A complex structure modeling system and method based on three-dimensional curved surface skeleton extraction

Abstract

This application provides a complex structure modeling system and method based on 3D surface skeleton extraction. The system acquires initial 3D point cloud data of pipelines and converts it into a 3D voxel mesh. A symbolic distance field is generated based on the pipeline voxel mesh. Local extremum traversal is performed on the symbolic distance field to obtain a discrete skeleton candidate point set. Topological homology extraction is performed based on the discrete skeleton candidate point set to generate an initial pipeline topological skeleton. Semantic topological classification is performed using the pipeline voxel mesh and the initial pipeline topological skeleton to obtain structural attribute labels for each skeleton segment. Local topological optimization is performed on the initial pipeline topological skeleton based on the structural attribute labels of all skeleton segments to obtain an optimized pipeline topological skeleton. A parameterized boundary representation model of the complex pipeline network structure is generated using the optimized pipeline topological skeleton. The technical solution provided in this application can effectively construct the topological skeleton of complex pipeline network structures, thereby improving the modeling accuracy of parametric modeling.

Description

Technical Field

[0001] This application relates to the field of complex structure modeling technology, and more specifically, to a complex structure modeling system and method based on three-dimensional surface skeleton extraction. Background Technology

[0002] Complex structure modeling based on 3D surface skeleton extraction is an engineering modeling process that uses a series of geometric and topological processing steps on raw 3D data to obtain an engineering model that preserves shape details while providing a parametric representation. A typical workflow includes data preprocessing, constructing a signed distance field or voxel representation, detecting curve and surface skeleton candidates on the distance field or contraction transformation, fusing multi-source candidates to generate a connected skeleton graph, estimating continuous radius / thickness functions on the skeleton, and finally reconstructing the closed boundary based on the parametric centerline / central plane using methods such as section sweeping, implicit field fusion, or parametric surface stitching, and deriving a mesh or CAD representation suitable for simulation and manufacturing.

[0003] However, in existing technologies, complex structures often exhibit high curvature, rich local details, or regions with multiple intersecting branches. This makes it easy for pseudo-branches, topological breaks, or oversimplification to occur during skeleton extraction, resulting in the final model failing to accurately reflect the original geometry. Furthermore, thin plates (flanges, baffles) and narrow pipes in complex pipeline structures are geometrically easily confused; a single skeleton method may mistakenly shrink thin plates to centerlines or expand narrow pipes to central axis surfaces. Therefore, how to effectively construct the topological skeleton of complex pipeline networks to improve the modeling accuracy of parametric modeling remains a challenge for the industry. Summary of the Invention

[0004] This application provides a complex structure modeling system and method based on three-dimensional surface skeleton extraction, which can effectively construct the topological skeleton of complex pipeline network structures to improve the modeling accuracy of parametric modeling.

[0005] In a first aspect, this application provides a method for modeling complex structures based on three-dimensional curved surface skeleton extraction, comprising the following steps:

[0006] The initial three-dimensional point cloud data of the pipeline network structure is obtained, and then the initial three-dimensional point cloud data of the pipeline network structure is converted into a three-dimensional voxel mesh of the pipeline network structure.

[0007] Based on the three-dimensional voxel mesh of the pipeline, a pipeline symbolic distance field is generated. The pipeline symbolic distance field is subjected to local extremum traversal to obtain a discrete skeleton candidate point set. Based on the discrete skeleton candidate point set, topological homology extraction is performed to generate the initial pipeline topological skeleton of the complex pipeline network structure.

[0008] Semantic topology classification is performed using the pipeline 3D voxel mesh and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton. Local topology optimization is performed on the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments to obtain the pipeline topology optimization skeleton of the complex pipeline network structure.

[0009] The parameterized boundary representation model of complex pipeline network structures is generated through the pipeline topology optimization skeleton.

[0010] In some embodiments, converting the initial pipeline 3D point cloud data into a pipeline 3D voxel mesh with a complex pipeline network structure specifically includes:

[0011] Fast neighborhood statistics are performed on the initial pipeline 3D point cloud data to obtain the local sampling scale of each point in the initial pipeline 3D point cloud data;

[0012] The local voxel size of each point is determined based on the corresponding local sampling scale;

[0013] Sparse voxel structures are generated using the local voxel size of all points and an adaptive octree.

[0014] The sparse voxel structure is subjected to three-dimensional morphological processing to obtain a three-dimensional voxel mesh of a complex pipeline network structure.

[0015] In some embodiments, generating a pipe symbolic distance field based on the pipe three-dimensional voxel mesh specifically includes:

[0016] The unsigned Euclidean distance from each voxel in the three-dimensional voxel mesh of the pipeline to the nearest pipeline boundary is determined, and thus the unsigned distance field is obtained;

[0017] By emitting rays from each voxel location and determining the number of intersections with the pipe surface, the internal and external relationships of each voxel are determined, and the symbol information of each voxel is obtained.

[0018] The pipeline symbolic distance field is generated based on the unsigned distance field and the symbolic information of all voxels.

[0019] In some embodiments, performing a local extremum traversal on the pipe symbol distance field to obtain a discrete skeleton candidate point set specifically includes:

[0020] Traverse all voxels marked as being inside the pipe in the pipe symbol distance field;

[0021] For each currently traversed voxel, obtain the distance values ​​of all neighboring voxels within the 26-neighborhood of the currently traversed voxel;

[0022] Candidate determination is performed by using the distance value of the currently traversed voxel and the distance values ​​of all adjacent voxels to obtain the candidate determination result of the currently traversed voxel, and then the discrete skeleton candidate point set is obtained.

