A method and system for repairing holes in low-density, large-scale point cloud data
By dynamically adjusting the interpolation density and analyzing local geometric features, the problem of poor hole repair in low-density, large-area point cloud data was solved, achieving efficient and accurate point cloud data repair, which is suitable for topographic mapping and cultural heritage digitization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG ZHENGYUAN AVIATION REMOTE SENSING TECH LTD
- Filing Date
- 2026-04-30
- Publication Date
- 2026-06-26
AI Technical Summary
Existing point cloud data hole repair methods are ineffective and inefficient when dealing with low-density, large-area point cloud data, and cannot meet practical needs.
By dynamically adjusting the interpolation density and combining it with local geometric feature analysis, the final point cloud data is generated by generating an irregular triangular network and elevation interpolation.
It effectively balances computational efficiency and restoration accuracy, avoids resource waste and loss of local details, and improves the integrity and accuracy of point cloud data, making it suitable for complex scenarios such as topographic mapping and cultural heritage digitization.
Smart Images
Figure CN122289079A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of point cloud restoration technology, and more specifically to a method and system for restoring holes in low-density, large-area point cloud data. Background Technology
[0002] In the process of point cloud data acquisition and application, low-density areas and holes often appear due to various factors (such as equipment limitations and environmental occlusion), affecting the integrity and accuracy of point cloud data, and consequently adversely impacting subsequent analysis, modeling, and other applications based on point cloud data. Existing point cloud data hole repair methods suffer from poor repair effects and low efficiency when processing low-density, large-area point cloud data, and cannot adequately meet practical needs.
[0003] Therefore, in view of the shortcomings of the existing technology, how to provide a method and system for repairing holes in low-density, large-area point cloud data is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0004] In view of this, the present invention provides a method and system for repairing holes in low-density, large-scale point cloud data. For large-scale, non-uniformly distributed point cloud data, the algorithm effectively balances computational efficiency and repair accuracy by dynamically adjusting the interpolation density, avoiding the resource waste or loss of local details caused by global uniform encryption in traditional methods.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for repairing holes in low-density, large-area point cloud data, comprising: acquiring original point cloud data; The original point cloud data is preprocessed to generate an original point cloud mesh; Define the boundary line of the point cloud hole region, and perform spatial calculations based on the boundary line and the original point cloud mesh to obtain the original point cloud set; A first planar simulated point set is generated within the specified range line according to the original point cloud density; The original point cloud and the first plane simulation point set are filtered and clustered to obtain the second plane simulation point set; The second set of simulated points in the plane is perturbed to generate a third set of simulated points in the plane; Based on the original point set, terrain is constructed to form an irregular triangular network; Traverse the third plane simulation point set, perform elevation interpolation based on the irregular triangular network, and generate the first simulation point cloud set; The elevation of the first simulated point cluster is disturbed to form a second simulated point cluster. The second simulated point cloud set is fused with the original point cloud set to output the final point cloud data result within the hole range.
[0006] Preferably, the raw point cloud data is preprocessed, including: The original point cloud data is divided into N×N intervals according to an empirical step size to form an original point cloud mesh W1, and a quadtree spatial index is established.
[0007] Preferably, spatial operations are performed based on the range line and the original point cloud mesh to obtain the original point cloud set, including: Spatial overlay analysis is performed on the range line and the original point cloud mesh to obtain the point cloud data contained when the range line is projected onto the original point cloud mesh, as well as the point cloud data within the range line with index, and to construct the original point cloud set.
[0008] Preferably, the original point cloud and the first planar simulated point set are filtered and clustered to obtain the second planar simulated point set, including: The original point cloud and the first plane simulation point set are preprocessed and fused to obtain a fused point set; The fusion point set is clustered using a density clustering algorithm or a region growing clustering algorithm to obtain clusters; The clusters are filtered to obtain candidate clusters; The candidate clusters are fitted to the RANSAC plane to generate a second plane simulation point set.
