Efficient Modeling Method, Apparatus, and Medium for Urban Building Entities Based on Compact Convex Hull

By constructing a compact convex hull and using adaptive spatial partitioning, the problem of dependence on normal vectors in existing technologies is solved, enabling efficient modeling of urban building entities and generating concise polygonal models with geometric consistency and detail preservation capabilities.

CN123118933APending Publication Date: 2026-11-17SHENZHEN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611621338.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing reconstruction methods have limitations in handling noise points and geometric completion of large-scale missing areas. Some methods rely heavily on point cloud normal vector information, but normal vectors are difficult to obtain accurately during actual data acquisition, which limits the accuracy and speed of urban building entity modeling.

Method used

An efficient modeling method for urban building entities based on compact convex hulls is adopted. By acquiring the original point cloud, planar primitive fitting and iterative merging are performed to construct a compact convex hull. Combined with adaptive spatial partitioning and optimized surface extraction, polygonal meshes are generated, reducing the dependence on normal vectors and improving the stability and accuracy of the reconstruction process.

Benefits of technology

Without relying on high-precision normal vectors, the accuracy and efficiency of urban building entity modeling are improved, the impact of noise points and occlusion loss on reconstruction results is reduced, and a concise polygonal model with geometric consistency and detail preservation is generated.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118933A_ABST
    Figure CN123118933A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of 3D modeling and discloses an efficient method, device, and medium for modeling urban building entities based on compact convex hulls. The method includes: acquiring the original point cloud of the target building, performing planar primitive fitting and iterative merging processing to obtain a merged plane, and dividing the plane according to its function to obtain multiple functional planes; performing projection processing, fitting processing, and cutting processing on the points in the multiple functional planes to obtain target polygons, calculating the polygon area and projected concave hull area of ​​the target polygons, and constructing a compact convex hull; performing external surface filtering and segmentation processing on the compact convex hull to obtain multiple convex polyhedra; determining the remaining set of planes and polyhedra, performing energy function solving and point cloud shape matching processing to obtain matching patches, and combining them with multiple convex polyhedra into polygon meshes to obtain the final urban building entity reconstruction result. This invention can effectively improve the accuracy and efficiency of urban building entity modeling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, and in particular to an efficient method, apparatus and medium for modeling urban building entities based on compact convex hulls. Background Technology

[0002] With the rapid development of large-scale 3D data acquisition technology and the continuous advancement of smart city applications, polygon mesh reconstruction for building point clouds has become an important research direction in the fields of 3D computer vision and computer graphics. Due to factors such as mutual occlusion between buildings, surface reflection, and complex acquisition environments, the actual acquired building point clouds are usually highly incomplete and accompanied by varying degrees of noise interference. These data quality issues pose a significant challenge to the accurate reconstruction of building geometry.

[0003] Existing reconstruction methods still have certain limitations in handling noise points and geometric completion of large-scale missing areas. Some methods rely heavily on point cloud normal vector information, but normal vectors are often difficult to obtain accurately during actual acquisition, thus limiting the accuracy and speed of urban building entity modeling.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide an efficient modeling method, device, and medium for urban building entities based on compact convex hulls. This invention aims to address the limitations of existing reconstruction methods in handling noise points and geometric completion of large-scale missing areas. Some methods heavily rely on point cloud normal vector information, but normal vectors are often difficult to obtain accurately during actual acquisition, thus limiting the accuracy and speed of urban building entity modeling.

[0006] To achieve the above objectives, the present invention provides an efficient modeling method for urban building entities based on compact convex hulls, the method comprising the following steps: The original point cloud of the target building is obtained, and the original point cloud is subjected to planar primitive fitting and iterative merging processing to obtain a merged plane. The merged plane is then divided according to planar function to obtain multiple functional planes. Projection processing, fitting processing, and cutting processing are performed on points in multiple functional planes to obtain target polygons. The polygon area and projected concave hull area of ​​the target polygons are calculated, and a compact convex hull is constructed based on the polygon area and the projected concave hull area. The compact convex hull is subjected to external surface filtering and segmentation to obtain multiple convex polyhedra; The remaining set of planes and the set of polyhedra are determined. Energy functions are solved and point cloud shape matching is performed based on the remaining set of planes and the set of polyhedra to obtain matching patches. The matching patches and multiple convex polyhedra are combined into polygon meshes to obtain the final urban building entity reconstruction result.

[0007] Optionally, the efficient modeling method for urban building entities based on compact convex hulls, wherein obtaining the original point cloud of the target building, performing planar primitive fitting and iterative merging processing on the original point cloud to obtain a merged plane, and dividing the merged plane according to planar function to obtain multiple functional planes, specifically includes: Identify the target building and obtain its original point cloud. The first energy function is constructed using the FPP method, and the original point cloud is subjected to planar primitive fitting based on the first energy function to obtain an initial plane; The initial plane is iteratively merged according to the PolyFit plane fusion rules to obtain the merged plane; The merged plane is divided according to its function to obtain multiple functional planes; The functional plane includes the ground, the vertical plane, and the remaining planes.

[0008] Optionally, in the efficient modeling method for urban building entities based on compact convex hulls, the expression for the first energy function is: ; in, For the first energy function, To fit the mass-energy term, for The corresponding weights For the energy term of model simplicity, for The corresponding weights For model integrity energy term, for The corresponding weights.

[0009] Optionally, the efficient modeling method for urban building entities based on compact convex hulls, wherein the projection processing, fitting processing, and cutting processing of points in multiple functional planes to obtain the target polygon specifically includes: Projecting points in the vertical plane onto the ground surface yields projection points; The projection points are fitted to line segments, and the support lines of the line segments are calculated; The ground is cut according to the support line to obtain an initial polygon; The corresponding support lines are sorted in descending order according to the length of the line segments, and the initial polygon is divided into two subspaces by the support lines in descending order until all the support lines are used to obtain the target polygon.

