features of the second level, the weight coefficient considers both the endpoint density and the line curvature; for the face features of the third level, the weight coefficient combines the feature point density and the normal vector variation of the face; and for the volume features of the fourth level, the weight coefficient considers both the density distribution and the shape complexity of the volume.

[0069] In the feature weighting process of the embodiment, a normalization processing mechanism for feature data is developed. First, the feature data of each level is standardized to eliminate the influence of different feature dimensions. Then, the weight coefficient obtained by calculation is multiplied by the standardized feature data to obtain weighted feature data. This weighting method ensures the reasonable retention of features of different levels in the simplification process.

[0070] In the construction of the adaptive octree grid of the embodiment, a recursive subdivision strategy based on weighted features is realized. First, an initial octree structure is established in the bounding box of the production line model, and then whether further subdivision is needed is determined according to the weighted feature value in each node. For areas with high weighted feature values, such as the joints of mechanical arms, the installation positions of sensors, and the like, more detailed grid division is performed.

[0071] In the grid node construction process of the embodiment, an adaptive node arrangement strategy is adopted. In areas with dramatic feature changes, such as device boundaries and irregular surfaces, the node density is increased; and in areas with gentle feature changes, such as flat panels and straight pipelines, the node density is appropriately reduced. This differentiated node arrangement strategy not only ensures the accurate expression of key features, but also reduces data redundancy.

[0072] The embodiment embeds Haar wavelet basis functions in the grid nodes, and realizes multi-resolution feature expression. On each grid node, a wavelet basis function of a suitable scale is selected according to the complexity of the local feature. A high-frequency wavelet basis function is used for a region with rich details, and a low-frequency wavelet basis function is used for a smooth region, so as to realize multi-scale decomposition of the feature.

[0073] In the wavelet coefficient thresholding process, the embodiment develops an adaptive threshold selection strategy. By analyzing the energy distribution of the wavelet coefficients, a threshold determination method based on the signal-to-noise ratio is established. Different thresholds are set for wavelet coefficients of different levels to ensure the retention of important features and the removal of secondary features. In the thresholding process, the correlation between coefficients of adjacent scales is considered to avoid feature mutation and discontinuity.

[0074] Finally, the embodiment simplifies the grid by wavelet inverse transform reconstruction. In the reconstruction process, a progressive feature recovery strategy is adopted. First, the main features are reconstructed, and then the secondary features are gradually added until the preset simplification accuracy is reached. The reconstructed grid retains the key features of the original model while significantly reducing the data volume.

[0075] Through the above implementation scheme, the embodiment establishes a complete feature weighting and grid simplification system. This scheme is particularly suitable for digital model simplification of complex automated production lines and can adaptively retain key structures according to the importance of features, providing an efficient data foundation for subsequent model applications. Practical applications show that this scheme can effectively reduce model complexity while maintaining the integrity of key features of the production line, providing reliable digital model support for equipment monitoring and maintenance.

[0076] Step S103: input the initial simplified grid into a multi-scale feature fusion module. The multi-scale feature fusion module extracts topological structure features of the initial simplified grid at a coarse scale and extracts local geometric features of the initial simplified grid at a fine scale, constructs a feature correspondence graph based on the topological structure features and the local geometric features, matches the feature correspondence graph with the shape parameters and position parameters in the significant feature database, adjusts the nodes of the initial simplified grid according to the matching result, and generates a final simplified model.

[0077] Optionally, in the multi-scale feature fusion process, the embodiment adopts an adaptive scale decomposition strategy for the initial simplified grid. For different functional units in the automated production line, such as conveyor belt systems, mechanical arms, control cabinets, etc., different feature analysis scales are adopted. In coarse scale analysis, the feature analysis window covers the entire functional unit; in fine scale analysis, the feature analysis window focuses on key components and connection structures.

[0078] The embodiment realizes a hierarchical topological structure analysis method in the coarse-scale feature extraction. First, a grid connectivity graph is constructed, where the nodes represent the grid cells and the edges represent the connection relationship between the cells. By analyzing the degree distribution characteristics of the connectivity graph, key connection points and functional nodes in the production line are identified. For example, the joints of a mechanical arm often exhibit high values, while the straight segments of a conveyor belt exhibit lower values.

[0079] The embodiment analyzes the global structural characteristics of the connectivity graph through a feature vector decomposition method. The eigenvalues and eigenvectors of the Laplacian matrix are calculated, and the grid is divided into sub-regions with similar topological characteristics using a spectral clustering method. This spectral analysis-based method can effectively identify the boundaries of functional modules and key connection structures in the production line.

[0080] In the fine-scale feature extraction link, the embodiment develops a local geometric feature analysis framework. By calculating the curvature tensor of each grid node, the principal curvature values and directions are obtained. For irregular surfaces in the production line, such as the groove structure of a guide rail and the mounting seat of a sensor, these local geometric features have important recognition value.

[0081] In the feature fusion process, the embodiment realizes a feature correspondence relationship learning mechanism based on graph neural networks. The coarse-scale topological features and fine-scale geometric features are respectively encoded into feature vectors and input into the graph neural network. The network gradually learns the correlation patterns between features through multi-layer graph convolution operations. In the convolution process, an attention mechanism is used to highlight the influence of important features and weaken the contribution of secondary features.

[0082] In constructing the feature correspondence graph, the embodiment adopts a multi-level feature matching strategy. First, the correspondence relationship between functional modules is established at the coarse-scale level, and then the matching of local features is improved at the fine-scale level. By calculating the similarity matrix between feature vectors, a soft correspondence relationship between feature points is established, avoiding the limitations of traditional hard matching methods.

[0083] In the feature matching process, the embodiment develops a reference matching mechanism based on a salient feature database. The node features in the feature correspondence graph are compared with the shape parameters and position parameters stored in the database. In the matching process, a fuzzy matching strategy is used, allowing a certain range of feature deviation, which improves the robustness of the matching.

[0084] In the grid node adjustment stage, the embodiment realizes an optimization-based node position refinement method. An objective function for node position optimization is constructed, which considers both feature preservation constraints and grid quality constraints. The feature preservation constraint ensures that the adjusted node position is consistent with the matched reference features, and the grid quality constraint avoids the generation of distorted cells.

[0085] The embodiment solves the node position by an iterative optimization algorithm. In each iteration, the deviation of the current mesh from the target feature is first calculated, and then the node position is updated based on the gradient information. To improve the optimization efficiency, an adaptive step strategy is adopted, which uses a larger step size in the early stage of optimization to quickly converge, and a smaller step size in the later stage to fine-tune.

[0086] The embodiment finally performs smoothing processing on the optimized mesh to eliminate local irregularities. The smoothing process uses a feature-preserving Laplacian smoothing algorithm to ensure that the mesh quality is improved without damaging important features. In this way, the final simplified model not only retains the key features of the production line, but also has good mesh quality.

[0087] The embodiment establishes a complete multi-scale feature fusion framework through the above implementation scheme. This scheme is particularly suitable for digital model simplification of complex automated production lines, and can accurately identify and retain key features in the production line while significantly reducing model complexity. Practical application shows that this scheme can effectively support digital monitoring and optimization analysis of production lines, providing a reliable digital model foundation for intelligent manufacturing.