[0023] In some embodiments, the process of extracting topological homology based on the discrete skeleton candidate point set to generate an initial pipeline topology skeleton for a complex pipeline network specifically includes:

[0024] Assign a height attribute to each skeleton candidate point in the discrete skeleton candidate point set, and treat each skeleton candidate point as an independent connected component;

[0025] Based on a set dynamic threshold, all connected components are dynamically merged to obtain multiple merged components, and then the persistence value of each merged component is determined.

[0026] All topology representative nodes are filtered out using a pre-defined persistence threshold and the persistence values ​​of all merged components;

[0027] All topological representative nodes are connected according to the merging relationship to form a graph structure, thereby generating the initial pipeline topology skeleton of the complex pipeline network structure.

[0028] In some embodiments, performing local topology optimization on the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments specifically includes:

[0029] When the structural attribute label of the skeleton segment is straight pipe, constrained linear optimization is performed between the two endpoints of the skeleton segment.

[0030] When the structural attribute label of the skeleton segment is elbow, perform arc fitting optimization between the two endpoints of the skeleton segment.

[0031] When the structural attribute label of a skeleton segment is "branch", the position of the skeleton segment is optimized.

[0032] In some embodiments, generating a parameterized boundary representation model of a complex pipeline network structure through the pipeline topology optimization skeleton involves inputting the pipeline topology optimization skeleton into CAD and generating the parameterized boundary representation model of the complex pipeline network structure through the geometric modeling function of the CAD kernel.

[0033] Secondly, this application provides a complex structure modeling system based on three-dimensional surface skeleton extraction, used to execute a complex structure modeling method based on three-dimensional surface skeleton extraction, including:

[0034] The mesh construction module is used to acquire initial 3D point cloud data of complex pipeline network structures, and then convert the initial 3D point cloud data of pipelines into 3D voxel meshes of complex pipeline network structures.

[0035] The skeleton generation module is used to generate a pipeline symbolic distance field based on the pipeline three-dimensional voxel mesh, perform local extremum traversal on the pipeline symbolic distance field to obtain a discrete skeleton candidate point set, and perform topological homology extraction based on the discrete skeleton candidate point set to generate an initial pipeline topology skeleton of a complex pipeline network structure.

[0036] The skeleton optimization module is used to perform semantic topology classification using the pipeline 3D voxel mesh and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton, and to perform local topology optimization on the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments, thereby obtaining the pipeline topology optimized skeleton of the complex pipeline network structure.

[0037] The model generation module is used to generate a parameterized boundary representation model of a complex pipeline network structure from the pipeline topology optimization skeleton.

[0038] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described complex structure modeling method based on three-dimensional curved surface skeleton extraction.

[0039] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for complex structure modeling based on three-dimensional surface skeleton extraction.

[0040] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0041] This application provides a complex structure modeling system and method based on 3D surface skeleton extraction. The system acquires initial 3D point cloud data of a complex pipeline network structure and converts this data into a 3D voxel mesh. A symbolic distance field is generated based on the voxel mesh. Local extremum traversal is performed on the symbolic distance field to obtain a discrete skeleton candidate point set. Topological homology extraction is performed on this candidate point set to generate an initial pipeline topological skeleton for the complex pipeline network structure. Semantic topological classification is performed using the voxel mesh and the initial topological skeleton to obtain structural attribute labels for each skeleton segment. Local topological optimization is performed on the initial topological skeleton based on the structural attribute labels of all skeleton segments to obtain an optimized topological skeleton for the complex pipeline network structure. Finally, a parameterized boundary representation model of the complex pipeline network structure is generated using the optimized topological skeleton.

[0042] Therefore, in this application, firstly, during the modeling of complex pipeline network structures, by acquiring high-precision 3D point cloud data and converting it into an adaptive voxelized 3D voxel mesh, irregular point clouds can be transformed into regular spatial discrete structures, effectively filtering out isolated points and local noise while preserving the main geometric features. Secondly, by first generating a signed distance field based on the pipeline's 3D voxel mesh, and then performing local extremum traversal within it to obtain a set of discrete skeleton candidate points, and further combining this with topological homology extraction to generate an initial pipeline topological skeleton, an automated transition from geometric information to topological structure can be achieved in complex pipeline networks. Furthermore, persistent homology effectively filters out locally fragmented pseudo-skeleton points, retaining only the truly important main trunks and branches, thus... An initial pipeline topology skeleton with topological consistency and geometric accuracy is formed. Then, by combining the 3D voxel mesh and the initial pipeline topology skeleton for semantic topology classification, and adopting corresponding local topology optimization strategies for different categories of skeleton segments, the accuracy and regularity of the skeleton geometric expression can be significantly improved while maintaining the overall topology structure. Finally, a parameterized boundary representation model of the complex pipeline network structure is generated by optimizing the pipeline topology skeleton. This enables an efficient transition from abstract skeleton to precise geometry, closely integrating skeleton information with the boundary modeling function of the CAD kernel. This ensures the correctness of the overall topology structure and allows for rapid modification and precise adjustment through parameter-driven methods, thereby significantly improving the modeling accuracy and flexibility of complex pipeline networks.