[0009] Preferably, the set of fusion points is clustered using a density clustering algorithm to obtain clusters, including: For the ε-neighborhood of each point in the fusion point set, mark the core point, boundary point, and noise point; Clusters are generated by merging core points with achievable density; Output the clustering results.
[0010] Preferably, a region growing clustering algorithm is used to cluster the fusion point set to obtain clusters, including: Starting from the seed point in the fusion point set, calculate the threshold of the angle between the normal vectors of neighboring points; Iteratively expand the points that satisfy the conditions until no further growth is possible, forming a planar cluster; Clusters with high overlap with the fusion point set in the planar clusters are selected as clustering clusters.
[0011] Preferably, terrain is constructed based on the original point set to form an irregular triangular network, including: Create a super triangle (a sufficiently large virtual triangle) containing all points as the initial state for the algorithm; The points in the original point cloud are inserted in spatial order. For each new point inserted, an existing triangle containing that point is found and split into three sub-triangles. Check if adjacent triangles satisfy the Delaunay condition, and perform local optimization by edge flipping; If terrain feature lines exist, these edges are retained as fixed edges of the triangulation, and the topology of the surrounding triangles is adjusted.
[0012] Preferably, a system for repairing holes in low-density, large-area point cloud data includes: The data acquisition module is used to acquire raw point cloud data; The preprocessing module is used to preprocess the raw point cloud data to generate a raw point cloud mesh. The spatial calculation module is used to set the range line of the point cloud hole region, and perform spatial calculations based on the range line and the original point cloud mesh to obtain the original point cloud set. The generation module is used to generate a first planar simulated point set within the range line according to the original point cloud density; The clustering analysis module is used to filter and cluster the original point cloud and the first plane simulation point set to obtain the second plane simulation point set; The planar perturbation module is used to perturb the second planar simulation point set to generate a third planar simulation point set; The terrain construction module is used to construct terrain based on the original point set, forming an irregular triangular network. The elevation interpolation module is used to traverse the third plane simulation point set, perform elevation interpolation based on the irregular triangular network, and generate the first simulation point cloud set. The elevation disturbance module is used to perform elevation disturbance on the first simulated point set to form a second simulated point set; The data fusion module is used to fuse the second simulated point cloud set with the original point cloud set and output the final point cloud data result within the hole range.
[0013] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for repairing holes in low-density, large-area point cloud data. The present invention has the following beneficial effects: (1) Efficiency and adaptability: For point cloud data with a large range and non-uniform distribution, the algorithm effectively balances computational efficiency and repair accuracy by dynamically adjusting the interpolation density, avoiding the waste of resources or loss of local details caused by global uniform encryption in traditional methods.
[0014] (2) Maintaining structural integrity: Combining local geometric feature analysis (such as curvature and normal vector consistency), key topological structures are restored first when filling holes, reducing human intervention and outperforming conventional methods that rely on fixed templates or boundary extensions.
[0015] (3) Application scalability: It is suitable for complex scenarios (such as topographic mapping and cultural heritage digitization). Its modular design makes it easy to integrate into existing point cloud processing workflows, reducing hardware dependence and making it particularly suitable for edge device deployment. It can improve the efficiency of existing point cloud data processing, especially when producing elevation data models, it can improve the precision of local elevation data and reduce the additional costs brought by data repair and measurement.
[0016] Overall, this method breaks through the limitations of traditional methods in terms of accuracy, efficiency, and generalization ability, providing a better solution for sparse point cloud repair. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of a method for repairing holes in low-density, large-area point cloud data provided by the present invention. Detailed Implementation
[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] This invention discloses a method for repairing holes in low-density, large-area point cloud data, such as... Figure 1 As shown, this includes: acquiring raw point cloud data; The original point cloud data is preprocessed to generate an original point cloud mesh; Define the boundary line of the point cloud hole region, and perform spatial calculations based on the boundary line and the original point cloud mesh to obtain the original point cloud set; A first planar simulated point set is generated within the specified range line according to the original point cloud density; The original point cloud and the first plane simulation point set are filtered and clustered to obtain the second plane simulation point set; The second set of simulated points in the plane is perturbed to eliminate regular arrangement and enhance natural distribution characteristics, thereby generating a third set of simulated points in the plane, P'. Based on the original point set, terrain is constructed to form an irregular triangular network; Traverse the third plane simulation point set P', perform elevation interpolation based on the irregular triangular network, and generate the first simulation point cloud set P1; The first simulated point set P1 is subjected to elevation perturbation to eliminate regular arrangement and enhance natural distribution characteristics, thereby forming a second simulated point set P. The second simulated point cloud set P is fused with the original point cloud set to output the final point cloud data result within the hole range.