[0088] From the above description, it can be seen that the automatic production line digital model simplification method provided by the embodiments of the present application can identify and save key feature information through a saliency feature extraction module, and establish a four-level topological structure tree. An adaptive feature weight calculation mechanism is innovatively designed, and combined with multi-resolution wavelet analysis to realize intelligent simplification of the mesh. The system extracts and matches topological structure features and local geometric features at different scales through a multi-scale feature fusion module, and realizes accurate preservation of key features during model simplification. This method breaks through the limitations of traditional model simplification and provides an intelligent and automated solution for efficient processing of large-scale digital models.

[0089] In an embodiment of the automatic production line digital model simplification method of the present application, referring to Figure 2 It can also specifically include the following content:

[0090] Step S201: Discretize the automatic production line digital model to obtain point cloud data, establish a three-dimensional mesh structure based on the point cloud data, calculate the curvature value, normal vector and neighborhood density of each vertex in the three-dimensional mesh structure, input the curvature value, normal vector and neighborhood density into a saliency calculation formula to obtain a saliency coefficient, and construct a saliency feature extraction module based on the saliency coefficient;

[0091] Step S202: comparing the saliency coefficients in the saliency feature extraction module with a first threshold value, extracting a vertex with a saliency coefficient greater than the first threshold value as a feature point, constructing a salient feature point set based on the spatial positional relationship of the feature points, and storing the geometric attribute parameters and the topological attribute parameters of the salient feature point set as morphological parameters and positional parameters, respectively.

[0092] Optionally, in the discretization process of the digital model of the automated production line, a multi-view scanning strategy is adopted. For different types of equipment in the production line, such as mechanical arms, conveyors, detection equipment, etc., optimal scanning angles and sampling densities are set respectively. In the connection between devices and the functional interaction area, a higher sampling density is used; in the flat area and simple structure, a lower sampling density is used to ensure the integrity and pertinence of the collected data.

[0093] In the point cloud data acquisition process, an adaptive noise reduction algorithm is implemented. First, the noise points are identified by statistical outlier analysis, and the distance distribution of each point with its neighborhood points is calculated. In the precision component area of the production line, such as sensor installation position, precision clamp, etc., a smaller noise threshold is used; in the large structural part area, such as support frame, protective cover, etc., a larger noise threshold is used.

[0094] In the construction of the three-dimensional grid structure, a grid division strategy based on local features is developed. First, an improved Poisson reconstruction algorithm is used to generate an initial triangular grid, and then the grid is optimized according to the local geometric features. In areas with sharp curvature changes, such as device corners, pipe bends, etc., the grid density is increased; in flat areas, the grid density is appropriately reduced, realizing the adaptive distribution of grid density.

[0095] In the vertex feature calculation link, a multi-scale feature extraction framework is implemented. For each grid vertex, first determine its feature calculation neighborhood radius. In areas with rich details, use a smaller neighborhood radius; in areas with simple features, use a larger neighborhood radius. Within this neighborhood, the principal curvature and Gaussian curvature are calculated by fitting the local quadratic surface, and the vector direction is estimated.

[0096] In the neighborhood density calculation process, an adaptive kernel function method is used. According to the local point cloud distribution characteristics, the bandwidth of the kernel function is dynamically adjusted. In sparse areas of point cloud, a larger bandwidth is used; in dense areas, a smaller bandwidth is used. This adaptive strategy ensures the accuracy of density estimation, especially in the special-shaped structures and precision components of the production line.

[0097] In the implementation of the embodiment, a comprehensive feature fusion model is developed in the calculation of saliency. The model combines curvature features, normal vector changes, and neighborhood density through nonlinear mapping into saliency coefficients. In the calculation process, dynamic weights are assigned to different features, such as increasing the curvature weight in the edge feature area, increasing the normal vector weight in the surface feature area, and increasing the density weight in the connecting structure area.

[0098] In the construction process of the saliency feature extraction module, the embodiment realizes a feature scale adaptive mechanism. By analyzing the spatial distribution of local features, the scale parameters of feature extraction are dynamically adjusted. For precision parts in the production line, a smaller feature extraction scale is used; for large structural parts, a larger feature extraction scale is used.

[0099] In the feature point selection process, the embodiment adopts a multi-level threshold screening strategy. First, a global threshold is used for preliminary screening, and then a local adaptive threshold is used for fine screening in different functional areas. This hierarchical screening method ensures the representativeness and integrity of the feature points, especially in the key functional areas of the production line.

[0100] In the construction of the salient feature point set, the embodiment develops a spatial clustering-based organization method. By analyzing the spatial distance and feature similarity between feature points, the feature points with correlation are organized into a feature point set. In the clustering process, the functional module division of the production line is considered to ensure that the feature point set has a good correspondence with the actual functional units.

[0101] In the parameter storage link, the embodiment realizes a hierarchical data organization structure. Morphological parameters include local geometric features, curvature distribution, surface properties, etc.; position parameters include spatial coordinates, relative position relationships, topological connectivity, etc. By establishing a parameter index table, efficient access and management of feature data are realized.

[0102] Through the above implementation scheme, the embodiment establishes a complete feature extraction and storage system. This scheme is particularly suitable for feature recognition in complex automated production lines and can accurately extract and save key feature information in the production line. Practical application shows that this scheme can effectively support the digital modeling and optimization analysis of the production line, providing a reliable data foundation for equipment monitoring and maintenance.

[0103] In an embodiment of the automated production line digital model simplification method of the present application, referring to Figure 3 The method can further include the following content:

[0104] Step S301: input the morphological parameters and the position parameters of the set of salient feature points into a feature encoder, the feature encoder performs geometric feature encoding on the morphological parameters to obtain a geometric feature vector, performs spatial position encoding on the position parameters to obtain a position feature vector, combines the geometric feature vector and the position feature vector to form feature data, and writes the feature data into a salient feature database according to a hierarchical structure;

[0105] Step S302: construct a four-level topological structure tree based on the set of salient feature points, take the point data in the set of salient feature points as a first-level node, take the connection relationship between adjacent points as a second-level node, take a face piece enclosed by a closed line segment as a third-level node, take a body element formed by a closed face piece as a fourth-level node, and store a corresponding feature data index in each node of the four-level topological structure tree.

[0106] Optionally, in the feature encoding process, a dual-stream feature encoding architecture is developed. For the morphological parameters of the production line, such as the surface curvature of the equipment, the edge features, the material properties, etc., a deep convolutional network is used for geometric feature encoding; for the position parameters, such as the spatial coordinates, the relative position relationship, the assembly constraints, etc., a graph attention network is used for spatial position encoding. This dual-stream encoding architecture can fully extract the geometric and spatial feature information in the production line.

[0107] In the geometric feature encoding link, a multi-scale feature extraction mechanism is realized. Through a pyramid pooling structure, morphological features are extracted at different spatial scales. For precise components in the production line, such as sensor mounting seats, positioning pins, etc., detailed features are extracted at a small scale; for large structural parts, such as racks, protective covers, etc., overall morphological features are extracted at a large scale. Each layer of the encoding network is provided with a residual connection to ensure effective transmission of feature information.