[0043] In summary, the technical solution adopted in this application can effectively construct the topological skeleton of complex pipeline network structures, thereby improving the modeling accuracy of parametric modeling. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this embodiment of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is an exemplary flowchart of a complex structure modeling method based on three-dimensional surface skeleton extraction, according to some embodiments of this application;

[0046] Figure 2 This is an exemplary flowchart illustrating the generation of an initial pipeline topology skeleton for a complex pipeline network structure according to some embodiments of this application;

[0047] Figure 3 This is a schematic diagram of a complex structure modeling system based on three-dimensional curved surface skeleton extraction, as shown in some embodiments of this application;

[0048] Figure 4 This is a schematic diagram of the structure of a computer device that implements a complex structure modeling method based on three-dimensional surface skeleton extraction, according to some embodiments of this application. Detailed Implementation

[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0050] This application provides a complex structure modeling system and method based on 3D surface skeleton extraction. The core of the system is to acquire initial 3D point cloud data of a complex pipeline network structure, and then convert this initial 3D point cloud data into a 3D voxel mesh of the complex pipeline network structure. A pipeline symbolic distance field is generated based on the pipeline 3D voxel mesh. Local extremum traversal is performed on the pipeline symbolic distance field to obtain a discrete skeleton candidate point set. Topological homology extraction is performed based on the discrete skeleton candidate point set to generate an initial pipeline topological skeleton of the complex pipeline network structure. Semantic topological classification is performed using the pipeline 3D voxel mesh and the initial pipeline topological skeleton to obtain structural attribute labels for each skeleton segment in the initial pipeline topological skeleton. Local topological optimization is performed on the initial pipeline topological skeleton based on the structural attribute labels of all skeleton segments to obtain a pipeline topologically optimized skeleton of the complex pipeline network structure. A parameterized boundary representation model of the complex pipeline network structure is generated using the pipeline topologically optimized skeleton. This approach can effectively construct the topological skeleton of complex pipeline network structures, thereby improving the modeling accuracy of parameterized modeling.

[0051] To better understand the above technical solutions, a detailed description of the technical solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. (Refer to...) Figure 1 This figure is an exemplary flowchart of a complex structure modeling method based on three-dimensional curved surface skeleton extraction according to some embodiments of this application. The figure mainly includes the following steps:

[0052] In step S101, initial three-dimensional point cloud data of the complex pipeline network structure is obtained, and then the initial three-dimensional point cloud data of the pipeline network structure is converted into a three-dimensional voxel mesh of the complex pipeline network structure.

[0053] In practice, tripod-mounted lidar and phase-detection laser scanning can be used to obtain initial 3D point cloud data of complex pipeline network structures.

[0054] In some embodiments, converting the initial 3D point cloud data of the pipeline into a 3D voxel mesh of a complex pipeline network structure can be achieved in the following manner:

[0055] Fast neighborhood statistics are performed on the initial pipeline 3D point cloud data to obtain the local sampling scale of each point in the initial pipeline 3D point cloud data;

[0056] The local voxel size of each point is determined based on the corresponding local sampling scale;

[0057] Sparse voxel structures are generated using the local voxel size of all points and an adaptive octree.

[0058] The sparse voxel structure is subjected to three-dimensional morphological processing to obtain a three-dimensional voxel mesh of a complex pipeline network structure.

[0059] In practical implementation, firstly, a rapid neighborhood statistics can be performed on the initial pipeline 3D point cloud data. This involves using the distance from each point to its k-th nearest neighbor as the local sampling scale. The value of K can be set based on historical experience. The local sampling scale reflects the sampling density and the smallest resolvable geometric feature at that point, thus obtaining the local sampling scale of each point in the initial pipeline 3D point cloud data. Then, the local voxel size of each point can be determined based on the corresponding local sampling scale. This involves mapping the local sampling scale of each point to a local voxel size, and using a preset scaling factor to enlarge or reduce the local sampling scale and limit it between a preset minimum and maximum voxel size to avoid excessive fineness leading to memory expansion or excessive coarseness resulting in detail loss. Furthermore, a sparse voxel structure can be generated using the local voxel sizes of all points and an adaptive octree. Based on the local voxel sizes of all points, an adaptive voxel structure is constructed using a top-down octree. Starting from the root node covering the scene, each node is further subdivided until the size of the leaf node approaches the target voxel size of the region, even if the number of points it contains or the local target size is not yet satisfied. The leaf nodes simultaneously record statistical information such as point count, average normal of points, and minimum / maximum depth. Occupied voxels are marked using occupancy criteria (e.g., the number of points within the leaf node exceeds a threshold or the occupancy ratio is higher than a threshold), and a sparse voxel structure is generated. Finally, three-dimensional morphological processing can be performed on the sparse voxel structure. This involves performing morphological processing on the sparse voxel structure to achieve noise reduction and smoothing. First, a closing operation (dilation followed by erosion) is performed on the region-adaptive structuring elements to fill small holes. The structuring element size is set according to the local voxel scale of the leaf node and scaled down in areas previously semantically determined to be thin plates or flanges. Small structuring elements are used to avoid filling thin sheets, and then an adaptive opening operation (erosion followed by dilation) is used to remove isolated noise clusters. The intensity of the opening operation can be increased in low-confidence regions, thereby performing connected component analysis on the processed voxels to remove isolated clumps with volumes below the physical or voxel number thresholds. Optionally, in narrow-band regions, multi-resolution voxels can be projected or resampled into a uniform fine-scale mesh for subsequent distance field calculations and skeleton extraction, thereby outputting a 3D voxel mesh of a complex pipeline network structure, where each voxel mesh contains an occupancy flag, point count, average normal, and local voxel size.

