Transformer substation plant extraction method based on point cloud morphological analysis and density clustering

Abstract

The invention provides a transformer substation plant extraction method based on point cloud morphological analysis and density clustering, and belongs to the technical field of transformer substation plant extraction, and the method comprises the steps: obtaining a transformer substation plant point cloud, removing the point cloud noise of the transformer substation plant point cloud data, and separating a ground point cloud and a high-altitude wire point cloud; in the separated substation point clouds, carrying out rapid density clustering analysis on the remaining point clouds, and extracting to obtain large-scale and dense clustering block point clouds; then PCA three-dimensional shape analysis is carried out, and clustering block point clouds which are regular in form and conform to factory building characteristics are extracted to serve as substation factory building point cloud candidates; and finally, projecting to the top view, and judging whether the point cloud is the substation plant point cloud by analyzing the projection form. According to the method, the ground points can be accurately removed in a complex transformer substation environment, and accurate extraction of the factory building is completed by means of rapid clustering analysis, PCA three-dimensional shape analysis, projection verification and the like.

Description

Technical Field

[0001] The invention relates to the technical field of substation plant extraction, and in particular to a substation plant extraction method based on point cloud morphological analysis and density clustering. Background Art

[0002] Laser radar (LiDAR) technology has developed rapidly in recent years. As a high-precision three-dimensional measurement method, it has been widely used in many fields such as urban modeling, road surveying, and substation inspection. LiDAR generates high-density three-dimensional point cloud data by emitting laser beams and recording reflected signals. These point cloud data can reflect key information such as the shape, size, and spatial distribution of objects.

[0003] However, despite the great potential of laser point cloud data in spatial information extraction, traditional manual processing methods are inefficient and have limited accuracy when faced with large-scale, high-density point cloud data. This is especially true in the extraction of building structures in complex environments. For example, in scenes such as substations, due to their special environment and complex structure, conventional detection methods are dangerous and complicated, and manual methods are difficult to achieve efficient automatic extraction. Therefore, how to quickly and accurately extract building structures from laser point cloud data, especially the geometric form of substation buildings, has become one of the hot spots and challenges of current research.

[0004] Existing building extraction methods mainly focus on the extraction of urban buildings or road surrounding facilities. Generally, these methods rely on the removal of ground points, clustering of point clouds, and extraction of geometric features. In relatively simple urban environments, these methods can better identify the outline and structure of buildings and become an effective means of building extraction. However, in complex industrial environments such as substations, existing methods clearly expose many shortcomings.

[0005] The substation environment is characterized by complex structures such as power lines and towers. These elements block the point cloud data of the plant from multiple perspectives, making it difficult to peel and extract the complete plant point cloud. In addition, many components in the substation are similar to the plant in volume and point cloud density, which makes it difficult for a single clustering method to effectively distinguish the two. Therefore, in order to meet the needs of building extraction in the special environment of the substation, it is urgent to propose a more advanced and efficient classification scheme to overcome the limitations of existing methods in such environments. Summary of the invention

[0006] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a substation building extraction method based on point cloud morphological analysis and density clustering, which can accurately remove ground points in a complex substation environment and complete the accurate extraction of the building through means such as rapid clustering analysis, PCA three-dimensional shape analysis and projection verification.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] A substation building extraction method based on point cloud morphological analysis and density clustering includes the following steps:

[0009] S1. Obtaining a point cloud of a substation building, removing point cloud noise from the point cloud data of the substation building, and separating the ground point cloud and the overhead wire point cloud;

[0010] S2. In the separated substation cloud, fast density clustering analysis is performed on the remaining point clouds to extract large-scale and dense cluster block point clouds;

[0011] S3, performing PCA three-dimensional shape analysis on the large-scale and dense cluster block point cloud, extracting the cluster block point cloud with regular morphology and conforming to the characteristics of the plant as a candidate for the substation plant point cloud;

[0012] S4. Project the obtained cluster block point cloud with regular shape and consistent with the characteristics of the plant onto the top view, and determine whether it is a substation plant point cloud by analyzing the projection shape.