[0108] In the position feature encoding process, a hierarchical spatial relationship modeling method is used. First, a local reference coordinate system is established to map each functional unit in the production line to a unified feature space. Then, through a graph attention mechanism, the spatial dependency relationship between different units is captured, and the attention weight is dynamically adjusted according to the distance and functional correlation between units. This encoding method is particularly suitable for expressing the complex spatial layout and connection relationship in the production line.

[0109] In the feature vector combination stage, an adaptive feature fusion strategy is developed. Through a gating mechanism, the fusion weight of the geometric feature vector and the position feature vector is dynamically adjusted, the weight of the position feature is increased at key positions such as equipment connections, and the weight of the geometric feature is increased in areas with complex shapes. The fused feature data not only retains local geometric details, but also contains global spatial information.

[0110] The embodiment realizes a multi-level indexing mechanism in the database writing process. A hierarchical storage structure of feature data is established, including a geometric feature layer, a spatial relationship layer, and a semantic attribute layer. Each layer of data is equipped with an efficient retrieval index, supporting multi-dimensional queries according to feature types, spatial positions, and functional attributes. This structure design significantly improves the access efficiency of feature data.

[0111] In the process of constructing the four-level topology structure tree, a bottom-up construction strategy is adopted. First, the set of significant feature points is mapped to the first level nodes, each node containing the coordinate information and feature attributes of the points. Then, the spatial adjacency relationship between the points is analyzed, and the minimum spanning tree algorithm is used to construct the connection relationship nodes of the second level, which reflect the physical connection and logical association in the production line.

[0112] In the patch construction link, an adaptive triangulation algorithm is developed. Through analyzing the closure and coplanar characteristics of the line segments, the third level patch nodes are identified and constructed. In the construction process, the specific geometric constraints of the production line, such as parallelism and perpendicularity, are considered to ensure that the generated patches conform to the actual structural characteristics.

[0113] In the process of constructing the volume element, a recognition method based on the closure of the patch is realized. By analyzing the connection relationship and normal vector distribution of the patch, the closed volume element structure is identified as the fourth level node. For complex components in the production line, such as valve assemblies and transmission devices, a hierarchical combination method is used to construct their complete volume element representation.

[0114] In the process of constructing the node feature index, a distributed storage strategy is adopted. In each level node, in addition to storing the feature data index of the current level, an associated index with other level nodes is also established. This multi-level index structure supports fast feature retrieval and association analysis.

[0115] Through the above implementation scheme, a complete feature coding and topology structure expression system is established. This scheme is particularly suitable for the digital expression of complex automated production lines and can effectively capture and organize multi-dimensional feature information in the production line. Practical application shows that this scheme reliably supports feature analysis and model simplification of the production line, providing a solid data foundation for digital twinning and intelligent operation and maintenance of the production line.

[0116] In an embodiment of the automated production line digital model simplification method of the present application, referring to Figure 4 It can also specifically include the following content:

[0117] Step S401: constructing a spatial density distribution map based on the four-level topology tree, calculating the number of significant feature points in the neighborhood range of each node in the spatial density distribution map, dividing the number by the neighborhood volume to obtain a local density value, normalizing the local density value to obtain a density distribution coefficient, and inputting the density distribution coefficient into a weight mapping function to generate a weight coefficient corresponding to each level;

[0118] Step S402: performing matrix multiplication operation on the weight coefficient and the feature data of each level node in the four-level topology tree to generate a weighted feature matrix, performing standardization processing on the weighted feature matrix to obtain weighted feature data, and reorganizing the weighted feature data into a tree structure according to the hierarchical relationship.

[0119] Optionally, in the embodiment, an adaptive grid division strategy is adopted when constructing the spatial density distribution map. The grid size is dynamically adjusted according to the feature distribution characteristics of different functional areas in the production line. Smaller grid size is adopted in feature-dense areas such as device connection and control panel, and larger grid size is adopted in feature-sparse areas such as conveyor belt and pipeline, so as to ensure that the density distribution map can accurately reflect the spatial distribution law of the production line features.

[0120] In the embodiment, a multi-scale neighborhood analysis method is implemented when calculating the neighborhood range of the node. First, the basic neighborhood radius of each node is determined, which is inversely proportional to the feature density of the area. Then, the number of significant feature points in the neighborhood is counted by using a spherical search algorithm. For key functional components in the production line, such as sensor assemblies and actuators, smaller neighborhood radius is adopted to ensure accurate expression of local features.

[0121] In the embodiment, an adaptive kernel density estimation algorithm is developed in the local density calculation process. By analyzing the spatial distribution pattern of the feature points, the bandwidth parameter of the kernel function is dynamically adjusted. In areas with uneven feature distribution, such as mechanical arm joints, a variable bandwidth strategy is adopted to ensure the accuracy of density estimation. The calculated local density value not only reflects the degree of feature aggregation, but also preserves the continuity of spatial distribution.

[0122] In the embodiment, a hierarchical normalization strategy is adopted when normalizing the density distribution coefficient. According to the feature distribution characteristics of different levels, corresponding normalization functions are designed. For the point level, maximum and minimum value normalization is adopted; for the line level, the length of the line segment is considered; for the face level, area weight is introduced; for the body level, normalization is combined with volume factor. This hierarchical normalization method ensures the comparability of features at different levels.

[0123] In the weight mapping function design, an adaptive mapping mechanism based on feature importance is realized. For different functional areas of the production line, such as stations, logistics channels, control areas, etc., a feature importance evaluation model is established. The model considers factors such as regional functional importance, operation frequency, and maintenance requirements, and maps the density distribution coefficient to a reasonable weight value.

[0124] In the weight coefficient generation process, a hierarchical weight balancing mechanism is developed. By analyzing the interdependence of features at each level, a weight adjustment rule is established. For example, when a certain element node has important functional attributes, the weights of the patches, line segments, and points it contains are increased accordingly, ensuring the hierarchical consistency of the features.

[0125] In the matrix multiplication operation section, a sparse matrix optimization algorithm is realized. Considering the sparsity characteristics of production line feature data, a compressed storage format and an efficient sparse matrix operation method are used. Through block calculation strategy, the processing efficiency of large-scale feature data is significantly improved.

[0126] In the weighted feature matrix standardization process, a feature correlation preservation strategy is adopted. By analyzing the covariance structure between features, a standardization method that preserves feature correlation is designed. This method not only unifies the scale of features, but also preserves important correlation information between features.

[0127] In the tree structure reorganization process, a feature clustering optimization algorithm is developed. Based on the similarity of weighted features, hierarchical clustering of feature nodes is performed. In the clustering process, the functional module division of the production line is considered to ensure that the reorganized tree structure reflects the actual functional hierarchy relationship.

[0128] Through the dynamic feature indexing mechanism, the efficient organization of the tree structure is realized. Multi-dimensional indexes are established for each node, including spatial location index, feature type index, and functional attribute index. This indexing mechanism supports fast feature retrieval and update operations.