[0060] It is worth noting that in the modeling of complex pipeline network structures, acquiring high-precision 3D point cloud data and converting it into an adaptive voxelized 3D voxel mesh plays a crucial role in data organization and structural representation. While point cloud data can completely record the geometric information of the pipeline surface, its distribution often contains unevenness and noise. Directly using it for topological skeleton extraction can lead to structural breaks or false branches. However, voxelization can transform irregular point clouds into regular spatial discrete structures. By leveraging the spatial occupancy information of voxels, a unified representation of the overall pipeline morphology can be achieved, while effectively filtering out isolated points and local noise while preserving the main geometric features.

[0061] In step S102, a pipeline symbolic distance field is generated based on the pipeline three-dimensional voxel mesh. The pipeline symbolic distance field is subjected to local extremum traversal to obtain a discrete skeleton candidate point set. Based on the discrete skeleton candidate point set, topological homology extraction is performed to generate the initial pipeline topology skeleton of the complex pipeline network structure.

[0062] In some embodiments, the generation of the pipe symbolic distance field based on the pipe three-dimensional voxel mesh can be specifically carried out in the following manner:

[0063] The unsigned Euclidean distance from each voxel in the three-dimensional voxel mesh of the pipeline to the nearest pipeline boundary is determined, and thus the unsigned distance field is obtained;

[0064] By emitting rays from each voxel location and determining the number of intersections with the pipe surface, the internal and external relationships of each voxel are determined, and the symbol information of each voxel is obtained.

[0065] The pipeline symbolic distance field is generated based on the unsigned distance field and the symbolic information of all voxels.

[0066] In practical implementation, firstly, the unsigned Euclidean distance from each voxel in the 3D voxel mesh of the pipeline to the nearest pipeline boundary can be determined. This is achieved by using a fast scan or distance transformation algorithm to traverse each voxel on the 3D voxel mesh of the pipeline and find the nearest boundary voxel, obtaining the shortest distance from each voxel to the pipeline surface, i.e., the unsigned Euclidean distance from the voxel to the nearest pipeline boundary. This unsigned Euclidean distance accurately reflects the spatial position of the voxel inside or outside the pipeline, thus combining the dataset of unsigned Euclidean distances from all voxels to the nearest pipeline boundary into an unsigned distance field. Then, by emitting rays from each voxel location and determining the number of intersections with the pipeline surface, i.e., emitting rays from the voxel center, the distances can be calculated. The number of intersections between the ray and the pipe surface is used to determine the internal / external relationship of each voxel. Based on the odd / even rule, it is determined whether the voxel is located inside or outside the pipe. If the number of intersections is odd, it is considered internal and assigned a positive sign; if it is even, it is considered external and assigned a negative sign, thus obtaining the sign information for each voxel. Finally, a pipe-signed distance field can be generated based on the unsigned distance field and the sign information of all voxels. This involves combining the obtained sign information with the unsigned distance field, adding the sign to the distance value of each voxel, and generating a complete signed distance field. This ensures that the distances of internal voxels are positive, external voxels are negative, and the pipe surface transitions to zero. Such a signed distance field preserves the precise geometric information of the pipe while clearly defining the internal / external spatial relationships of the voxels.

[0067] In some embodiments, performing a local extremum traversal on the pipe symbol distance field to obtain a discrete skeleton candidate point set can be done in the following manner:

[0068] Traverse all voxels marked as being inside the pipe in the pipe symbol distance field;

[0069] For each currently traversed voxel, obtain the distance values ​​of all neighboring voxels within the 26-neighborhood of the currently traversed voxel;

[0070] Candidate determination is performed by using the distance value of the currently traversed voxel and the distance values ​​of all adjacent voxels to obtain the candidate determination result of the currently traversed voxel, and then the discrete skeleton candidate point set is obtained.

[0071] In specific implementation, firstly, all voxels marked as being inside the pipe can be traversed in the pipe symbol distance field. Then, during the traversal, for each currently traversed voxel, the distance values ​​of all neighboring voxels within its 26-neighborhood can be obtained, including the six directions of up, down, left, right, front, and back, as well as all diagonal directions, totaling 26 neighbors. This yields the corresponding distance values ​​of the 26 neighboring voxels in the pipe symbol distance field. Furthermore, candidate determination can be performed by comparing the distance value of the currently traversed voxel in the pipe symbol distance field with the distance values ​​of all neighboring voxels. That is, the distance value of the currently traversed voxel is compared with the distance values ​​of its neighboring voxels. If the distance value of the currently traversed voxel is greater than the distance values ​​of all neighboring voxels, it is determined as a local maximum point. In other words, the candidate determination result of the currently traversed voxel is that it belongs to the skeleton candidate point. The skeleton candidate point is a candidate for potential skeleton nodes, thus the dataset composed of all skeleton candidate points can be converted into a discrete skeleton candidate point set.

[0072] Preferably, in some embodiments, reference is made to Figure 2 As shown in the figure, this is an exemplary flowchart of generating an initial pipeline topology skeleton for a complex pipeline network structure according to some embodiments of this application. In this embodiment, the initial pipeline topology skeleton for a complex pipeline network structure is generated by extracting topological homology based on the discrete skeleton candidate point set, which can be achieved by the following steps:

[0073] In step S1021, each skeleton candidate point in the discrete skeleton candidate point set is assigned a height attribute, and each skeleton candidate point is treated as an independent connected component.