[0010] Optionally, the efficient modeling method for urban building entities based on compact convex hulls, wherein calculating the polygon area and projected concave hull area of ​​the target polygon, and constructing a compact convex hull based on the polygon area and the projected concave hull area, specifically includes: All points in the original point cloud are projected onto the target polygon to obtain the projected polygon, and the projected concave area is calculated based on the points in the projected polygon. Calculate the polygon area of ​​the projected polygon, and calculate the point cloud coverage based on the polygon area and the projected hull area; Extract the polygons in the target polygon whose point cloud coverage is greater than a preset threshold to obtain a compact convex hull.

[0011] Optionally, the efficient modeling method for urban building entities based on compact convex hulls, wherein the external surface filtering and segmentation of the compact convex hull to obtain multiple convex polyhedra specifically includes: Filter the outer surfaces outside the compact convex hull to obtain the remaining surfaces, and sort the remaining surfaces in descending order according to the area of ​​the projected concave hull; Based on the sorted remaining faces, the compact convex hull is initially segmented to obtain multiple convex spaces; Each convex space is divided into two subspaces based on the remaining surfaces, resulting in multiple convex polyhedra.

[0012] Optionally, the efficient urban building entity modeling method based on compact convex hulls, wherein determining the set of remaining planes and the set of polyhedra, performing energy function solving and point cloud shape matching processing based on the set of remaining planes and the set of polyhedra to obtain matching patches, and combining the matching patches and multiple convex polyhedra into a polygon mesh to obtain the final urban building entity reconstruction result, specifically includes: The remaining planes are clustered based on planar adjacency relationships to obtain a set of remaining planes and a set of polyhedra. A second energy function is constructed based on the set of remaining planes and the set of polyhedra, and the second energy function is solved to obtain the solution result; Obtain the constituent facets of the polyhedron set, and perform point cloud shape matching on the constituent facets of the polyhedron set according to the solution results to obtain the matching facets; The final urban building entity reconstruction result is obtained by combining polygonal meshes based on the matching facets.

[0013] Optionally, in the efficient modeling method for urban building entities based on compact convex hulls, the expression for the second energy function is: ; For the second energy function, For parameters, S For each facet located at the boundary of its set of polyhedra, The facets that make up the set of polyhedra, F S for S China does not belong to A collection of noodle pieces, for F S Corresponding patch index, With | F S | represents the number of face pieces in the corresponding face piece set. and k For the corresponding patch index, and For binary variables, For the area of ​​the corresponding facet, R To select the contribution corresponding to the selected facet.

[0014] Furthermore, to achieve the above objectives, the present invention also provides an efficient modeling device for urban building entities based on compact convex hulls, wherein the efficient modeling device for urban building entities based on compact convex hulls includes: The plane partitioning module is used to acquire the original point cloud of the target building, perform plane primitive fitting and iterative merging processing on the original point cloud to obtain a merged plane, and divide the merged plane according to plane function to obtain multiple functional planes; The compact convex hull construction module is used to perform projection processing, fitting processing, and cutting processing on points in multiple functional planes to obtain target polygons, calculate the polygon area and projected concave hull area of ​​the target polygons, and construct a compact convex hull based on the polygon area and the projected concave hull area. A convex polyhedron construction module is used to perform external surface filtering and segmentation processing on the compact convex hull to obtain multiple convex polyhedra. The reconstruction result generation module is used to determine the remaining set of planes and the set of polyhedra, and to perform energy function solving and point cloud shape matching processing based on the remaining set of planes and the set of polyhedra to obtain matching patches. The matching patches and multiple convex polyhedra are then combined into polygon meshes to obtain the final reconstruction result of the urban building entity.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a compact convex hull-based efficient modeling program for urban building entities stored in the memory and executable on the processor, wherein when the compact convex hull-based efficient modeling program for urban building entities is executed by the processor, it implements the steps of the compact convex hull-based efficient modeling method for urban building entities as described above.

[0016] Furthermore, to achieve the above objectives, the present invention also provides a medium, wherein the medium stores an efficient modeling program for urban building entities based on compact convex hulls, and when the efficient modeling program for urban building entities based on compact convex hulls is executed by a processor, it implements the steps of the efficient modeling method for urban building entities based on compact convex hulls as described above.

[0017] In this invention, the original point cloud of the target building is acquired, and planar primitive fitting and iterative merging processes are performed on the original point cloud to obtain a merged plane. This merged plane is then divided according to planar function to obtain multiple functional planes. Points in these multiple functional planes are subjected to projection, fitting, and cutting processes to obtain target polygons. The polygon area and projected concave hull area of ​​the target polygons are calculated, and a compact convex hull is constructed based on these areas. The compact convex hull undergoes external surface filtering and segmentation processes to obtain multiple convex polyhedra. A set of remaining planes and a set of polyhedra are determined, and energy functions are solved and point cloud shape matching is performed based on these sets to obtain matching patches. These matching patches and the multiple convex polyhedra are then combined into a polygon mesh to obtain the final urban building entity reconstruction result. This invention provides global spatial constraints for building point cloud reconstruction by constructing a compact convex hull generated based on two-dimensional projection, limiting the reconstruction process to a reasonable range and reducing the impact of noise points, occlusion, missing elements, and fragmented structures on the reconstruction results. This effectively improves the accuracy and efficiency of urban building entity modeling. Attached Figure Description