[0129] Through the above implementation scheme, a complete feature weighting and organization system is established. This scheme is particularly suitable for feature importance analysis of complex automated production lines, and can reasonably allocate feature weights according to actual application requirements, providing a reliable weight basis for subsequent model simplification. Practice shows that this scheme can effectively support feature optimization and simplification of the production line, significantly improving the practicality of digital models in actual applications.

[0130] In an embodiment of the automated production line digital model simplification method of the present application, referring to Figure 5 It can also specifically include the following content:

[0131] Step S501: determine a space bounding box based on the weighted feature data, recursively divide the space bounding box into an octree structure, calculate a local variance of the weighted feature data at each node of the octree structure, continue to subdivide the node when the local variance of the node is greater than a second threshold value until a preset number of subdivision levels is reached or the local variance is less than the second threshold value, construct an adaptive grid on the octree node where the subdivision is completed, and map a Haar wavelet basis function to a node coordinate of the adaptive grid;

[0132] Step S502: perform coefficient decomposition on the Haar wavelet basis function on the adaptive grid node, calculate an energy value of each layer of wavelet coefficients, set an adaptive threshold value according to the energy value, set wavelet coefficients less than the adaptive threshold value to zero, perform a reconstruction operation on the remaining wavelet coefficients to generate an initial simplified grid, and establish a corresponding relationship between nodes of the initial simplified grid and the original weighted feature data.

[0133] Optionally, in the embodiment, a feature distribution adaptive strategy is adopted when determining the space bounding box. According to the spatial distribution characteristics of the weighted feature data, the size and direction of the bounding box are dynamically adjusted. For the region of movable components such as mechanical arms in the production line, the bounding box range is appropriately expanded to accommodate the motion space; for the fixed equipment region, the bounding box is set close to the equipment contour to improve the space utilization efficiency.

[0134] In the recursive division process of the octree structure, the embodiment realizes an adaptive subdivision mechanism based on feature distribution. By analyzing the distribution characteristics of the weighted feature data in each node, a variance calculation model is established. The model not only considers the dispersion degree of the feature values, but also combines the weight information of the features, ensuring more detailed division in key functional areas.

[0135] In the local variance calculation link, the embodiment develops a multi-scale feature aggregation algorithm. For feature distributions of different scales, a weighted statistical method is used to calculate the local variance. In the precision control region of the production line, such as the sensor installation position and precision positioning structure, a smaller statistical window is used; in the region of large structural parts, a larger statistical window is used.

[0136] In the octree subdivision control, the embodiment realizes a double-threshold adaptive mechanism. In addition to using the local variance threshold value, a depth control threshold value is also introduced. In regions where the features change dramatically, such as device interfaces and functional conversion areas, deeper subdivision is allowed; in regions where the features change smoothly, the subdivision depth is appropriately controlled to avoid the computational burden caused by excessive subdivision.

[0137] In the adaptive mesh construction process, the embodiment adopts a feature-preserving mesh generation strategy. Based on the distribution characteristics of octree nodes, a tetrahedral mesh with variable distribution density is constructed. In key functional areas of the production line, such as mechanical interfaces and control panels, the mesh density is increased; in simple structure areas, the mesh density is appropriately reduced.

[0138] In the process of mapping the Haar wavelet basis function, the embodiment develops a mapping mechanism based on feature importance. According to the feature weight of the area where the mesh node is located, the appropriate scale of the wavelet basis function is selected. For high-weight areas, more basis functions are used to improve the expression accuracy; for low-weight areas, fewer basis functions are used to reduce the data volume.

[0139] In the process of wavelet coefficient decomposition, the embodiment realizes a multi-level energy analysis method. Through wavelet transform, the feature information is decomposed into different frequency levels, and the energy distribution of each level is calculated. In the energy calculation, the weight information of the feature is considered to ensure that important features are fully reflected in the energy distribution.

[0140] In the adaptive threshold setting link, the embodiment adopts a dynamic threshold strategy based on energy distribution. By analyzing the energy distribution law of the wavelet coefficients, an adaptive threshold determination model is established. This model effectively removes the secondary features with low energy while preserving important features, achieving efficient compression of data.

[0141] In the process of wavelet coefficient reconstruction, the embodiment develops a feature-preserving reconstruction algorithm. By layer-by-layer superposition of the retained wavelet coefficients, the simplified mesh is reconstructed. In the reconstruction process, a progressive strategy is adopted, first reconstructing the main features, then gradually adding the secondary features, until the preset accuracy requirement is met.

[0142] In the process of establishing the node correspondence relationship, the embodiment realizes a mapping mechanism based on feature similarity. By calculating the geometric distance and feature similarity between the simplified mesh nodes and the original feature data, an optimal matching relationship is established. This mapping mechanism ensures the feature consistency between the simplified model and the original model.

[0143] Through the above implementation scheme, the embodiment establishes a complete mesh simplification and feature preservation system. This scheme is particularly suitable for digital model simplification of complex automated production lines, and can significantly reduce the data volume while preserving key features. Practice shows that this scheme can effectively support digital monitoring and optimization analysis of the production line, providing an efficient digital model foundation for intelligent manufacturing.

[0144] In an embodiment of the automated production line digital model simplification method of the present application, referring to Figure 6 , the following contents can also be specifically included:

[0145] Step S601: Multi-level decomposition is performed on the initial simplified mesh to obtain mesh representations of different scales, a mesh connectivity graph is constructed at a coarse scale level, a degree distribution and connectivity property of each node in the mesh connectivity graph are calculated, a mesh skeleton is constructed based on the degree distribution and connectivity property, and a branch structure and connection relationship of the mesh skeleton are extracted as topological structure features;

[0146] Step S602: A local curvature tensor of a node of the initial simplified mesh is calculated at a fine scale level, a principal curvature value and a principal curvature direction are calculated based on the curvature tensor, the principal curvature value and the principal curvature direction are combined to form a curvature descriptor, and local geometric features are obtained by clustering analysis on the curvature descriptor, and a mapping relationship between the local geometric features and the mesh node is established.

[0147] Optionally, in the multi-level decomposition process, an adaptive mesh simplification strategy is adopted. Different simplification parameters are set for different functional modules in the production line, such as conveying devices, mechanical arms, control units, etc. Higher mesh precision is maintained in key functional areas, and precision is appropriately reduced in secondary areas, so as to ensure that the simplified mesh can accurately express the geometric features of the production line.

[0148] In the mesh connectivity graph construction link, a connection weight mechanism based on functional association is implemented. By analyzing the spatial relationship and functional dependence between mesh nodes, different weights are given to the connection edges. For example, a higher weight is given to the connection at the device interface, and a lower weight is given to the connection at the ordinary structure, so as to highlight the key connection structure in the production line.

[0149] In the degree distribution calculation process, a multi-scale statistical analysis method is developed. First, the degree features of the nodes in the local neighborhood are calculated, and then the degree distribution at different scales is obtained by hierarchical aggregation. This method can effectively identify key connection points in the production line, such as multi-device intersection, functional module interface, etc.