[0074] In step S1022, all connected components are dynamically merged according to the set dynamic threshold to obtain multiple merged components, and then the persistence value of each merged component is determined.

[0075] In step S1023, all topological representative nodes are filtered out using a pre-set persistence threshold and the persistence values ​​of all merged components;

[0076] In step S1024, all topology representative nodes are connected into a graph structure according to the merging relationship, thereby generating the initial pipeline topology skeleton of the complex pipeline network structure.

[0077] In specific implementation, firstly, each skeleton candidate point in the discrete skeleton candidate point set can be assigned a height attribute, that is, the distance value corresponding to the skeleton candidate point in the symbolic distance field can be used as the height attribute, and each skeleton candidate point is initially regarded as an independent connected component. Then, all connected components can be dynamically merged according to a set dynamic threshold, that is, the "flooding" process is simulated from high to low according to the height attribute. An efficient disjoint-set data structure and spatial index (such as kd-tree) are used for neighborhood search. When a point at a dynamic threshold is activated, its activated neighbors within a given connection radius are searched and their corresponding connected components are merged. The dynamic threshold and the information of the participating components are recorded for each merge, thus obtaining the complete merged component. Here, the dynamic threshold is a dynamic height threshold. Furthermore, the persistence value of each merged component can be determined. The persistence value is used to represent the importance of the merged component. The "birth height" of each original component is its own height, and the death height is the dynamic threshold when it is merged into a larger component. The difference between the two is the persistence value of the merged component.

[0078] In addition, in specific implementation, all topological representative nodes can be filtered out using a pre-set persistence threshold and the persistence values ​​of all merged components. That is, merged components with persistence values ​​greater than the pre-set persistence threshold are filtered out, and the point with the largest height in the filtered merged components is selected as the topological representative node of that merged component, thus filtering out all topological representative nodes. Finally, all topological representative nodes can be connected into a graph structure according to the merging relationship. That is, each time two merged components are merged at a certain dynamic threshold, a weighted edge is established between their topological representative nodes. The weight can be the Euclidean distance between the topological representative nodes. In this way, the graph directly generated by the merging event reflects both the branch hierarchy and the merging sequence. To retain possible real closed loops, a local adjacency complex can be additionally constructed on the representative nodes, and a supplementary persistence analysis can be performed on the one-dimensional topological features. The high-persistence loop edges are added back to the skeleton graph. In this way, the initial pipeline topology skeleton of the complex pipeline network structure can be generated.

[0079] It should be noted that by first generating a symbolic distance field based on the three-dimensional voxel mesh of the pipeline, then performing local extremum traversal within it to obtain a discrete skeleton candidate point set, and further combining it with topological homology extraction to generate an initial pipeline topological skeleton, an automated transition from geometric information to topological structure can be achieved in complex pipeline networks. This not only utilizes the symbolic distance field to ensure that the skeleton points are located in the geometric center region inside the pipeline, avoiding offsets caused by noise or sparse sampling, but also effectively filters out locally fragmented pseudo-skeleton points through persistent homology, retaining only the truly important trunks and branches in the structure, thus forming an initial pipeline topological skeleton with topological consistency and geometric accuracy.

[0080] In step S103, semantic topology classification is performed using the three-dimensional voxel mesh of the pipeline and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton. Based on the structural attribute labels of all skeleton segments, local topology optimization is performed on the initial pipeline topology skeleton to obtain the pipeline topology optimization skeleton of the complex pipeline network structure.

[0081] In some embodiments, semantic topology classification is performed using the pipeline 3D voxel mesh and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton. Specifically, this can be achieved in the following manner:

[0082] For each skeleton segment in the initial pipeline topology skeleton, a support domain for the skeleton segment is constructed based on the pipeline three-dimensional voxel mesh;

[0083] Based on the support domain, multidimensional descriptive features of the skeleton segment are extracted;

[0084] The multidimensional descriptive features are normalized and dimensionality reduced, and the skeleton segments are semantically classified using a pre-trained classifier to obtain the structural attribute labels of the skeleton segments, thereby obtaining the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton.

[0085] In practical implementation, firstly, for each skeleton segment in the initial pipeline topology skeleton, a support domain for the skeleton segment can be constructed based on the pipeline's 3D voxel mesh. That is, each skeleton segment in the initial pipeline topology skeleton is treated as a basic object. By spatially extending the skeleton segment outward into a domain based on a local radius, voxels falling into this domain and the original point cloud point set in the pipeline's 3D voxel mesh are mapped back to the skeleton segment, thus establishing the association between the segment and the voxels. The resulting domain is then used as the support domain for the skeleton segment. Next, multidimensional descriptive features of the skeleton segment are extracted based on the support domain. Specifically, the segment length, node degree, average radius and radius variance, axial gradient of the radius, and approximate curvature and torsion of the skeleton segment can be calculated within the support domain. The system considers factors such as the value of the segment, the angle between adjacent segments, the density and normal consistency of voxels around the segment, the local thickness statistics of voxels, the durability score, and the matching degree with the predefined component template in geometric scale and feature space. These features are then used as a set of multidimensional descriptive features. Finally, the multidimensional descriptive features are normalized and dimensionality reduced, and a pre-trained classifier is used to perform semantic classification of the skeleton segments. This involves inputting the processed multidimensional descriptive features into a pre-trained classifier for semantic classification, such as straight pipes, elbows, and branches, while outputting confidence scores. This yields the structural attribute labels for the skeleton segments. The above method allows for the acquisition of the structural attribute labels for each skeleton segment in the initial pipeline topology skeleton.