[0018] Figure 1 This is a flowchart of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention; Figure 2 This is a schematic diagram of the overall operation process of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention; Figure 3 This is a schematic diagram of the compact convex hull generation process, which is a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls according to the present invention. Figure 4 This is a schematic diagram of the spatial partitioning process of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention. Figure 5 This is a schematic diagram of the surface extraction process of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention; Figure 6 This is a schematic diagram of the missing point cloud reconstruction result of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention. Figure 6 In r v (Indicates the proportion of points retained). Figure 7 This is a schematic diagram of the point cloud reconstruction results from the UrbanBIS dataset, representing a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls according to the present invention. Figure 8 This is a schematic diagram of the point cloud reconstruction results from the PolyFit dataset, representing a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls according to the present invention. Figure 9 This is a schematic diagram comparing the reconstruction results of different methods of the preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention. Figure 10 This is a schematic diagram of the reconstruction result visualization of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention. Figure 11 This is a schematic diagram comparing the reconstruction results of different methods in the same scene of a preferred embodiment of the efficient modeling method for urban building entities based on compact convex hulls of the present invention. Figure 12 This is a structural diagram of a preferred embodiment of the efficient modeling device for urban building entities based on compact convex hulls of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] With the rapid development of large-scale 3D data acquisition technology and the continuous advancement of smart city applications, polygon mesh reconstruction of building point clouds has become an important research direction in the fields of 3D computer vision and computer graphics. Due to factors such as mutual occlusion between buildings, surface reflection, and complex acquisition environments, the actual acquired building point clouds are usually highly incomplete and accompanied by varying degrees of noise interference. These data quality issues pose a significant challenge to the accurate restoration of building geometry. Existing reconstruction methods still have limitations in handling noise points and geometric completion of large-scale missing areas. Furthermore, some methods heavily rely on point cloud normal vector information, but normal vectors are often difficult to obtain accurately during actual acquisition, thus limiting their applicability in scenarios where normal vector input is not required. Therefore, how to effectively restore a complete, concise, and structurally consistent 3D building model using only incomplete and noisy building point clouds while preserving the main geometric structure remains a worthy research problem.

[0021] Specifically, 3D reconstruction is a major research direction in the fields of computer graphics and computer vision. This invention will introduce the technical background from the aspects of explicit mesh reconstruction, implicit mesh reconstruction, low-poly mesh reconstruction, and reconstruction based on building base contours.

[0022] 1. Explicit Mesh Reconstruction: Explicit mesh reconstruction directly establishes topological edges and faces from discrete data. Some classic methods attempt to search for point-based topological connections within local neighborhoods, including sphere rotation methods, scale-space methods, Delaunay triangulation, and voxel-based reconstruction methods. These methods are essentially heuristic reconstruction methods based on geometric information. Their main drawback is the lack of semantic priors, making it difficult to guarantee the reconstruction quality of local semantic details. To compensate for this deficiency, some methods establish topological relationships by fusing deep semantic features, including deep geometric prior methods, MeshCNN (Mesh Convolutional Neural Network), Point2Mesh (Point Cloud to Mesh Network), BSP-Net (Binary Space Partitioning Network), cube shape abstraction methods, and the Neuro-Galogen solution. These methods utilize geometric and semantic priors for mesh reconstruction. However, a major limitation of these methods is their sensitivity to the distribution of input data and noisy points. Furthermore, due to the lack of global geometric constraints, the reconstruction results are prone to problems such as face flipping and topological errors.

[0023] 2. Implicit Mesh Reconstruction: Unlike explicit mesh reconstruction, implicit mesh reconstruction solves for an implicit function and defines the zero-level set of this function as the target surface. The greatest advantage of this type of method is its ability to utilize the global constraints provided by the scalar function, thereby ensuring the manifold consistency of the output mesh. Representative methods include Poisson reconstruction and radial basis function methods. These methods offer high flexibility and can handle shapes with complex topologies. In recent years, some research has made progress in global normal orientation, further improving the normal sensitivity problem during reconstruction and enhancing practicality. However, these methods still suffer from the loss of local geometric details, resulting in overly smooth reconstructed meshes. Even improving fitting accuracy cannot fundamentally solve this problem.

[0024] 3. Low-Polygon Mesh Reconstruction: The main goal of low-polygon mesh reconstruction is to construct lightweight polygonal meshes using structured geometric units while preserving geometric features. Representative methods include block-based methods, structure-preserving methods, and planar assumption methods. These methods extract basic geometric elements from the original point cloud to construct polygonal meshes, while approximating the target object within incomplete geometric structures. However, these methods are often limited by their dependence on normals and have low computational efficiency. Some methods achieve shape reconstruction by combining geometric primitives. However, these methods are mainly geared towards CAD models and still have limitations in curvature matching. A novel strategy, WindPoly, has been proposed in the prior art, which reduces the dependence on normal vectors through global orientation optimization. Although this method improves usability to some extent, it still suffers from the problem of geometric detail loss. Furthermore, some studies employ learning-based methods to generate polygonal meshes, such as a continuous manifold representation for polygonal mesh reconstruction; a shape generation method based on a diffusion probability model; and a graph neural network based on polyhedra for reconstructing architectural models from point clouds. However, these methods often struggle to adequately preserve fine-grained geometric details when dealing with diverse architectural structures.

[0025] 4. Reconstruction Based on Building Base Profile: Building base profiles are typically used to locate walls, constrain roof and facade stretching, and provide structural priors for subsequent spatial division. Existing building reconstruction methods often require varying degrees of trade-offs between robustness, geometric regularity, semantic expressiveness, and scalability. Existing technologies propose a 2.5D dual-profile method to generate a seamless roof-wall model while maintaining sharp geometric features. Subsequently, a regularity-based modeling method further improves the geometric rationality of the reconstruction results by constraining the consistency between the roof plane and boundaries, but this method relies on accurate and reliable geometric primitives. Hybrid primitive methods further extend the application scope to buildings, vegetation, and other scenarios, but are still susceptible to profile errors and rely on certain scale assumptions. Existing technologies integrate GIS data, grid models, and street view images to generate rich urban models. However, their optimization process and the assumption of straight profiles limit the interaction efficiency and the ability to express curved surface geometries to some extent. Existing technologies construct instance-level proxy models from sparse SfM point clouds for UAV path planning, but their reconstruction results often oversimplify the roof structure. Overall, this type of method can maintain the regularity of the building structure relatively well, but it also inherits the limitations of 2.5D representation, such as the difficulty in fully expressing detailed facade geometry.