[0021] Specifically, the raw point cloud data is preprocessed, including: The original point cloud data is divided into N×N intervals according to an empirical step size to form an original point cloud mesh W1, and a quadtree spatial index is established to facilitate faster data processing in the later stages. Specifically, spatial operations are performed based on the range line and the original point cloud mesh to obtain the original point cloud set, including: Spatial overlay analysis is performed on the range line and the original point cloud mesh to obtain the point cloud data contained when the range line is projected onto the original point cloud mesh, as well as the point cloud data within the range line with index, and to construct the original point cloud set.
[0022] Specifically, a boundary line F for the point cloud hole region is defined, and spatial operations are performed with the original point cloud grid W1 to obtain the original point cloud set T1={t1, t2, t3, t4, ...} within the boundary line with an index. The specific method for obtaining the boundary line is to directly obtain the boundary line F in the constructed 3D spatial scene of the point cloud by interacting with the mouse and software. The boundary line F is then spatially overlaid with the original point cloud grid W1 to obtain the point cloud data contained in the point cloud grid data projected on the boundary line, as well as the original point cloud set T1 within the boundary line with an index.
[0023] Specifically, generating a first planar simulated point set within the specified range line according to the original point cloud density includes: Within the defined point cloud data aperture range F, generate a plane simulation point set P'1={P'1} according to the original point cloud density. 11 , P' 12 , P' 13 , P' 14 ,........}.
[0024] Specifically, the original point cloud and the first plane simulation point set are filtered and clustered to obtain the second plane simulation point set, including: The original point cloud and the first plane simulation point set are preprocessed and fused to obtain a fused point set; The fusion point set is clustered using a density clustering algorithm or a region growing clustering algorithm to obtain clusters; The clusters are filtered to obtain candidate clusters; The candidate clusters are fitted to the RANSAC plane to generate a second plane simulation point set.
[0025] Specifically, a density-based clustering algorithm is used to cluster the fusion point set to obtain clusters, including: For the ε-neighborhood of each point in the fusion point set, mark the core point, boundary point, and noise point; Clusters are generated by merging core points with achievable density; Output the clustering results.
[0026] Specifically, the fusion point set is clustered using a region growing clustering algorithm to obtain clusters, including: Starting from the seed point in the fusion point set, calculate the threshold of the angle between the normal vectors of neighboring points; Iteratively expand the points that satisfy the conditions until no further growth is possible, forming a planar cluster; Clusters with high overlap with the fusion point set in the planar clusters are selected as clustering clusters.
[0027] In a specific embodiment of the present invention, the original point cloud T1 within the range line and the first plane simulation point set P'1 are filtered and clustered to obtain the second plane simulation point set P'2={P' 21 , P' 22 , P' 23 , P' 24 ,........},include: The following is the detailed procedure for performing cluster analysis by combining the original point cloud T1 within the range line with the simulated point set P'1 of the first plane: 1. Data Preprocessing and Fusion Point cloud alignment: The simulated point set P'1 of the first plane is registered with the original point cloud set T1 using the ICP (Iterative Closest Point) algorithm to ensure that the coordinate systems are consistent.
[0028] Feature enhancement: Attributes (such as normal vector, curvature, and elevation) are added to each point to facilitate subsequent clustering in distinguishing between planar and non-planar regions.