[0086] In some embodiments, local topology optimization of the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments can be performed in the following manner:

[0087] When the structural attribute label of the skeleton segment is straight pipe, constrained linear optimization is performed between the two endpoints of the skeleton segment.

[0088] When the structural attribute label of the skeleton segment is elbow, perform arc fitting optimization between the two endpoints of the skeleton segment.

[0089] When the structural attribute label of a skeleton segment is "branch", the position of the skeleton segment is optimized.

[0090] In specific implementation, firstly, a set of support points is collected for each skeleton segment (all surface points falling within the influence range of the skeleton segment are captured from the pipe's 3D voxel mesh according to the support domain of the skeleton segment), and weights are assigned to these points based on historical experiments and data analysis, which will not be elaborated here. Then, when the structural attribute label of the skeleton segment is straight pipe, constrained linear optimization is performed between the two endpoints of the skeleton segment, that is, constrained straight line least squares fitting is used. First, the optimal fitting direction of the segment is obtained by singular value decomposition. Then, under the premise of ensuring the connection constraints of the two end nodes (e.g., the endpoints or the endpoints are allowed to move slightly along the adjacent segments) and maintaining the consistency of length or endpoints, the linear optimization problem is solved with the goal of minimizing the vertical residual from the point on the segment to the straight line. If necessary, a regularization term is added to constrain the movement to not exceed the proportion of the local radius to prevent overcorrection. When the structural attribute label of the skeleton segment is elbow, circular arc fitting optimization is performed between the two endpoints of the skeleton segment, that is, first, an optimal plane is fitted to the point set of the skeleton segment (the local plane is determined by principal component analysis), the points are projected onto this plane and a robust circular arc is used. The center and radius of the circle are obtained using fitting methods (such as algebraic circle fitting or nonlinear least squares based on minimizing the projection residual). Then, nonlinear minimization is used to further apply endpoint tangent continuity constraints to ensure first-order continuity with adjacent segments. Second-order curvature continuity can also be maintained if necessary. When the structural attribute label of a skeleton segment is "branch," the skeleton segment undergoes position optimization, i.e., node position optimization. Branch nodes are treated as free variables, and the objective function is established as the weighted squared distance from the center of the support section of each connected segment to the node (or in the form of local energy to minimize angular misalignment and curvature abrupt changes). Under constraints, an iterative solver (such as the Newton or Levenberg-Marquardt method) is used to adjust the node positions, while limiting the maximum node displacement, preserving the original topology, and avoiding intersections with voxel shapes. Finally, the local topology optimization of the initial pipeline skeleton is completed through the above methods, thus using the locally optimized initial pipeline skeleton as the pipeline topology optimization skeleton for complex pipeline network structures.

[0091] It should be noted that by combining 3D voxel meshes and the initial pipeline topology skeleton for semantic topology classification, and employing corresponding local topology optimization strategies for different categories of skeleton segments, the accuracy and regularity of the skeleton's geometric representation can be significantly improved while maintaining the overall topology structure. Straight pipe segments, after linear constraint optimization, better reflect the axis of the actual pipeline; elbow segments achieve smooth transitions through arc fitting; and branch nodes achieve reasonable connection angles and spatial layouts through position optimization, effectively avoiding skeleton distortion and irregularities caused by scanning noise or uneven point clouds. This semantically driven optimization ensures that the final generated topology skeleton possesses both clear structural semantics and accurately corresponds to actual geometric features, providing stable and high-precision skeleton support for subsequent parametric modeling, thus improving the overall modeling accuracy and engineering usability.

[0092] In step S104, a parameterized boundary representation model of the complex pipeline network structure is generated through the pipeline topology optimization skeleton.

[0093] In some embodiments, generating a parameterized boundary representation model of a complex pipeline network structure through the pipeline topology optimization skeleton involves inputting the pipeline topology optimization skeleton into CAD and generating the parameterized boundary representation model of the complex pipeline network structure through the geometric modeling function of the CAD kernel.

[0094] In practice, the CAD kernel first automatically calls the corresponding modeling module based on the structural attribute tags (such as straight pipes, elbows, and branches) of the skeleton segments. For example, it generates a cylindrical boundary representation with adjustable radius and length on straight pipe segments, generates bend elements with parameterized curvature radii from the skeleton arc fitting results on elbow segments, and constructs multi-path pipe segment models that satisfy connection constraints at branch nodes. Then, these elements are automatically spliced ​​into a complete parameterized boundary representation model through Boolean operations, boundary fusion, and continuous smoothing. This parameterized boundary representation model not only accurately reflects the geometric characteristics of complex pipeline networks but also has high adjustability and editability, providing high-precision parameterized support for subsequent design modifications, engineering simulations, and construction applications.

[0095] It should be noted that generating a parameterized boundary representation model of a complex pipeline network structure through a pipeline topology optimization skeleton can achieve an efficient transition from an abstract skeleton to precise geometry. This allows the model to not only possess realistic spatial topological relationships but also to parametrically express the geometric features and dimensional constraints of different components such as straight pipes, elbows, and branches. By closely integrating skeleton information with the boundary modeling function of the CAD kernel, the correctness of the overall topology structure is ensured, while parameter-driven rapid modification and precise adjustment are also possible, thereby significantly improving the modeling accuracy and flexibility of complex pipeline networks.