[0026] Existing 3D reconstruction methods often struggle to simultaneously achieve model compactness, geometric consistency, local detail, and reconstruction robustness. Traditional triangular mesh methods require large amounts of data, and implicit methods are prone to producing overly smoothed models. Existing structured reconstruction methods typically rely on accurate normal vectors or geometric primitives and are sensitive to noise, occlusion, and fragmented structures. Furthermore, spatial construction methods based on large-scale planar arrangements have high computational costs and may lead to structural non-closure or even algorithm non-convergence under complex or missing data conditions. To address these issues, this invention aims to propose a lightweight polygon reconstruction method for urban building point clouds. Without requiring additional high-quality geometric features, it achieves a concise polygon model that combines robustness, geometric consistency, model closure, and detail preservation through global spatial constraints, adaptive spatial partitioning, and optimized surface extraction.

[0027] To address the aforementioned problems, this invention proposes a polygon mesh reconstruction method for urban building point clouds. First, the original point cloud is mapped to a two-dimensional space, and a compact convex hull is constructed to form a 2.5D closed space constraining the reconstruction process. Then, principal planes are extracted from the point cloud, and these planes are used to adaptively partition the compact convex hull, obtaining multiple convex polyhedra representing the building's interior space. Based on this, candidate polygonal surfaces are constructed, and an optimized model considering geometric consistency, model closure, and local geometric details is established. The final polygonal mesh is obtained through surface extraction. This method reduces the dependence on geometric features such as normal vectors and improves the stability and geometric representation capability of reconstructing complex, noisy, and incomplete building point clouds while reducing the amount of model data.

[0028] The preferred embodiment of the present invention describes an efficient modeling method for urban building entities based on compact convex hulls, such as... Figure 1 and Figure 2 As shown, the efficient modeling method for urban building entities based on compact convex hulls includes the following steps: Step S10: Obtain the original point cloud of the target building, perform planar primitive fitting and iterative merging processing on the original point cloud to obtain a merged plane, and divide the merged plane according to planar function to obtain multiple functional planes.

[0029] The purpose of this invention is to address the problems of existing urban building point cloud reconstruction methods, such as strong dependence on high-quality normal vectors and geometric primitives, sensitivity to noise and missing data, and the tendency for models to become non-closed or even convergent under complex and fragmented structures. By constructing a compact convex hull and combining adaptive spatial partitioning and optimized surface extraction, the geometric consistency, closure, detail preservation ability, and overall robustness of the reconstruction results are improved while reducing the amount of model data.

[0030] Specifically, the target building is identified, and its original point cloud is obtained; a first energy function is constructed using the FPP method, and the original point cloud is subjected to planar primitive fitting processing based on the first energy function to obtain an initial plane; the initial plane is iteratively merged according to the PolyFit planar fusion rule to obtain a merged plane; the merged plane is divided according to its function to obtain multiple functional planes; wherein, the functional planes include the ground, vertical planes, and other planes.

[0031] The expression for the first energy function is: ; in, For the first energy function, To fit the mass-energy term, for The corresponding weights For the energy term of model simplicity, for The corresponding weights For model integrity energy term, for The corresponding weights.

[0032] like Figure 2 As shown, the reconstruction process mainly consists of four parts: plane detection, compact convex hull generation, spatial partitioning, and surface extraction. Figure 2 In the diagram, (a) represents the input point cloud, while (b)-(e) illustrate the four parts of the reconstruction process, namely, planar detection (i.e.,...). Figure 2 (b) in the middle), compact convex hull generation (i.e. Figure 2 (c) in the middle), spatial division (i.e. Figure 2 (d) and surface extraction (i.e. Figure 2 (e)). Given a point cloud (i.e., the original point cloud of the target building in this invention), planar primitives (such as...) are first detected from the point cloud. Figure 2 As shown in (b) in the figure, the initial structured information is obtained, and the plane is divided into three categories: ground, vertical plane and other planes.

[0033] The specific process of plane detection is as follows: To construct a lightweight polygonal mesh, planes must first be initialized from the original point cloud. The plane extraction process needs to consider geometric consistency and curvature adaptability, while also effectively handling noise and scattered point sets. This invention employs the existing plane extraction method FPP to fit plane primitives, which defines an energy function consisting of three energy terms. (i.e., the first energy function): ; in, The average distance between a plane and its interior points is used to represent the fit quality (i.e., the fit quality energy term). The simplicity of the model is measured (i.e., the model simplicity energy term). Measure the integrity of the model (i.e., the model integrity energy term). , and These are the weights corresponding to the three energy terms. The three weights are set to 1 / 3 by default, so that the three energy terms contribute equally. As the energy terms are optimized, the discrete points are clustered to form an initial plane.

[0034] While the FPP method provides a basic framework for plane detection, it generates fragmented surfaces when processing low-quality point clouds (e.g., missing points, outliers, noise), negatively impacting subsequent computations. To address this issue, this invention iteratively merges candidate planes based on the plane fusion rules in PolyFit (an algorithm for reconstructing polygonal surfaces from point clouds), avoiding fragmented planes and improving plane quality. Finally, based on the plane's function, the plane is divided into ground planes. P G Vertical plane { P V}, and the remaining faces { P R}

[0035] The specific implementation process is as follows: PRR defines the condition that a plane must satisfy to merge, namely, when the included angle between the two planes is less than 10°. And the number of common points between the two planes is greater than Planes can be merged, as shown below: ; ; in and For plane and The normal vector, and Representing a plane and The corresponding number of points , The threshold value is [value], and the default value is [value]. .