[0150] In the mesh skeleton construction, a feature-preserving skeleton extraction algorithm is adopted. By analyzing the degree distribution and connectivity property, the main structure line of the mesh is identified. In the extraction process, special attention is paid to the functional flow lines of the production line, such as material transmission path, signal transmission channel, etc., so as to ensure that the skeleton structure can reflect the functional layout of the production line.

[0151] In the branch structure extraction link, a hierarchical extraction mechanism based on functional modules is implemented. First, the main functional line is identified, such as the main conveying belt path, the main control channel, etc., and then the secondary branch is extracted, such as the auxiliary conveying device, the connection of the monitoring device, etc. This hierarchical extraction method effectively expresses the functional hierarchical relationship of the production line.

[0152] In the curvature tensor calculation process of the embodiment, an adaptive neighborhood selection strategy is developed. The size of the calculation neighborhood is dynamically adjusted according to the complexity of the local surface. A smaller calculation neighborhood is used at precise structures in the production line, such as sensor mounting seats and positioning reference surfaces, and a larger calculation neighborhood is used at large planar structures.

[0153] In the principal curvature analysis link of the embodiment, a robust feature direction estimation method is used. The principal curvature values and directions are obtained by eigenvalue decomposition of the curvature tensor. In the calculation process, the influence of noise and local disturbance is considered to ensure that the extracted curvature features have good stability.

[0154] In the curvature descriptor construction process of the embodiment, a multi-scale feature fusion mechanism is realized. The principal curvature information at different scales is weighted and combined to form a comprehensive curvature descriptor. The weight coefficient is dynamically adjusted according to the importance of the feature to ensure that the descriptor can accurately express the local geometric feature.

[0155] In the clustering analysis stage of the embodiment, an adaptive clustering algorithm based on geometric similarity is developed. The number of clusters and cluster centers are dynamically determined by analyzing the distribution characteristics of the curvature descriptor. In the clustering process, the unique geometric feature types of the production line, such as cylindrical surfaces, planes, and irregular surfaces, are considered.

[0156] In the feature mapping relationship establishment process of the embodiment, a bidirectional association mechanism is used. Not only is the forward mapping from geometric features to grid nodes established, but also the reverse index from grid nodes to geometric features is constructed. This bidirectional mapping mechanism supports fast feature query and local update operations.

[0157] Through the above implementation scheme, the embodiment establishes a complete multi-scale feature analysis system. This scheme is particularly suitable for feature extraction and expression of complex automated production lines and can accurately capture the geometric and topological features of the production line at different scales. Practical application shows that this scheme can effectively support feature analysis and optimization design of the production line and provides reliable technical support for digital transformation.

[0158] In an embodiment of the automated production line digital model simplification method of the present application, referring to Figure 7 The method can further include the following steps.

[0159] Step S701: input the topological structure features and the local geometric features into a feature fusion network, the feature fusion network generates a feature correspondence graph based on graph convolution operation, calculates a similarity matrix between nodes in the feature correspondence graph, performs matching operation on the similarity matrix and the shape parameters and position parameters stored in a significant feature database, generates a feature matching metric value, and constructs a node optimization objective function based on the feature matching metric value.

[0160] Step S702: iteratively solving the node optimization objective function, adjusting the node position and connection relationship of the initial simplified mesh, stopping iteration when the objective function value is less than a third threshold or the maximum number of iterations is reached, reconstructing the final mesh node coordinates and connection relationship into a three-dimensional model, and performing mesh smoothing processing on the reconstructed three-dimensional model to obtain a final simplified model.

[0161] Optionally, in the feature fusion network design of the embodiment, a multi-layer graph convolution structure is adopted. The network first receives topological structure features and local geometric features as input, and gradually fuses these feature information through multi-layer graph convolution operation. In key functional areas of the production line, such as device interfaces and control units, the network uses deeper convolution layers to capture more complex feature associations.

[0162] In the graph convolution operation process of the embodiment, an adaptive receptive field mechanism is realized. For different types of functional modules in the production line, the receptive field range of the graph convolution is dynamically adjusted. For example, for complex motion components such as robotic arms, a larger receptive field is used to capture overall motion features; for precision positioning structures, a smaller receptive field is used to preserve local details.

[0163] In the similarity matrix calculation link of the embodiment, a multi-dimensional feature similarity measurement method is developed. By combining geometric similarity, topological similarity and functional similarity, a comprehensive similarity evaluation model is constructed. When calculating the similarity of device connections, the model emphasizes functional consistency; when evaluating the similarity of structural parts, it focuses more on the matching degree of geometric shape.

[0164] In the feature matching operation process of the embodiment, a hierarchical matching strategy is adopted. First, fast matching is performed at the coarse-grained level to select candidate feature pairs; then, accurate matching is performed at the fine-grained level to determine the optimal feature correspondence. This hierarchical matching method significantly improves the processing efficiency of large-scale feature data.

[0165] In the construction of the objective function of the embodiment, a multi-objective balancing mechanism is realized. The objective function not only considers the feature matching metric value, but also introduces a geometric preservation term and a topological preservation term. In the optimization process, the weights of each term are dynamically adjusted according to the functional importance of different regions to ensure that the simplified model retains key features.

[0166] In the iterative solving process of the embodiment, an adaptive step size control strategy is developed. By analyzing the gradient information of the objective function, the optimization step size is dynamically adjusted. A smaller step size is used in areas with drastic feature changes, and a larger step size is used in areas with smooth features, improving optimization efficiency while ensuring convergence stability.

[0167] The embodiment adopts a constraint-based optimization method in the node position adjustment link. By establishing geometric constraints and topological constraints, the functional characteristics of the production line are ensured during node adjustment. For example, the device mounting surface maintains the planeness constraint, the bearing seat maintains the cylindrical constraint, and the precision positioning structure maintains the relative position constraint.

[0168] In the connection relationship optimization process, the embodiment realizes a topology-preserving reconstruction mechanism. By analyzing the connection importance between nodes, the key connections are preferentially maintained, and the secondary connections are appropriately simplified. This differentiated processing strategy ensures that the simplified model maintains functional integrity while significantly reducing data volume.

[0169] In the three-dimensional model reconstruction stage, the embodiment develops a feature-aware reconstruction algorithm. Based on the optimized grid nodes and connection relationships, a progressive reconstruction strategy is adopted. First, the main functional structure is reconstructed, and then the detailed features are gradually added to ensure the quality and efficiency of the reconstructed model.

[0170] In the grid smoothing processing link, the embodiment adopts a local feature-preserving smoothing algorithm. By analyzing the local curvature and feature line information, the sharpness of key features is maintained during smoothing. For important features such as precision mating surfaces and positioning references in the production line, a smaller smoothing strength is used; for general transition surfaces, a larger smoothing strength can be used.

[0171] Through the above implementation scheme, a complete model simplification and optimization system is established. This scheme is particularly suitable for digital model simplification of complex automated production lines and can significantly reduce data volume while maintaining key functional features. Practical application shows that this scheme reliably supports virtual simulation and digital twin applications of production lines, providing an efficient digital foundation for intelligent manufacturing.