[0013] Preferably, in step S1, it specifically includes:

[0014] Obtain the point cloud of the substation building through laser scanning or drone aerial survey;

[0015] Remove point cloud noise and outliers through statistical filtering algorithm;

[0016] The initial ground plane is established by randomly selecting three points, setting the number of iterations and the inlier threshold, identifying the inliers and recording the plane model, and separating the ground point cloud and the overhead wire point cloud.

[0017] Preferably, the formula of the statistical filtering algorithm is:

[0018]

[0019] Where, d i is the average neighborhood distance of the i-th point, k is the number of neighboring points, and p i and p j are the coordinates of point i and neighboring point j respectively.

[0020] Preferably, the initial ground plane is established by randomly selecting three points, and the number of iterations and the inlier threshold are set, the inliers are identified and the plane model is recorded, and the ground point cloud and the overhead wire point cloud are separated, including:

[0021] After removing the point cloud noise from the substation building point cloud, three points are randomly selected to establish the initial ground plane, and the equation of the initial ground plane is:

[0022] Ax+By+Cz+D=0;

[0023] Use the coordinates of three random points to establish a plane model, and calculate the distances from all points to the plane:

[0024]

[0025] Set the number of iterations k, the inlier threshold t and the minimum inlier ratio d. If the distance from a point to the plane model is less than the inlier threshold, the point can be determined to be a point P in the plane. inliers :

[0026] P inliers ={p i |d i <t};

[0027] If the number of inliers is greater than the set minimum inlier ratio, the plane model is recorded and all inliers are removed from the point cloud to separate the ground point cloud and the overhead wire point cloud in the substation building point cloud.

[0028] Preferably, in step S2, it specifically includes:

[0029] Perform voxel network downsampling on the separated substation cloud, divide the substation cloud into several voxels, and assign a center point to each voxel;

[0030] According to the substation cloud after downsampling of the voxel network, the neighborhood radius is defined based on the local density histogram and the minimum number of points is determined;

[0031] According to the DBSCAN algorithm, the core points are determined by density gradient, and the substation cloud is density clustered to extract a large-scale and dense cluster block point cloud.

[0032] Preferably, defining the neighborhood radius and determining the minimum number of points based on the local density histogram includes:

[0033] The neighborhood radius ε is defined based on the local density histogram:

[0034] ε=elbow method({d1,d2,...,d n});

[0035] And determine the minimum number of points through a heuristic search method:

[0036]

[0037] Among them, MinPts is the minimum number of points.

[0038] Preferably, in step S3, it specifically includes:

[0039] Based on step S2, the coordinate data matrix M of the neighborhood point cloud set P of the target point is identified, and the covariance matrix is ​​calculated:

[0040]

[0041] Where N is the number of points;

[0042] Then, the covariance matrix is ​​decomposed into eigenvalues ​​CV = VΛ to obtain the eigenvalue diagonal matrix Λ and the eigenvector V. The morphological information of the point cloud is determined by the size of the eigenvectors λ1, λ2, and λ3 corresponding to the diagonal elements in the eigenvalue diagonal matrix Λ.

[0043] Use a L 、a P 、a V They represent the possibility of the point cloud shape being linear, planar, and solid, respectively, and are calculated by the eigenvectors λ1, λ2, and λ3. The formula is:

[0044]

[0045] when a P >a L And a P >a V , then the part of the point cloud is judged to be a plane shape, and a cluster block point cloud with regular shape and consistent with the characteristics of the factory building is obtained.

[0046] Preferably, in step S4, it specifically includes:

[0047] The obtained cluster block point cloud is projected onto the top view plane, and then the projected point cloud data is rasterized to generate a binary image:

[0048] Grid(i,j)={p k |p k ∈X proj and(x k ,y k )∈cell(i,j)};

[0049] In the formula, X proj is the point cloud to be projected, cell(i,j) is the grid with row and column index (i,j), p k is any point in the grid cell (i, j);

[0050] Then find all connected domains in the projected point cloud data, assign a unique label to each connected domain, and set the threshold N threshold , where all connected domains larger than the threshold are considered as candidates for substation buildings;

[0051] Contour detection is performed on all candidate connected domains, and the Douglas-Peucker algorithm is used to obtain the contour points and edges of the point cloud point by point, and the parallel relationship between the edges is calculated one by one; if the number of parallel relationships is greater than two pairs, the connected domain is a candidate for the substation building location, and then the connected domain is compared with the substation building point cloud candidate in step S3 to obtain the final substation building point cloud.