[0096] Therefore, in this application, firstly, during the modeling of complex pipeline network structures, by acquiring high-precision 3D point cloud data and converting it into an adaptive voxelized 3D voxel mesh, irregular point clouds can be transformed into regular spatial discrete structures, effectively filtering out isolated points and local noise while preserving the main geometric features. Secondly, by first generating a signed distance field based on the pipeline's 3D voxel mesh, and then performing local extremum traversal within it to obtain a set of discrete skeleton candidate points, and further combining this with topological homology extraction to generate an initial pipeline topological skeleton, an automated transition from geometric information to topological structure can be achieved in complex pipeline networks. Furthermore, persistent homology effectively filters out locally fragmented pseudo-skeleton points, retaining only the truly important main trunks and branches, thus... An initial pipeline topology skeleton with topological consistency and geometric accuracy is formed. Then, by combining the 3D voxel mesh and the initial pipeline topology skeleton for semantic topology classification, and adopting corresponding local topology optimization strategies for different categories of skeleton segments, the accuracy and regularity of the skeleton geometric expression can be significantly improved while maintaining the overall topology structure. Finally, a parameterized boundary representation model of the complex pipeline network structure is generated by optimizing the pipeline topology skeleton. This enables an efficient transition from abstract skeleton to precise geometry, closely integrating skeleton information with the boundary modeling function of the CAD kernel. This ensures the correctness of the overall topology structure and allows for rapid modification and precise adjustment through parameter-driven methods, thereby significantly improving the modeling accuracy and flexibility of complex pipeline networks.

[0097] In summary, the technical solution adopted in this application can effectively construct the topological skeleton of complex pipeline network structures, thereby improving the modeling accuracy of parametric modeling.

[0098] Furthermore, in another aspect of this application, in some embodiments, this application provides a complex structure modeling system based on three-dimensional surface skeleton extraction, referencing... Figure 3 The figure is a schematic diagram of a complex structure modeling system based on three-dimensional surface skeleton extraction, according to some embodiments of this application. The complex structure modeling system based on three-dimensional surface skeleton extraction includes:

[0099] Mesh construction module 201 is used to acquire initial three-dimensional point cloud data of pipelines in complex pipeline network structure, and then convert the initial three-dimensional point cloud data of pipelines into a three-dimensional voxel mesh of pipelines in complex pipeline network structure.

[0100] The skeleton generation module 202 is used to generate a pipeline symbolic distance field based on the pipeline three-dimensional voxel mesh, perform local extremum traversal on the pipeline symbolic distance field to obtain a discrete skeleton candidate point set, and perform topological homology extraction based on the discrete skeleton candidate point set to generate an initial pipeline topology skeleton of a complex pipeline network structure.

[0101] The skeleton optimization module 203 is used to perform semantic topology classification using the pipeline three-dimensional voxel mesh and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton, and to perform local topology optimization on the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments, thereby obtaining the pipeline topology optimized skeleton of the complex pipeline network structure.

[0102] The model generation module 204 is used to generate a parameterized boundary representation model of a complex pipeline network structure through the pipeline topology optimization skeleton.

[0103] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described complex structure modeling method based on three-dimensional curved surface skeleton extraction.

[0104] In some embodiments, reference Figure 4 The figure is a schematic diagram of a computer device implementing a complex structure modeling method based on three-dimensional surface skeleton extraction, according to some embodiments of this application. The complex structure modeling method based on three-dimensional surface skeleton extraction in the above embodiments can be implemented through... Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0105] The processor 301 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the complex structure modeling method based on three-dimensional surface skeleton extraction in this application.

[0106] The communication bus 302 can be used to transmit information between the aforementioned components.

[0107] The memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0108] The memory 303 stores program code for executing the scheme of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the determination of the complex structure modeling method based on three-dimensional curved surface skeleton extraction can be achieved by the processor 301 and one or more software modules in the program code in the memory 303.

[0109] Communication interface 304 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0110] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0111] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0112] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described complex structure modeling method based on three-dimensional curved surface skeleton extraction.