[0172] In order to break through the limitations of traditional model simplification and provide an intelligent and automated solution for efficient processing of large-scale digital models, an embodiment of an automated production line digital model simplification device for implementing all or part of the contents of the automated production line digital model simplification method is provided, as shown in Figure 8 , which specifically includes the following contents:

[0173] The feature extraction module 10 is used to construct a salient feature extraction module that calculates saliency scores for point, line, surface, and volume data in the automated production line digital model. The feature points with saliency scores exceeding a first threshold value are extracted as a set of salient feature points, and the corresponding morphological parameters and position parameters are stored in a salient feature database. A four-level topological structure tree is established based on the set of salient feature points.

[0174] The architecture building module 20 is used for inputting the four-level topology structure tree into an adaptive feature weight calculation module. The adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the set of salient feature points. The weight coefficient is multiplied by feature data of the corresponding level to obtain weighted feature data. The weighted feature data is input into an adaptive octree grid construction module. The adaptive octree grid construction module constructs a grid node based on the weighted feature data. A multi-resolution wavelet basis function is embedded into the grid node. The wavelet basis function coefficients are processed by hierarchical thresholding. The thresholded wavelet coefficients are reconstructed into an initial simplified grid.

[0175] The model simplification module 30 is used for inputting the initial simplified grid into a multi-scale feature fusion module. The multi-scale feature fusion module extracts a topology structure feature of the initial simplified grid at a coarse scale and extracts a local geometric feature of the initial simplified grid at a fine scale. A feature correspondence graph is constructed based on the topology structure feature and the local geometric feature. The feature correspondence graph is matched with a shape parameter and a position parameter in the salient feature database. The initial simplified grid is adjusted based on a matching result. A final simplified model is generated.

[0176] As can be seen from the above description, the automatic production line digital model simplification device provided by the embodiments of the present application can identify and save key feature information through the saliency feature extraction module and establish a four-level topology structure tree. An adaptive feature weight calculation mechanism is innovatively designed. The intelligent simplification of the grid is realized by combining multi-resolution wavelet analysis. The system extracts and matches the topology structure feature and the local geometric feature at different scales through the multi-scale feature fusion module. The accurate preservation of key features in the model simplification process is realized. The method breaks through the limitations of traditional model simplification and provides an intelligent and automatic solution for efficient processing of large-scale digital models.

[0177] From the hardware level, in order to break through the limitations of traditional model simplification and provide an intelligent and automatic solution for efficient processing of large-scale digital models, the present application provides an embodiment of an electronic device for realizing all or part of the contents of the automatic production line digital model simplification method. The electronic device specifically includes the following contents:

[0178] The processor, the memory, the communications interface, and the bus; wherein the processor, the memory, the communications interface complete the communication among each other through the bus; the communications interface is used for realizing the information transmission between the automatic production line digital model simplification device and the core business system, the user terminal, and the related database and other related equipment; the logic controller can be a desktop computer, a tablet computer, a mobile terminal and the like, and the embodiment is not limited thereto. In the embodiment, the logic controller can be implemented with reference to the embodiments of the automatic production line digital model simplification method and the embodiments of the automatic production line digital model simplification device, the contents of which are incorporated herein, and the repeated parts will not be described herein.

[0179] It can be understood that the user terminal can include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, and the like. The smart wearable device can include smart glasses, a smart watch, a smart bracelet, and the like.

[0180] In actual application, part of the automatic production line digital model simplification method can be executed on the electronic device as described above, or all operations can be completed in the client device. Specifically, the selection can be made according to the processing capability of the client device and the limitation of the user usage scenario, and the like. The present application is not limited thereto. If all operations are completed in the client device, the client device can further include a processor.

[0181] The client device described above can have a communication module (i.e., a communication unit) and can be communicatively connected with a remote server to realize the data transmission with the server. The server can include a server of the task scheduling center side, and the server of the intermediate platform can also be included in other implementation scenarios, such as the server of the third-party server platform communicatively connected with the server of the task scheduling center. The server can include a single computer device, a server cluster composed of multiple servers, or a distributed server structure.

[0182] Figure 9 A schematic block diagram of the system configuration of the electronic device 9600 of the embodiment of the present application is shown in FIG. 9. As shown in FIG. 9, the electronic device 9600 can include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. It is worth noting that the structure shown in FIG. 9 is exemplary; other types of structures can also be used to supplement or replace the structure to realize the telecommunication function or other functions. Figure 9 Figure 9 The structure shown in FIG. 9 is exemplary; other types of structures can also be used to supplement or replace the structure to realize the telecommunication function or other functions.

[0183] ​In an embodiment, the automated production line digital model simplification method function can be integrated into the central processor 9100. The central processor 9100 can be configured to control as follows:

[0184] Step S101: A saliency feature extraction module is constructed, which calculates saliency scores for point, line, surface, and body data in the automated production line digital model, extracts feature points with saliency scores exceeding a first threshold value as a saliency feature point set, stores morphological parameters and position parameters corresponding to the saliency feature point set in a saliency feature database, and establishes a four-level topological structure tree based on the saliency feature point set;

[0185] Step S102: The four-level topological structure tree is input into an adaptive feature weight calculation module, which calculates weight coefficients for each level based on the spatial distribution density of the saliency feature point set, multiplies the weight coefficients by feature data of the corresponding level to obtain weighted feature data, and inputs the weighted feature data into an adaptive octree grid construction module, which constructs grid nodes based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid nodes, performs hierarchical thresholding processing on the wavelet basis function coefficients, and reconstructs the thresholded wavelet coefficients into an initial simplified grid;

[0186] Step S103: The initial simplified grid is input into a multi-scale feature fusion module, which extracts topological structure features of the initial simplified grid at a coarse scale and local geometric features of the initial simplified grid at a fine scale, constructs a feature correspondence graph based on the topological structure features and the local geometric features, matches the feature correspondence graph with morphological parameters and position parameters in the saliency feature database, adjusts nodes of the initial simplified grid according to the matching result, and generates a final simplified model.

[0187] As can be seen from the above description, the electronic device provided by the embodiments of the present application identifies and saves key feature information through a saliency feature extraction module and establishes a four-level topological structure tree. An adaptive feature weight calculation mechanism is innovatively designed, and intelligent simplification of the grid is realized in combination with multi-resolution wavelet analysis. The system extracts and matches topological structure features and local geometric features at different scales through a multi-scale feature fusion module, and realizes accurate preservation of key features in the model simplification process. This method breaks through the limitations of traditional model simplification and provides an intelligent and automated solution for efficient processing of large-scale digital models.

[0188] In another embodiment, the automated production line digital model simplification device can be configured separately from the central processing unit 9100. For example, the automated production line digital model simplification device can be configured as a chip connected to the central processing unit 9100, and the automated production line digital model simplification method function can be implemented through the control of the central processing unit.

[0189] 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.

[0190] 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.

[0191] 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.

[0192] 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.

[0193] 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.

[0194] The memory 9140 can further include a data storage 9143 for storing data such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. A driver storage 9144 of the memory 9140 can include various drivers of the electronic device for communication functions and / or for performing other functions of the electronic device (e.g., a messaging application, a phonebook application, etc.).