[0036] Finally, the planes are categorized based on their function. First, the bottom surface of the point cloud bounding box is defined as the initial ground. Then, according to the PRR rule, planes that can be merged with the ground are combined to form the final ground. P G A plane perpendicular to the ground is defined as a vertical plane. P V The remaining plane is defined as { P R The final plane is divided into ground. P G Vertical plane { P V} and the remaining faces { P R Three categories.

[0037] Step S20: Perform projection processing, fitting processing, and cutting processing on points in multiple functional planes to obtain target polygons, calculate the polygon area and projected concave hull area of ​​the target polygons, and construct a compact convex hull based on the polygon area and the projected concave hull area.

[0038] Specifically, points in the vertical plane are projected onto the ground to obtain projection points; the projection points are fitted into line segments, and the support lines of the line segments are calculated; the ground is cut according to the support lines to obtain an initial polygon; the corresponding support lines are sorted in descending order according to the length of the line segments, and the initial polygon is divided into two subspaces in descending order using the support lines until all the support lines are used to obtain the target polygon.

[0039] All points in the original point cloud are projected onto the target polygon to obtain the projected polygon, and the area of ​​the projected concave hull is calculated based on the points in the projected polygon; the polygon area of ​​the projected polygon is calculated, and the point cloud coverage is calculated based on the polygon area and the area of ​​the projected concave hull; polygons in the target polygon whose point cloud coverage is greater than a preset threshold are extracted to obtain a compact convex hull.

[0040] like Figure 2 (c) and Figure 3 As shown, this invention generates a compact convex hull based on a vertical plane. The process of generating the compact convex hull is as follows: A compact convex hull is essentially a 2.5D structure representing the contour of a point cloud. Constructing a compact convex hull can be divided into two steps: first, constructing a 2D bounding box based on the ground plane, and then using 2D projection points to optimize the bounding box.

[0041] For the first step, first identify the points that make up the vertical plane (such as...). Figure 3 As shown in (a) in the figure, the projection onto the ground (as shown in the figure) Figure 3 As shown in (b)), these points (i.e., projection points) are fitted into line segments (e.g., ...). Figure 3 As shown in (c)), the support lines are calculated (based on the principle that two points determine a straight line, for each line segment, its two endpoints can determine a straight line, which is the support line). Then, the support lines are used to cut the ground to generate an initial polygon, representing the initial 2D space. To avoid redundant calculations, this invention sorts the corresponding vertical planes and support lines in descending order according to the length of the line segments, and uses the support lines to divide the 2D space in which it is located into two subspaces in turn. Before the next division, it is determined which subspace the next support line is located in, until all support lines have been traversed, the process ends, and finally a set of polygons (i.e., the target polygons) is obtained, such as... Figure 3 As shown in (d) in the figure.

[0042] The second step involves projecting all points in the point cloud (i.e., the original point cloud) onto a polygon, such as... Figure 3 As shown in (e) in the diagram. For each polygon B i The alpha shape (i.e., the projected concave hull) is calculated using the CGAL library based on the points located inside the polygon, and then defined. A ( B i () is a polygon B i The area, and A ( B i ) α for B i The alpha shape area (i.e., the projected concave hull area). A ( B i ) α / A ( B i This represents the coverage of the point cloud. The larger the value, the closer the shape of the point cloud is to the shape of the polygon. Searching for a polygon that satisfies this property... A ( B i ) α / A ( B i > T r polygons (such as) Figure 3 (as shown in (f) in the figure), where T r The threshold (i.e., the preset threshold) is used as the threshold. Finally, the found polygons are extruded to generate a 2.5D shape, which is the compact convex hull.

[0043] Step S30: Perform external surface filtering and segmentation on the compact convex hull to obtain multiple convex polyhedra.

[0044] Specifically, the outer surfaces outside the compact convex hull are filtered to obtain the remaining surfaces, and the remaining surfaces are sorted in descending order according to the projected concave hull area; the sorted remaining surfaces are used to perform initial segmentation processing on the compact convex hull to obtain multiple convex spaces; each convex space is divided into two subspaces according to the remaining surfaces to obtain multiple convex polyhedra.

[0045] like Figure 2 (d) and Figure 4 As shown, the remaining faces are used to partition the space around the convex hull, resulting in several convex polyhedra. The specific implementation process of this spatial partitioning is as follows: The spatial partitioning is based on the intersection and division of the remaining planes. Figure 4 (a) shows the positions of the polyhedron forming the compact convex hull and the remaining planes. Before segmentation begins, the outer faces are identified to perform initial segmentation of the polyhedron. The detection of the outer faces is based on the following rule: for planes... P If all other planes are located on the same side of this plane, then the plane... P For the external surface, such as Figure 4 As shown in (b), the model contains two external faces. During the initial segmentation, the external faces divide all polyhedra in two, with each polyhedron located on one side of the external face. Polyhedra carrying geometric information are concentrated on one side of the external face, while polyhedra on the other side are removed to improve computational efficiency, such as... Figure 4 As shown in (c) in the figure.

[0046] After the above division, each polyhedron can be considered as a convex space. Finally, traverse the remaining planes to determine the convex space in which they reside (e.g., ...). Figure 4 As shown in (d) in the figure, this plane divides the convex space into two subspaces (as shown in the figure). Figure 4 As shown in (e) in the diagram), the loop continues until all planes have been traversed, ending the loop. Figure 4 As shown in (f), the original convex hull is further divided into several convex polyhedra.

[0047] Step S40: Determine the remaining set of planes and the set of polyhedra. Solve the energy function and perform point cloud shape matching processing based on the remaining set of planes and the set of polyhedra to obtain matching patches. Combine the matching patches and multiple convex polyhedra into a polygon mesh to obtain the final urban building entity reconstruction result.