[0113] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0114] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A complex structure modeling method based on three-dimensional curved surface skeleton extraction, characterized in that, Includes the following steps: The initial three-dimensional point cloud data of the pipeline network structure is obtained, and then the initial three-dimensional point cloud data of the pipeline network structure is converted into a three-dimensional voxel mesh of the pipeline network structure. Based on the three-dimensional voxel mesh of the pipeline, a pipeline symbolic distance field is generated. The pipeline symbolic distance field is subjected to local extremum traversal to obtain a discrete skeleton candidate point set. Based on the discrete skeleton candidate point set, topological homology extraction is performed to generate the initial pipeline topology skeleton of the complex pipeline network structure. Semantic topology classification is performed using the pipeline 3D voxel mesh and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton. Based on the structural attribute labels of all skeleton segments, local topology optimization is performed on the initial pipeline topology skeleton to obtain the pipeline topology optimization skeleton of the complex pipeline network structure. The parameterized boundary representation model of complex pipeline network structure is generated through the pipeline topology optimization skeleton. Specifically, the process of extracting topological homology based on the discrete skeleton candidate point set to generate the initial pipeline topology skeleton for the complex pipeline network structure includes: Each skeleton candidate point in the discrete skeleton candidate point set is assigned a height attribute, and each skeleton candidate point is treated as an independent connected component. All connected components are dynamically merged according to a set dynamic threshold to obtain multiple merged components, and then the persistence value of each merged component is determined. All topology representative nodes are selected using a pre-set persistence threshold and the persistence values ​​of all merged components. All topology representative nodes are connected into a graph structure according to the merging relationship, thereby generating the initial pipeline topology skeleton of the complex pipeline network structure. The process involves dynamically merging all connected components based on a set dynamic threshold. This simulates a flooding process from high to low height attributes, using a disjoint-set data structure and spatial indexing for neighborhood lookup. When a point at a dynamic threshold is activated, its activated neighbors within a given connection radius are searched, and their corresponding connected components are merged. The dynamic threshold and information of the participating components are recorded for each merge, resulting in a complete merged component. The dynamic threshold is a dynamic height threshold. The persistence value is an indicator of the importance of the merged component. The birth height of each original component is its own height, and the death height is the dynamic threshold at which it is merged into a larger component. The difference between the birth height and the death height is used as the persistence value of the merged component, thus determining the persistence value of each merged component.

2. The method of claim 1, wherein, Converting the initial 3D point cloud data of the pipeline into a 3D voxel mesh of a complex pipeline network structure specifically includes: Fast neighborhood statistics are performed on the initial pipeline 3D point cloud data to obtain the local sampling scale of each point in the initial pipeline 3D point cloud data; The local voxel size of each point is determined based on the corresponding local sampling scale; Sparse voxel structures are generated using the local voxel size of all points and an adaptive octree. The sparse voxel structure is subjected to three-dimensional morphological processing to obtain a three-dimensional voxel mesh of a complex pipeline network structure.

3. The method of claim 1, wherein the three-dimensional curved surface skeleton is extracted from the complex structure model. The generation of the pipe symbolic distance field based on the three-dimensional voxel mesh of the pipe specifically includes: The unsigned Euclidean distance from each voxel in the three-dimensional voxel mesh of the pipeline to the nearest pipeline boundary is determined, and thus the unsigned distance field is obtained; By emitting rays from each voxel location and determining the number of intersections with the pipe surface, the internal and external relationships of each voxel are determined, and the symbol information of each voxel is obtained. The pipeline symbolic distance field is generated based on the unsigned distance field and the symbolic information of all voxels.

4. The method of claim 1, wherein the complex structure modeling based on three-dimensional curved surface skeleton extraction is characterized by, The local extremum traversal of the distance field of the pipeline symbol yields a discrete skeleton candidate point set, specifically including: Traverse all voxels marked as being inside the pipe in the pipe symbol distance field; For each currently traversed voxel, obtain the distance values ​​of all neighboring voxels within the 26-neighborhood of the currently traversed voxel; Candidate determination is performed by using the distance value of the currently traversed voxel and the distance values ​​of all adjacent voxels to obtain the candidate determination result of the currently traversed voxel, and then the discrete skeleton candidate point set is obtained.

5. The method of claim 1, wherein the complex structure modeling based on three-dimensional curved surface skeleton extraction is characterized by, Local topology optimization of the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments specifically includes: When the structural attribute label of the skeleton segment is straight pipe, constrained linear optimization is performed between the two endpoints of the skeleton segment. When the structural attribute label of the skeleton segment is elbow, perform arc fitting optimization between the two endpoints of the skeleton segment. When the structural attribute label of a skeleton segment is "branch", the position of the skeleton segment is optimized.

6. The method of claim 1, wherein, The process of generating a parameterized boundary representation model of a complex pipeline network structure using the pipeline topology optimization skeleton involves inputting the pipeline topology optimization skeleton into a CAD system and then using the geometric modeling function of the CAD kernel to generate the parameterized boundary representation model of the complex pipeline network structure.

7. A complex structure modeling system based on three-dimensional curved surface skeleton extraction, for performing a complex structure modeling method based on three-dimensional curved surface skeleton extraction according to any one of claims 1 to 6, characterized in that, include: The mesh construction module is used to acquire initial 3D point cloud data of complex pipeline network structures, and then convert the initial 3D point cloud data of pipelines into 3D voxel meshes of complex pipeline network structures. The skeleton generation module is used to generate a pipeline symbolic distance field based on the pipeline three-dimensional voxel mesh, perform local extremum traversal on the pipeline symbolic distance field to obtain a discrete skeleton candidate point set, and perform topological homology extraction based on the discrete skeleton candidate point set to generate an initial pipeline topology skeleton of a complex pipeline network structure. The skeleton optimization module is used to perform semantic topology classification using the pipeline 3D voxel mesh and the initial pipeline topology skeleton to obtain the structural attribute labels of each skeleton segment in the initial pipeline topology skeleton, and to perform local topology optimization on the initial pipeline topology skeleton based on the structural attribute labels of all skeleton segments, thereby obtaining the pipeline topology optimized skeleton of the complex pipeline network structure. The model generation module is used to generate a parameterized boundary representation model of a complex pipeline network structure from the pipeline topology optimization skeleton.

8. A computer device, comprising: The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the complex structure modeling method based on three-dimensional surface skeleton extraction as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by a processor to implement the complex structure modeling method based on three-dimensional curved surface skeleton extraction according to any one of claims 1 to 6.

Images