[0195] The communication module 9110 is a transmitter / receiver that transmits and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processor 9100 to provide input signals and receive output signals, as in the case of a conventional mobile communication terminal.

[0196] Based on different communication technologies, a plurality of communication modules 9110 can be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. The communication module 9110 (transmitter / receiver) is further coupled to the speaker 9131 and the microphone 9132 via the audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby implementing a conventional telecommunication function. The audio processor 9130 can include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is coupled to the central processor 9100, thereby enabling recording on the local device via the microphone 9132 and enabling playing of a sound stored on the local device via the speaker 9131.

[0197] The embodiment of the present application further provides a computer readable storage medium capable of implementing all steps of the automatic production line digital model simplification method with the execution subject being a server or a client in the above-mentioned embodiment, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement all steps of the automatic production line digital model simplification method with the execution subject being a server or a client in the above-mentioned embodiment, for example, the processor executes the computer program to implement the following steps:

[0198] Step S101: constructing a saliency feature extraction module, wherein the saliency feature extraction module calculates saliency scores of point, line, surface and body data in the automatic production line digital model, extracts feature points with saliency scores exceeding a first threshold value as a saliency feature point set, stores morphological parameters and position parameters corresponding to the saliency feature point set in a saliency feature database, and establishes a four-level topological structure tree based on the saliency feature point set;

[0199] Step S102: input the four-level topology tree into an adaptive feature weight calculation module, the adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the set of salient feature points, multiplies the weight coefficient by feature data of the corresponding level to obtain weighted feature data, and inputs the weighted feature data into an adaptive octree grid construction module, the adaptive octree grid construction module constructs a grid node based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid node, performs hierarchical thresholding processing on the wavelet basis function coefficient, and reconstructs the thresholded wavelet coefficient into an initial simplified grid;

[0200] Step S103: input the initial simplified grid into a multi-scale feature fusion module, the multi-scale feature fusion module extracts a topological structure feature of the initial simplified grid at a coarse scale and extracts a local geometric feature of the initial simplified grid at a fine scale, constructs a feature correspondence graph based on the topological structure feature and the local geometric feature, matches the feature correspondence graph with a shape parameter and a position parameter in the salient feature database, adjusts nodes of the initial simplified grid according to a matching result, and generates a final simplified model.

[0201] As can be seen from the above description, the computer readable storage medium provided by the embodiments of the application identifies and saves key feature information through the salient feature extraction module, and establishes a four-level topology tree. An adaptive feature weight calculation mechanism is innovatively designed, and intelligent simplification of the grid is realized in combination with multi-resolution wavelet analysis. The system extracts and matches topological structure features and local geometric features at different scales through the multi-scale feature fusion module, and realizes accurate preservation of key features in the model simplification process. The method breaks through the limitations of traditional model simplification and provides an intelligent and automated solution for efficient processing of large-scale digital models.

[0202] The embodiments of the application also provide a computer program product capable of realizing all steps in the automatic production line digital model simplification method in which the execution subject in the above embodiments is a server or a client. The computer program / instruction is executed by a processor to realize the steps of the automatic production line digital model simplification method, for example, the computer program / instruction realizes the following steps:

[0203] Step S101: construct a salient feature extraction module, the salient feature extraction module calculates a saliency score for point, line, surface, and volume data in an automatic production line digital model, extracts feature points with a saliency score exceeding a first threshold value as a set of salient feature points, stores shape parameters and position parameters corresponding to the set of salient feature points in a salient feature database, and establishes a four-level topology tree based on the set of salient feature points;

[0204] Step S102: input the four-level topology tree into an adaptive feature weight calculation module, the adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the salient feature point set, multiplies the weight coefficient by feature data of the corresponding level to obtain weighted feature data, and inputs the weighted feature data into an adaptive octree grid construction module, the adaptive octree grid construction module constructs a grid node based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid node, performs hierarchical thresholding processing on a wavelet basis function coefficient, and reconstructs the thresholded wavelet coefficient into an initial simplified grid;

[0205] Step S103: input the initial simplified grid into a multi-scale feature fusion module, the multi-scale feature fusion module extracts a topology structure feature of the initial simplified grid at a coarse scale and extracts a local geometric feature of the initial simplified grid at a fine scale, constructs a feature correspondence graph based on the topology structure feature and the local geometric feature, matches the feature correspondence graph with a shape parameter and a position parameter in the salient feature database, adjusts nodes of the initial simplified grid according to a matching result, and generates a final simplified model.

[0206] From the above description, it can be known that the computer program product provided by the embodiment of the present application identifies and saves key feature information through the salient feature extraction module, and establishes a four-level topology structure tree. An adaptive feature weight calculation mechanism is innovatively designed, and intelligent simplification of the grid is realized in combination with multi-resolution wavelet analysis. The system extracts and matches the topology structure feature and the local geometric feature at different scales through the multi-scale feature fusion module, and realizes accurate preservation of key features in the model simplification process. The method breaks through the limitation of traditional model simplification, and provides an intelligent and automatic solution for efficient processing of large-scale digital models.

[0207] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, device, or computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product in the form of being implemented 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.

[0208] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0209] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0210] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 Figure 1 one or more flowcharts and / or blocks

[0211] The principles and implementations of the present application are described in the specific embodiments of the present application, and the above descriptions of the embodiments are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation and application range will be changed, and the above descriptions should not be understood as the limitation of the present application.

Claims

1. An automated production line digital model simplification method, characterized by, The method includes: A salient feature extraction module is constructed. The salient feature extraction module calculates the salient score of point, line, surface and volume data in the digital model of automated production line. The feature points with the salient score exceeding the first threshold are extracted as a salient feature point set. At the same time, the morphological parameters and positional parameters corresponding to the salient feature point set are stored in the salient feature database. A four-level topological structure tree is established based on the salient feature point set. The four-level topological structure tree is input into the adaptive feature weight calculation module. The adaptive feature weight calculation module calculates the weight coefficient of each level based on the spatial distribution density of the salient feature point set. The weight coefficient is multiplied by the feature data of the corresponding level to obtain weighted feature data. The weighted feature data is input into the adaptive octree mesh construction module. The adaptive octree mesh construction module constructs mesh nodes based on the weighted feature data, embeds multi-resolution wavelet basis functions into the mesh nodes, performs hierarchical thresholding processing on the wavelet basis function coefficients, and reconstructs the thresholded wavelet coefficients into an initial simplified mesh. The initial simplified mesh is input into a multi-scale feature fusion module. The multi-scale feature fusion module extracts the topological features of the initial simplified mesh at a coarse scale and extracts the local geometric features of the initial simplified mesh at a fine scale. Based on the topological features and the local geometric features, a feature correspondence graph is constructed. The feature correspondence graph is matched with the morphological parameters and positional parameters in the salient feature database. Based on the matching results, the nodes of the initial simplified mesh are adjusted to generate the final simplified model.