[0048] Specifically, the remaining planes are clustered based on planar adjacency relationships to obtain a set of remaining planes and a set of polyhedra; a second energy function is constructed based on the set of remaining planes and the set of polyhedra, and the second energy function is solved to obtain the solution result; the constituent facets of the polyhedron set are obtained, and point cloud shape matching is performed on the constituent facets of the polyhedron set based on the solution result to obtain matching facets; polygon meshes are combined based on the matching facets to obtain the final urban building entity reconstruction result. The expression for the second energy function is: ; in, For the second energy function, For parameters, S For each facet located at the boundary of its set of polyhedra, The facets that make up the set of polyhedra, F S for S China does not belong to A collection of noodle pieces, for F S Corresponding patch index, With | F S | represents the number of face pieces in the corresponding face piece set. and k For the corresponding patch index, and For binary variables, For the area of ​​the corresponding facet, R To select the contribution corresponding to the selected facet.

[0049] like Figure 2 (e) and Figure 5 As shown, finally, an energy function is constructed based on the facet information in the polyhedron, the energy function is solved, and the polygon mesh is extracted. The specific process of surface extraction is as follows: Surface extraction aims to extract polygonal meshes from convex polyhedra obtained through spatial partitioning. This invention designs an optimization method to solve this problem. First, the set of facets constituting the polyhedron is defined as { F}, where the facet located at the boundary is defined as S This represents the outer surface structure. Subsequently, the remaining planes are clustered based on planar adjacency relationships, with adjacent planes grouped into the same class. Each class of planes is surrounded by a set of convex hulls, and the polyhedra inside the convex hulls are defined as { C in Based on the remaining plane set { P R} and polyhedron set { C in Define the energy function. (That is, the second energy function in this invention), the expression is as follows:

[0050]

[0051] in, max This indicates the goal of maximizing energy, where 'st' represents the solution. x The conditions that need to be met.

[0052] This invention will form a polyhedron { C in} face definition , S China does not belong to The surface is defined as { F S In the above formula With | FS | then represent the set of facets respectively. and{ F S The number of pieces in}, | E | indicates the number of edges. l , k , i These represent the indices of the faces or edges of the corresponding sets, respectively. By solving the energy function, we can... E U To maximize, three conditions must be met simultaneously, from Find the facet that matches the shape of the point cloud and match it with { F S These are combined to form a polygonal mesh. The first two conditions... and express This is a binary variable used to indicate whether a patch is selected; This is a parameter, with a default value of 1. The third condition. As a manifold constraint, the manifold nature of the reconstruction result is guaranteed by controlling the number of faces connected by each edge to 0 or 2. , This indicates the contribution provided by the selected slice. A ( F ) represents the total area of ​​the face. The indicator function is defined as follows:

[0053] in, This represents the area of ​​the corresponding facet. This represents the alpha shape area of ​​the corresponding facet. For the first i A piece of dough. Figure 5 An example of surface extraction involving two iterations is shown. Figure 5 In Figure 5, (a) represents the shape of the model before the solution is obtained. Figures (b) and (c) show the first solution process, with (c) being a two-dimensional schematic diagram of (b). Then { F S}and Swap the components and perform a second solution, such as... Figure 5 As shown in (d) and (e), (e) is a two-dimensional schematic diagram of (d). The final result of the urban building reconstruction is as follows. Figure 5 (f) and Figure 8 As shown.

[0054] The beneficial effects of this invention are: This invention effectively constrains the reconstruction space by constructing a compact convex hull, and combines adaptive space partitioning and optimized surface extraction. This improves the adaptability to noise, occlusion and missing data without relying on additional geometric features such as high-precision normal vectors, thereby enhancing the stability and robustness of the reconstruction process.

[0055] Key aspects of this invention include: 1. By constructing a tight convex hull that matches the main building structure, effective constraints are achieved on the reconstruction space, providing a stable boundary for subsequent structural restoration.

[0056] 2. Utilize building structure information to establish hierarchical spatial division results, reduce complex geometric search processes, and improve structural analysis capabilities.

[0057] 3. By jointly optimizing geometric consistency, manifold constraints, and local features, lightweight and high-quality architectural polygon model generation is achieved.

[0058] Compared with the prior art, the present invention has the following advantages: 1. This invention provides global spatial constraints for building point cloud reconstruction by constructing a compact convex hull based on two-dimensional projection, which limits the reconstruction process to a reasonable range, reduces the impact of noise points, occlusion, missing parts, and fragmented structures on the reconstruction results, and reduces the problems of model non-closure and algorithm non-convergence in traditional methods.

[0059] 2. This invention combines adaptive spatial partitioning and optimized surface extraction methods to construct polygonal models using only structural information in point clouds. It does not rely on accurate normal vectors or additional high-quality geometric primitives. Even under conditions of complex building point clouds, noisy input, and incomplete scan data, it can still generate structured models with good geometric consistency.

[0060] 3. This invention avoids a large number of invalid intersection calculations between planes by using compact convex hull constraints and convex polyhedral space partitioning, and extracts a small number of high-quality polygonal surfaces through surface optimization. While maintaining the main structure, sharp features and local details of the building, it effectively reduces the amount of model data and improves the visualization and analysis efficiency of large-scale urban 3D scenes.

[0061] The experimental results of this invention are as follows: The point cloud data used for testing in this invention comes from the BuildingNet dataset, the UrbanBIS dataset, and the PolyFit dataset. BuildingNet contains point clouds of buildings in various styles; UrbanBIS is a database of urban scenes containing real buildings; and PolyFit contains point clouds of buildings collected from real scenes by various sensing devices. The method of this invention is evaluated based on these datasets to verify its robustness. Quantitative evaluation metrics are designed from two main aspects: geometric consistency and data simplification rate. Regarding geometric consistency, this invention uses Hausdorff distance (Dis... H ), mean distance (Dis) M ) and Chebyshev distance (Dis C The evaluation is performed by calculating the simplification rate R. Avg. The simplicity of the reconstruction results is measured, and the accuracy of the reconstruction results is evaluated using the Intersection over Union (IoU) ratio.