[0052] According to the specific embodiments provided by the present invention, compared with the prior art, the present invention discloses the following technical effects:

[0053] (1) The present invention can accurately identify and remove points of non-building structures, such as cables, equipment brackets, etc., thereby improving the authenticity and reliability of point cloud data. At the same time, through rapid density clustering, the building areas in the point cloud can be quickly identified and classified, thereby improving processing efficiency. In addition, the geometric features of the point cloud are analyzed by PCA to extract key shape features from multidimensional data, which helps to accurately identify the plant structure. A projection verification step is introduced in the point cloud extraction process to ensure that the extracted plant structure has a high degree of geometric consistency and accuracy, effectively avoiding the problem of misextraction and completing the accurate extraction of the substation plant.

[0054] (2) The present invention adopts the above-mentioned substation building extraction method based on point cloud morphological analysis and density clustering, which realizes efficient and accurate substation building structure extraction and improves the quality and efficiency of data processing; at the same time, the extracted results are suitable for fields such as automated inspection, facility monitoring and engineering measurement, and have broad application prospects; and the automated extraction process reduces manual intervention, reduces work risks, and improves the safety and efficiency of inspection and maintenance work. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0056] Figure 1 It is a flow chart of a substation plant extraction method based on point cloud morphology analysis and density clustering of the present invention;

[0057] Figure 2 A schematic diagram of the first embodiment of the present invention;

[0058] Figure 3 A schematic diagram of ground point removal provided in Embodiment 1 of the present invention;

[0059] Figure 4 A schematic diagram of point cloud density clustering provided in the first embodiment of the present invention;

[0060] Figure 5 A schematic diagram of an original point cloud provided in the first embodiment of the present invention;

[0061] Figure 6 This is a schematic diagram of the projected point cloud provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] Embodiment 1

[0065] like Figure 1 and Figure 2 As shown, the present invention provides a substation building extraction method based on point cloud morphological analysis and density clustering, comprising the following steps:

[0066] S1. Obtaining a point cloud of a substation building, removing point cloud noise from the point cloud data of the substation building, and separating the ground point cloud and the overhead wire point cloud;

[0067] S2. In the separated substation cloud, fast density clustering analysis is performed on the remaining point clouds to extract large-scale and dense cluster block point clouds;

[0068] S3, performing PCA three-dimensional shape analysis on the large-scale and dense cluster block point cloud, extracting the cluster block point cloud with regular morphology and conforming to the characteristics of the plant as a candidate for the substation plant point cloud;

[0069] S4. Project the obtained cluster block point cloud with regular shape and consistent with the characteristics of the plant onto the top view, and determine whether it is a substation plant point cloud by analyzing the projection shape.

[0070] Wherein, in step S1, specifically including:

[0071] Obtain the point cloud of the substation building through laser scanning or drone aerial survey;

[0072] Remove point cloud noise and outliers through statistical filtering algorithm;

[0073] The formula of the statistical filtering algorithm is:

[0074]

[0075] Where, d i is the average neighborhood distance of the i-th point, k is the number of neighboring points, and p i and p j are the coordinates of point i and neighboring point j respectively.

[0076] The initial ground plane is established by randomly selecting three points, and the number of iterations and the inlier threshold are set. The inliers are identified and the plane model is recorded. The ground point cloud and the overhead wire point cloud are separated, including:

[0077] After removing the point cloud noise from the substation building point cloud, three points are randomly selected to establish the initial ground plane, and the equation of the initial ground plane is:

[0078] Ax+By+Cz+D=0;

[0079] Use the coordinates of three random points to establish a plane model, and calculate the distances from all points to the plane:

[0080]

[0081] Set the number of iterations k, the inlier threshold t and the minimum inlier ratio d. If the distance from a point to the plane model is less than the inlier threshold, the point can be determined to be a point P in the plane. inliers :

[0082] P inliers ={p i |d i <t};

[0083] If the number of inliers is greater than the set minimum inlier ratio, the plane model is recorded and all inliers are removed from the point cloud to separate the ground point cloud and the overhead wire point cloud in the substation building point cloud. The result is as follows: Figure 3 shown.