2. The automated production line digital model simplification method of claim 1, wherein, The salient feature extraction module calculates salient scores for point, line, surface, and volume data in the automated production line digital model, and extracts feature points with salient scores exceeding a first threshold as a salient feature point set, including: The digital model of the automated production line is discretized to obtain point cloud data. A three-dimensional mesh structure is established based on the point cloud data. The curvature value, normal vector, and neighborhood density of each vertex in the three-dimensional mesh structure are calculated. The curvature value, normal vector, and neighborhood density are input into the significance calculation formula to obtain the significance coefficient. A significant feature extraction module is constructed based on the significance coefficient. The saliency coefficient in the saliency feature extraction module is compared with a first threshold. Vertices with saliency coefficients greater than the first threshold are extracted as feature points. A saliency feature point set is constructed based on the spatial positional relationship of the feature points. The geometric attribute parameters and topological attribute parameters of the saliency feature point set are stored as morphological parameters and positional parameters, respectively.

3. The automated production line digital model simplification method of claim 1, wherein, The step of storing the morphological and positional parameters corresponding to the salient feature point set into a salient feature database and establishing a four-level topological structure tree based on the salient feature point set includes: The shape parameters and the position parameters of the set of significant feature points are input into a feature encoder, the feature encoder performs geometric feature encoding on the shape parameters to obtain a geometric feature vector, performs spatial position encoding on the position parameters to obtain a position feature vector, combines the geometric feature vector and the position feature vector to form feature data, and writes the feature data into a significant feature database according to a hierarchical structure; A four-level topological structure tree is constructed based on the set of significant feature points, point data in the set of significant feature points is taken as a first-level node, a connection relationship between adjacent points is taken as a second-level node, a face patch enclosed by a closed line segment is taken as a third-level node, and a body element formed by a closed face patch is taken as a fourth-level node, and corresponding feature data indexes are stored in each node of the four-level topological structure tree.

4. The automated production line digital model simplification method of claim 1, wherein, The four-level topological structure tree is input into an adaptive feature weight calculation module, the adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the set of significant feature points, multiplies the weight coefficient by feature data of a corresponding level to obtain weighted feature data, and includes: A spatial density distribution map is constructed based on the four-level topological structure tree, a number of significant feature points in a neighborhood range of each node in the spatial density distribution map is calculated, the number is divided by a neighborhood volume to obtain a local density value, the local density value is normalized to obtain a density distribution coefficient, and the density distribution coefficient is input into a weight mapping function to generate a corresponding weight coefficient of each level; The weight coefficient and feature data of each level node in the four-level topological structure tree are subjected to matrix multiplication to generate a weighted feature matrix, the weighted feature matrix is subjected to standardization processing to obtain weighted feature data, and the weighted feature data is reorganized into a tree structure according to a hierarchical relationship.

5. The automated production line digital model simplification method of claim 1, wherein, The weighted feature data is input into an adaptive octree grid construction module, the adaptive octree grid construction module constructs a grid node based on the weighted feature data, embeds a multi-resolution wavelet basis function into the grid node, performs hierarchical thresholding processing on coefficients of the wavelet basis function, and reconstructs the thresholded wavelet coefficients into an initial simplified grid, and includes: A spatial bounding box is determined based on the weighted feature data, the spatial bounding box is recursively divided into an octree structure, a local variance of the weighted feature data is calculated at each node of the octree structure, the node is further subdivided when the local variance is greater than a second threshold value, until a preset number of subdivision levels is reached or the local variance is less than the second threshold value, an adaptive grid is constructed on the subdivided octree node, and a Haar wavelet basis function is mapped to node coordinates of the adaptive grid; Coefficients of the Haar wavelet basis function on the adaptive grid node are decomposed, an energy value of each layer of wavelet coefficients is calculated, an adaptive threshold value is set according to the energy value, wavelet coefficients less than the adaptive threshold value are set to zero, and a reconstruction operation is performed on the remaining wavelet coefficients to generate an initial simplified grid, and nodes of the initial simplified grid are associated with original weighted feature data.

6. The automated production line digital model simplification method of claim 1, wherein, The initial simplified mesh is input into a multi-scale feature fusion module, which extracts topological structure features of the initial simplified mesh at a coarse scale and local geometric features of the initial simplified mesh at a fine scale, including: Multi-level decomposition is performed on the initial simplified mesh to obtain mesh representations of different scales, a mesh connectivity graph is constructed at a coarse scale level, a degree distribution and connectivity property of each node in the mesh connectivity graph are calculated, a mesh skeleton is constructed based on the degree distribution and connectivity property, and a branch structure and connection relationship of the mesh skeleton are extracted as topological structure features; A local curvature tensor of a node of the initial simplified mesh is calculated at a fine scale level, principal curvature values and directions are calculated based on the curvature tensor, the principal curvature values and directions are combined to form a curvature descriptor, and local geometric features are obtained by clustering analysis on the curvature descriptor, and a mapping relationship between the local geometric features and the mesh node is established.

7. The automated production line digital model simplification method of claim 1, wherein, The topological structure features and the local geometric features are input into a feature fusion network, the feature fusion network generates a feature correspondence graph based on graph convolution operation, calculates a similarity matrix between nodes in the feature correspondence graph, performs matching operation on the similarity matrix and morphological parameters and position parameters stored in a significant feature database, generates a feature matching metric value, and constructs a node optimization objective function based on the feature matching metric value; The node optimization objective function is iteratively solved, the node position and connection relationship of the initial simplified mesh are adjusted, and when the objective function value is less than a third threshold value or the maximum iteration number is reached, the iteration is stopped, the final mesh node coordinates and connection relationship are reconstructed into a three-dimensional model, and the reconstructed three-dimensional model is subjected to mesh smoothing processing to obtain a final simplified model. The device comprises:

8. An automated production line digital model simplification apparatus, characterized by, A feature extraction module is configured to construct a saliency feature extraction module, calculate saliency scores of point, line, surface and volume data in an automated production line digital model, extract feature points with saliency scores exceeding a first threshold value as a set of salient feature points, store morphological parameters and position parameters corresponding to the set of salient feature points in a salient feature database, and establish a four-level topological structure tree based on the set of salient feature points. ​ An architecture building module is configured to input the four-level topology tree into an adaptive feature weight calculation module. The adaptive feature weight calculation module calculates a weight coefficient of each level based on a spatial distribution density of the set of salient feature points. The weight coefficient is multiplied by feature data of the corresponding level to obtain weighted feature data. The weighted feature data is input into an adaptive octree grid construction module. The adaptive octree grid construction module constructs grid nodes based on the weighted feature data. A multi-resolution wavelet basis function is embedded into the grid nodes. The wavelet basis function coefficients are processed by hierarchical thresholding. The thresholded wavelet coefficients are reconstructed into an initial simplified grid. A model simplification module is configured to input the initial simplified grid into a multi-scale feature fusion module. The multi-scale feature fusion module extracts topology structure features of the initial simplified grid at a coarse scale and extracts local geometric features of the initial simplified grid at a fine scale. A feature correspondence graph is constructed based on the topology structure features and the local geometric features. The feature correspondence graph is matched with morphological parameters and position parameters in the salient feature database. The initial simplified grid is adjusted based on the matching result to generate a final simplified model.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the automatic production line digital model simplification method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the automatic production line digital model simplification method of any one of claims 1 to 7.

Images