[0062] The method of this invention (i.e., Ours in Table 1) is compared with PolyFit (polygon fitting algorithm), KSR (ablation experiment), LowPoly (low-polygon reconstruction algorithm), R-LowPoly (improved low-polygon reconstruction algorithm), WindPoly (wrap-around polygon reconstruction algorithm), COMPOD, and other methods. The experimental results are as follows: Figure 9 As shown. KSR and COMPOD are reconstruction methods that rely on normal vectors. LowPoly and R-LowPoly require an initial mesh as input. To meet their input requirements, this invention needs to obtain an initial mesh. From Figure 9 It can be seen that PolyFit cannot guarantee that the generated model has a clear and well-defined basic geometry. KSR is susceptible to erroneous normal vectors and produces redundant patches. LowPoly and R-LowPoly struggle to obtain flat, structured planes. COMPOD relies on accurate normal estimation but lacks manifold constraints. In contrast, the method of this invention can effectively capture more refined internal geometry. Figure 6 In section 8, the present invention demonstrates the reconstruction results of the method on different data samples, including architectural models with missing regions and complex structures. Figure 10 This demonstrates the reconstruction results of buildings in more different styles, verifying the effectiveness of the invention in practical applications.

[0063] This invention has also been validated on large-scale urban point cloud data containing multiple building instances, and the reconstruction results of different methods have been compared, such as... Figure 11 As shown, the corresponding quantitative evaluation indicators are shown in Table 1 (Dis H Dis represents Hausdorff distance. M Dis represents the average distance. C R represents the chamfer distance. Avg (IoU represents the average regularity error).

[0064] Table 1: Quantitative evaluation indicators for different methods

[0065] It can be seen that PolyFit's reconstruction results are often oversimplified, resulting in the loss of a large amount of geometric details; KSR relies heavily on accurate normal information and struggles to obtain complete reconstruction results in the absence of normal vectors; LowPoly and R-LowPoly both suffer from severe patch fragmentation problems; WindPoly is prone to missing regions near the bottom of the model, while COMPOD generates fragmented structural blocks. In contrast, the method of this invention can stably generate more complete and structurally consistent reconstruction results, outperforming other methods overall. Furthermore, this method can effectively resist random noise introduced by environmental factors during data acquisition, further demonstrating its value in practical applications.

[0066] In addition, the plane detection part in this invention can be replaced by other similar solutions. For example, the region growing method can be used to fit the plane by clustering points.

[0067] Furthermore, such as Figure 12 As shown, based on the above-mentioned efficient modeling method for urban building entities based on compact convex hulls, the present invention also provides an efficient modeling device for urban building entities based on compact convex hulls, wherein the efficient modeling device for urban building entities based on compact convex hulls includes: The plane partitioning module 51 is used to acquire the original point cloud of the target building, perform plane primitive fitting and iterative merging processing on the original point cloud to obtain a merged plane, and divide the merged plane according to plane function to obtain multiple functional planes. The compact convex hull construction module 52 is used to perform projection processing, fitting processing and cutting processing on points in multiple functional planes to obtain target polygons, calculate the polygon area and projected concave hull area of ​​the target polygons, and construct a compact convex hull based on the polygon area and the projected concave hull area. The convex polyhedron construction module 53 is used to perform external surface filtering and segmentation processing on the compact convex hull to obtain multiple convex polyhedra. The reconstruction result generation module 54 is used to determine the remaining set of planes and the set of polyhedra, perform energy function solving and point cloud shape matching processing based on the remaining set of planes and the set of polyhedra to obtain matching patches, and combine the matching patches and multiple convex polyhedra into polygon meshes to obtain the final urban building entity reconstruction result.

[0068] The present invention also provides a medium, wherein the medium stores an efficient modeling program for urban building entities based on compact convex hulls, and the efficient modeling program for urban building entities based on compact convex hulls, when executed by a processor, implements the steps of the efficient modeling method for urban building entities based on compact convex hulls as described above.

[0069] In summary, this invention provides an efficient modeling method, apparatus, and storage medium for urban building entities based on compact convex hulls. The method includes: acquiring the original point cloud of the target building; performing planar primitive fitting and iterative merging processing on the original point cloud to obtain a merged plane; dividing the merged plane according to planar functions to obtain multiple functional planes; performing projection processing, fitting processing, and cutting processing on points in the multiple functional planes to obtain target polygons; calculating the polygon area and projected concave hull area of ​​the target polygons; constructing a compact convex hull based on the polygon area and the projected concave hull area; performing external surface filtering and segmentation processing on the compact convex hull to obtain multiple convex polyhedra; determining the remaining set of planes and the set of polyhedra; performing energy function solving and point cloud shape matching processing based on the remaining set of planes and the set of polyhedra to obtain matching patches; and combining the matching patches and the multiple convex polyhedra into a polygon mesh to obtain the final urban building entity reconstruction result. This invention provides global spatial constraints for building point cloud reconstruction by constructing a compact convex hull based on two-dimensional projection, which limits the reconstruction process to a reasonable range and reduces the impact of noise points, occlusion, missing parts, and fragmented structures on the reconstruction results. This can effectively improve the accuracy and efficiency of urban building entity modeling.

[0070] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0071] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable medium, and when executed, it can include the processes described in the above method embodiments. The medium can be a memory, magnetic disk, optical disk, etc.

[0072] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A compact convex hull-based efficient modeling method for urban building entities, characterized in that, The efficient modeling method for urban building entities based on compact convex hulls includes: The original point cloud of the target building is obtained, and the original point cloud is subjected to planar primitive fitting and iterative merging processing to obtain a merged plane. The merged plane is then divided according to planar function to obtain multiple functional planes. Projection processing, fitting processing, and cutting processing are performed on points in multiple functional planes to obtain target polygons. The polygon area and projected concave hull area of ​​the target polygons are calculated, and a compact convex hull is constructed based on the polygon area and the projected concave hull area. The compact convex hull is subjected to external surface filtering and segmentation to obtain multiple convex polyhedra; The remaining set of planes and the set of polyhedra are determined. Energy functions are solved and point cloud shape matching is performed based on the remaining set of planes and the set of polyhedra to obtain matching patches. The matching patches and multiple convex polyhedra are combined into polygon meshes to obtain the final urban building entity reconstruction result.