[0029] 2. Selection of Cluster Analysis Method
[0030] (1) Density-based clustering (such as DBSCAN)
[0031] Principle: Clusters are defined by neighborhood radius (ε) and minimum number of points (MinPts), which is suitable for non-uniformly distributed point clouds.
[0032] step: Search the ε-neighborhood of each point in the fusion point set T1∪P'1 and label the core point, boundary point and noise point.
[0033] Core points with achievable density are merged to form clusters, while noise points (such as isolated points) are directly eliminated.
[0034] Output clustering results C={C1,C2,...,C n Each cluster may correspond to a plane or a surface.
[0035] (2) Region growing clustering (for planar features)
[0036] Principle: Merge adjacent points based on normal vector consistency or elevation continuity.
[0037] step: Starting from the seed point (such as the point in P'1), calculate the threshold of the angle between the normal vectors of neighboring points (such as <10°).
[0038] Iteratively expand the points that meet the conditions until they can no longer grow, forming a planar cluster C_plane.
[0039] Select clusters in C_plane that have a high degree of overlap with P'1 as candidate sets for P'2.
[0040] 3. Generation of the planar simulation point set P'2
[0041] Cluster selection criteria: Geometric consistency: The elevation variance (σ²) of points within a cluster is below a threshold, or the residual of the fitted plane is small.
[0042] Coverage: The proportion of clusters within the projection area of P'1 exceeds a preset value (e.g., 80%).
[0043] Optimize output: Perform RANSAC plane fitting on the candidate clusters to remove outliers.
[0044] Finally, P'2={P' 21 ,P' 22 ,...}, retain the high-confidence plane points and supplement the interpolation points in the missing regions of T1.
[0045] Specifically, terrain is constructed based on the original point set to form an irregular triangular network, including: Based on the original point cluster T1 within the boundary line, terrain is constructed to form an irregular triangular network (TIN). The specific process is as follows: Create a super triangle (a sufficiently large virtual triangle) containing all points as the initial state for the algorithm; Insert the points in the original point set T1 in spatial order. For each new point inserted, find the existing triangle containing that point and split it into 3 sub-triangles. Check if adjacent triangles satisfy the Delaunay condition, and perform local optimization by edge flipping; Constraint handling (optional): If terrain feature lines exist, these edges are forcibly retained as fixed edges of the triangulation network, and the topology of the surrounding triangles is adjusted.
[0046] In one specific embodiment of the present invention, a system for repairing holes in low-density, large-area point cloud data includes: The data acquisition module is used to acquire raw point cloud data; The preprocessing module is used to preprocess the raw point cloud data to generate a raw point cloud mesh. The spatial calculation module is used to set the range line of the point cloud hole region, and perform spatial calculations based on the range line and the original point cloud mesh to obtain the original point cloud set. The generation module is used to generate a first planar simulated point set within the range line according to the original point cloud density; The clustering analysis module is used to filter and cluster the original point cloud and the first plane simulation point set to obtain the second plane simulation point set; The planar perturbation module is used to perturb the second planar simulation point set to generate a third planar simulation point set; The terrain construction module is used to construct terrain based on the original point set, forming an irregular triangular network. The elevation interpolation module is used to traverse the third plane simulation point set, perform elevation interpolation based on the irregular triangular network, and generate the first simulation point cloud set. The elevation disturbance module is used to perform elevation disturbance on the first simulated point set to form a second simulated point set; The data fusion module is used to fuse the second simulated point cloud set with the original point cloud set and output the final point cloud data result within the hole range.
[0047] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0048] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for repairing holes in low-density, large-area point cloud data, characterized in that, include: Obtain raw point cloud data; The original point cloud data is preprocessed to generate an original point cloud mesh; Define the boundary line of the point cloud hole region, and perform spatial calculations based on the boundary line and the original point cloud mesh to obtain the original point cloud set; A first planar simulated point set is generated within the specified range line according to the original point cloud density; The original point cloud and the first plane simulation point set are filtered and clustered to obtain the second plane simulation point set; The second set of simulated points in the plane is perturbed to generate a third set of simulated points in the plane; Based on the original point set, terrain is constructed to form an irregular triangular network; Traverse the third plane simulation point set, perform elevation interpolation based on the irregular triangular network, and generate the first simulation point cloud set; The elevation of the first simulated point cluster is disturbed to form a second simulated point cluster. The second simulated point cloud set is fused with the original point cloud set to output the final point cloud data result within the hole range.