[0084] In addition, in step S2, it specifically includes:

[0085] Perform voxel network downsampling on the separated substation cloud, divide the substation cloud into several voxels, and assign a center point to each voxel;

[0086] According to the substation cloud after downsampling of the voxel network, the neighborhood radius is defined based on the local density histogram and the minimum number of points is determined;

[0087] The neighborhood radius ε is defined based on the local density histogram:

[0088] ε=elbow method({d1,d2,...,d n});

[0089] And determine the minimum number of points MinPts through a heuristic search method:

[0090]

[0091] According to the DBSCAN algorithm, the core points are determined by density gradient, and the substation cloud is clustered by density to extract a large-scale and dense cluster block point cloud. The results are as follows: Figure 4 shown.

[0092] Secondly, in step S3, it specifically includes:

[0093] Based on step S2, the coordinate data matrix M of the neighborhood point cloud set P of the target point is identified, and the covariance matrix is ​​calculated:

[0094]

[0095] Where N is the number of points;

[0096] Then, the covariance matrix is ​​decomposed into eigenvalues ​​CV = VΛ to obtain the eigenvalue diagonal matrix Λ and the eigenvector V. The morphological information of the point cloud is determined by the size of the eigenvectors λ1, λ2, and λ3 corresponding to the diagonal elements in the eigenvalue diagonal matrix Λ.

[0097] Use a L 、a P 、a V They represent the possibility of the point cloud shape being linear, planar, and solid, respectively, and are calculated by the eigenvectors λ1, λ2, and λ3. The formula is:

[0098]

[0099] when a P >a L And a P >a V , then the part of the point cloud is judged to be a plane shape, and a cluster block point cloud with regular shape and consistent with the characteristics of the factory building is obtained.

[0100] Finally, in step S4, it specifically includes:

[0101] The obtained cluster block point cloud is projected onto the top view plane, and then the projected point cloud data is rasterized to generate a binary image:

[0102] Grid(i,j)={p k |p k ∈X proj and(x k ,y k )∈cell(i,j)};

[0103] In the formula, X proj is the point cloud to be projected, cell(i,j) is the grid with row and column index (i,j), p k is any point in the grid cell (i, j); the results of the original point cloud and the projected point cloud are as follows Figure 5 and Figure 6 shown.

[0104] Then find all connected domains in the projected point cloud data, assign a unique label to each connected domain, and set the threshold N threshold , where all connected domains larger than the threshold are considered as candidates for substation buildings;

[0105] Contour detection is performed on all candidate connected domains, and the Douglas-Peucker algorithm is used to obtain the contour points and edges of the point cloud point by point, and the parallel relationship between the edges is calculated one by one; if the number of parallel relationships is greater than two pairs, the connected domain is a candidate for the substation building location, and then the connected domain is compared with the substation building point cloud candidate in step S3 to obtain the final substation building point cloud.

[0106] Therefore, the above-mentioned substation building extraction method based on point cloud morphological analysis and density clustering can accurately remove ground points in a complex substation environment and complete the accurate extraction of the building through rapid clustering analysis, PCA three-dimensional shape analysis and projection verification.

[0107] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A substation building extraction method based on point cloud morphological analysis and density clustering, characterized in that: The following steps are involved: S1. Obtaining a point cloud of a substation building, removing point cloud noise from the point cloud data of the substation building, and separating the ground point cloud and the overhead wire point cloud; S2. In the separated substation cloud, fast density clustering analysis is performed on the remaining point clouds to extract large-scale and dense cluster block point clouds; S3, performing PCA three-dimensional shape analysis on the large-scale and dense cluster block point cloud, extracting the cluster block point cloud with regular morphology and conforming to the characteristics of the plant as a candidate for the substation plant point cloud; S4. Project the obtained cluster block point cloud with regular shape and consistent with the characteristics of the plant onto the top view, and determine whether it is a substation plant point cloud by analyzing the projection shape.

2. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 1 is characterized in that: In step S1, it specifically includes: Obtain the point cloud of the substation building through laser scanning or drone aerial survey; Remove point cloud noise and outliers through statistical filtering algorithm; The initial ground plane is established by randomly selecting three points, setting the number of iterations and the inlier threshold, identifying the inliers and recording the plane model, and separating the ground point cloud and the overhead wire point cloud.

3. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 2 is characterized in that: The formula of the statistical filtering algorithm is: Where, d i is the average neighborhood distance of the i-th point, k is the number of neighboring points, and p i and p j are the coordinates of point i and neighboring point j respectively.

4. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 2 is characterized in that: The initial ground plane is established by randomly selecting three points, and the number of iterations and the inlier threshold are set. The inliers are identified and the plane model is recorded. The ground point cloud and the overhead wire point cloud are separated, including: After removing the point cloud noise from the substation building point cloud, three points are randomly selected to establish the initial ground plane, and the equation of the initial ground plane is: Ax+By+Cz+D=0; Use the coordinates of three random points to establish a plane model, and calculate the distances from all points to the plane: Set the number of iterations k, the inlier threshold t and the minimum inlier ratio d. If the distance from a point to the plane model is less than the inlier threshold, the point can be determined to be a point P in the plane. inliers : P inliers ={p i |d i <t}; If the number of inliers is greater than the set minimum inlier ratio, the plane model is recorded and all inliers are removed from the point cloud to separate the ground point cloud and the overhead wire point cloud in the substation building point cloud.

5. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 1 is characterized in that: In step S2, it specifically includes: Perform voxel network downsampling on the separated substation cloud, divide the substation cloud into several voxels, and assign a center point to each voxel; According to the substation cloud after downsampling of the voxel network, the neighborhood radius is defined based on the local density histogram and the minimum number of points is determined; According to the DBSCAN algorithm, the core points are determined by density gradient, and the substation cloud is density clustered to extract a large-scale and dense cluster block point cloud.

6. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 5 is characterized in that: Define the neighborhood radius and determine the minimum number of points based on the local density histogram, including: The neighborhood radius ε is defined based on the local density histogram: ε=elbow method({d1,d2,...,d n }); And determine the minimum number of points through a heuristic search method: Among them, MinPts is the minimum number of points.

7. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 1 is characterized in that: In step S3, it specifically includes: Based on step S2, the coordinate data matrix M of the neighborhood point cloud set P of the target point is identified, and the covariance matrix is ​​calculated: Where N is the number of points; Then, the covariance matrix is ​​decomposed into eigenvalues ​​CV = VΛ to obtain the eigenvalue diagonal matrix Λ and the eigenvector V. The morphological information of the point cloud is determined by the size of the eigenvectors λ1, λ2, and λ3 corresponding to the diagonal elements in the eigenvalue diagonal matrix Λ. Use a L 、a P 、a V They represent the possibility of the point cloud shape being linear, planar, and solid, respectively, and are calculated by the eigenvectors λ1, λ2, and λ3. The formula is: when a P >a L And a P >a V , then the part of the point cloud is judged to be a plane shape, and a cluster block point cloud with regular shape and consistent with the characteristics of the factory building is obtained.

8. The method for extracting substation buildings based on point cloud morphological analysis and density clustering according to claim 1 is characterized in that: In step S4, it specifically includes: The obtained cluster block point cloud is projected onto the top view plane, and then the projected point cloud data is rasterized to generate a binary image: Grid(i,j)={p k |p k ∈X proj and(x k ,y k )∈cell(i,j)}; Where, X proj is the point cloud to be projected, cell(i,j) is the grid with row and column index (i,j), p k is any point in the grid cell (i, j); Then find all connected domains in the projected point cloud data, assign a unique label to each connected domain, and set the threshold N threshold , where all connected domains larger than the threshold are considered as candidates for substation buildings; Contour detection is performed on all candidate connected domains, and the Douglas-Peucker algorithm is used to obtain the contour points and edges of the point cloud point by point, and the parallel relationship between the edges is calculated one by one; if the number of parallel relationships is greater than two pairs, the connected domain is a candidate for the substation building location, and then the connected domain is compared with the substation building point cloud candidate in step S3 to obtain the final substation building point cloud.

Images