2. The compact convex hull based urban building entity efficient modeling method according to claim 1, wherein, The process involves acquiring the original point cloud of the target building, performing planar primitive fitting and iterative merging on the original point cloud to obtain a merged plane, and then dividing the merged plane according to its function to obtain multiple functional planes, specifically including: Identify the target building and obtain its original point cloud. The first energy function is constructed using the FPP method, and the original point cloud is subjected to planar primitive fitting based on the first energy function to obtain an initial plane; The initial plane is iteratively merged according to the PolyFit plane fusion rules to obtain the merged plane; The merged plane is divided according to its function to obtain multiple functional planes; The functional plane includes the ground, the vertical plane, and the remaining planes.

3. The compact convex hull based urban building entity efficient modeling method according to claim 2, characterized in that, The expression for the first energy function is: ; wherein, is a first energy function, is a fitting quality energy term, is is a corresponding weight, is a model simplicity energy term, is is a corresponding weight, is a model completeness energy term, is is a corresponding weight.

4. The compact convex hull based urban building entity efficient modeling method according to claim 2, characterized in that, The process of projecting, fitting, and cutting points in multiple functional planes to obtain a target polygon specifically includes: Projecting points in the vertical plane onto the ground surface yields projection points; The projection points are fitted to line segments, and the support lines of the line segments are calculated; The ground is cut according to the support line to obtain an initial polygon; The corresponding support lines are sorted in descending order according to the length of the line segments, and the initial polygon is divided into two subspaces by the support lines in descending order until all the support lines are used to obtain the target polygon.

5. The compact convex hull based urban building entity efficient modeling method according to claim 1, wherein, The calculation of the polygon area and projected concave hull area of ​​the target polygon, and the construction of a compact convex hull based on the polygon area and the projected concave hull area, specifically includes: All points in the original point cloud are projected onto the target polygon to obtain the projected polygon, and the projected concave area is calculated based on the points in the projected polygon. Calculate the polygon area of ​​the projected polygon, and calculate the point cloud coverage based on the polygon area and the projected hull area; Extract the polygons in the target polygon whose point cloud coverage is greater than a preset threshold to obtain a compact convex hull.

6. The compact convex hull based urban building entity efficient modeling method according to claim 5, characterized in that, The external surface filtering and segmentation of the compact convex hull to obtain multiple convex polyhedra specifically includes: Filter the outer surfaces outside the compact convex hull to obtain the remaining surfaces, and sort the remaining surfaces in descending order according to the area of ​​the projected concave hull; Based on the sorted remaining faces, the compact convex hull is initially segmented to obtain multiple convex spaces; Each convex space is divided into two subspaces based on the remaining surfaces, resulting in multiple convex polyhedra.

7. The compact convex hull based urban building entity efficient modeling method according to claim 6, characterized in that, The process of determining the remaining set of planes and the set of polyhedra, solving for the energy function and performing point cloud shape matching based on the remaining set of planes and the set of polyhedra to obtain matching patches, and combining the matching patches and multiple convex polyhedra into a polygon mesh to obtain the final urban building entity reconstruction result, specifically includes: The remaining planes are clustered based on planar adjacency relationships to obtain a set of remaining planes and a set of polyhedra. A second energy function is constructed based on the set of remaining planes and the set of polyhedra, and the second energy function is solved to obtain the solution result; Obtain the constituent facets of the polyhedron set, and perform point cloud shape matching on the constituent facets of the polyhedron set according to the solution results to obtain the matching facets; The final urban building entity reconstruction result is obtained by combining polygonal meshes based on the matching facets.

8. The compact convex hull based urban building entity efficient modeling method according to claim 7, characterized in that, The expression for the second energy function is: ; wherein, is a second energy function, is a parameter, is a face of the polyhedral set located at the boundary, is a face of the polyhedral set, F S is S a set of faces not belonging to , is F S a corresponding face index, and F S are the number of faces of the corresponding face set, and k is a corresponding face index, and is a binary variable, is the area of the corresponding face, is the selected face corresponding contribution.

9. A compact convex hull based urban building entity efficient modeling device, characterized in that, The efficient urban building entity modeling device based on compact convex hull is used to implement the efficient urban building entity modeling method based on compact convex hull as described in any one of claims 1-8, wherein the efficient urban building entity modeling device based on compact convex hull includes: The plane partitioning module is used to acquire the original point cloud of the target building, perform plane primitive fitting and iterative merging processing on the original point cloud to obtain a merged plane, and divide the merged plane according to plane function to obtain multiple functional planes; The compact convex hull construction module is used to perform projection processing, fitting processing, and cutting processing on points in multiple functional planes to obtain target polygons, calculate the polygon area and projected concave hull area of ​​the target polygons, and construct a compact convex hull based on the polygon area and the projected concave hull area. A convex polyhedron construction module is used to perform external surface filtering and segmentation processing on the compact convex hull to obtain multiple convex polyhedra. The reconstruction result generation module is used to determine the remaining set of planes and the set of polyhedra, and to perform energy function solving and point cloud shape matching processing based on the remaining set of planes and the set of polyhedra to obtain matching patches. The matching patches and multiple convex polyhedra are then combined into polygon meshes to obtain the final reconstruction result of the urban building entity.

10. A medium characterized by, The medium stores an efficient modeling program for urban building entities based on compact convex hulls. When the efficient modeling program for urban building entities based on compact convex hulls is executed by a processor, it implements the steps of the efficient modeling method for urban building entities based on compact convex hulls as described in any one of claims 1-8.