2. The method for repairing holes in low-density, large-area point cloud data according to claim 1, characterized in that, Preprocessing the raw point cloud data includes: The original point cloud data is divided into N×N intervals according to an empirical step size to form an original point cloud mesh W1, and a quadtree spatial index is established.
3. The method for repairing holes in low-density, large-area point cloud data according to claim 1, characterized in that, Spatial operations are performed based on the defined boundary line and the original point cloud mesh to obtain the original point cloud set, including: Spatial overlay analysis is performed on the range line and the original point cloud mesh to obtain the point cloud data contained when the range line is projected onto the original point cloud mesh, as well as the point cloud data within the range line with index, and to construct the original point cloud set.
4. The method for repairing holes in low-density, large-area point cloud data according to claim 1, characterized in that, The original point cloud and the first plane simulation point set are filtered and clustered to obtain the second plane simulation point set, including: The original point cloud and the first plane simulation point set are preprocessed and fused to obtain a fused point set; The fusion point set is clustered using a density clustering algorithm or a region growing clustering algorithm to obtain clusters; The clusters are filtered to obtain candidate clusters; The candidate clusters are fitted to the RANSAC plane to generate a second plane simulation point set.
5. The method for repairing holes in low-density, large-area point cloud data according to claim 4, characterized in that, The fusion point set is clustered using a density clustering algorithm to obtain clusters, including: For the ε-neighborhood of each point in the fusion point set, mark the core point, boundary point, and noise point; Clusters are generated by merging core points with achievable density; Output the clustering results.
6. The method for repairing holes in low-density, large-area point cloud data according to claim 4, characterized in that, The fusion point set is clustered using a region growing clustering algorithm to obtain clusters, including: Starting from the seed point in the fusion point set, calculate the threshold of the angle between the normal vectors of neighboring points; Iteratively expand the points that satisfy the conditions until no further growth is possible, forming a planar cluster; Clusters with high overlap with the fusion point set in the planar clusters are selected as clustering clusters.
7. The method for repairing holes in low-density, large-area point cloud data according to claim 1, characterized in that, Based on the original point set, terrain is constructed to form an irregular triangular network, including: Create a triangle containing all the points as the initial state for the algorithm; The points in the original point cloud are inserted in spatial order. For each new point inserted, an existing triangle containing that point is found and split into three sub-triangles. Check whether adjacent triangles satisfy the Delaunay condition, and perform local optimization by edge flipping; If terrain feature lines exist, these edges are retained as fixed edges of the triangulation, and the topology of the surrounding triangles is adjusted.
8. A system for repairing low-density, large-area point cloud data holes, employing the method for repairing low-density, large-area point cloud data holes as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire raw point cloud data; The preprocessing module is used to preprocess the raw point cloud data to generate the raw point cloud mesh. The spatial calculation module is used to set the range line of the point cloud hole region, and perform spatial calculations based on the range line and the original point cloud mesh to obtain the original point cloud set. The generation module is used to generate a first planar simulated point set within the range line according to the original point cloud density; The clustering analysis module is used to filter and cluster the original point cloud and the first plane simulation point set to obtain the second plane simulation point set; The planar perturbation module is used to perform planar perturbation on the second planar simulation point set to generate a third planar simulation point set; The terrain construction module is used to construct terrain based on the original point set, forming an irregular triangular network. The elevation interpolation module is used to traverse the third plane simulation point set, perform elevation interpolation based on the irregular triangular network, and generate the first simulation point cloud set. The elevation disturbance module is used to perform elevation disturbance on the first simulated point set to form a second simulated point set; The data fusion module is used to fuse the second simulated point cloud set with the original point cloud set and output the final point cloud data result within the hole range.