Wire distance calculation method fusing regular part extraction and clustering growth

Abstract

The invention provides a wire distance calculation method fusing regular component extraction and clustering growth, and belongs to the technical field of power equipment monitoring, and the method comprises the steps: obtaining point cloud data of a transformer substation, carrying out the noise and outlier removal and ground point cloud elimination processing, and obtaining the residual point cloud data containing power transformation facilities and various wires; a deep learning model is utilized to extract components with regular structures from the residual point cloud data, and various wires and power transformation facilities connected with the wires are preliminarily labeled and identified; then classifying various types of wires, cutting out a point cloud area where the power transformation facility is located, and further refining and extracting a power transformation facility point cloud connected with the wires by applying a clustering growth algorithm; and finally, calculating the distance between the wires, the distance between the wires and the ground and the distance between the wires and the power transformation facility, and generating an analysis report for operation, maintenance and monitoring of a power transformation transmission line. According to the method, the problems of high difficulty and incapability of accurately extracting the lead and the power transformation facility in the existing method are solved.

Description

Technical Field

[0001] The invention relates to the technical field of power equipment monitoring, and in particular to a wire distance calculation method integrating regular component extraction and cluster growth. Background Art

[0002] With the rapid development of power facilities and the improvement of automation, the daily inspection and maintenance of power transmission lines in complex areas of substation power facilities are of great significance to improving power supply safety and reliability. Traditional manual inspection methods can no longer meet the requirements of modern power facility maintenance, especially in vast power facility areas and high-altitude operation scenarios, which pose great safety risks and time costs. Therefore, intelligent analysis technology based on point cloud data has gradually become a popular research direction for power facility detection.

[0003] In power transmission lines, the distances between conductors, between conductors and power facilities, and between conductors and the ground are key parameters that are directly related to the safe operation of the transmission system. Improper distances between power conductors and equipment may lead to safety issues such as short circuits and fires. Especially in high-voltage transmission environments, accurate distance calculation is particularly important. Point cloud data, due to the rich spatial geometric information it contains, makes it possible to automatically measure such distances. However, due to the complex structure of various components in power facilities, the huge scale of point cloud data, and the complex spatial relationships between different facilities, how to efficiently and accurately extract conductors and power facilities from massive point cloud data and perform related distance calculations has become a major challenge.

[0004] Most current methods focus on extracting conductors or facilities directly from point clouds, but lack full utilization of regular components, especially cross-arm structure facilities, which are important components with regularity. As a result, existing methods are difficult to accurately extract conductors and substation facilities. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a conductor distance calculation method that integrates regular component extraction and cluster growth, uses a deep learning model to accurately extract cross-arm structure facilities and complex power facility point clouds in point cloud data, and uses classification and clustering algorithms to carefully analyze the conductor type and its spatial distribution, thereby achieving accurate calculation of the distance between power conductors and equipment, providing an efficient and accurate solution for intelligent detection and automated maintenance of power transmission systems, and greatly improving the accuracy of facility monitoring and the degree of automation of maintenance.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A wire distance calculation method integrating regular component extraction and cluster growth comprises the following steps:

[0008] S1. Obtaining substation point cloud data, and performing noise and outlier removal and ground point cloud culling processing on the point cloud data to obtain remaining point cloud data including substation facilities and various conductors;

[0009] S2. Use the deep learning model to extract components with regular structures from the remaining point cloud data, and preliminarily mark and identify various types of conductors and the substation facilities connected to them;

[0010] S3. Classify various types of conductors based on the extracted regular structural components, and cut out the point cloud area where the substation facilities are located according to the spatial directions of the classified conductors;

[0011] S4, based on step S3, applying a cluster growth algorithm to further refine and extract the substation facility point cloud connected to the conductor;

[0012] S5. After the extraction and classification of substation facilities and conductors, the conductor spacing, the distance between the conductors and the ground, and the distance between the conductors and substation facilities are calculated, and an analysis report is generated for the operation, maintenance and monitoring of substation transmission lines.

[0013] Preferably, in step S1, the point cloud data is subjected to noise and outlier removal and ground point cloud culling processing, including:

[0014] Select a statistical filter to determine the outliers by calculating the distribution of neighboring points of each point, and set the point cloud set P = {p i |i=1,2,...,N}, each point p i In three-dimensional space, the coordinates are (x i ,y i ,z i ), select each point p i The k nearest points of the point p are calculated. i The average distance d to the k nearest points i :

[0015]

[0016] Among them, d i For point p i The average Euclidean distance between its k nearest neighbors;

[0017] By counting the average distance distribution of all points, the mean μ and standard deviation σ are calculated, and d i Points that exceed the following distance threshold conditions are considered as outliers and filtered out:

[0018] d i >μ+ασ;

[0019] Among them, α is a user-defined parameter ranging from 1 to 3, so that outliers are filtered out of the point cloud in the overall scene of the substation area by statistical outlier filtering;

[0020] Then, the RANSAC random sampling consensus algorithm is used to fit and filter out the ground points, and three non-collinear points are randomly selected to form a candidate plane. The maximum inner point set is iteratively obtained to form a fitting plane, thereby identifying the ground plane and retaining the height information of the ground points.

[0021] Preferably, in step S2, it specifically includes: firstly, using a voxel-based Transformer network model to process the input point cloud data, voxelize the point cloud data, and form a grid of fixed size; secondly, using a multi-scale feature extraction strategy to extract each local feature of the point cloud in the voxel and output a feature of dimension (H, C), where H is the number of voxels and C is the feature dimension corresponding to the voxel; then, using the Transformer network to extract the context features between voxels in the vertical and horizontal directions, obtain the corresponding feature representation through the QKV weight matrix, and perform self-attention mechanism calculation, and the output size is (H, C 1 ) voxel features; the output voxel features are resampled again to restore to point-by-point features, and connected with the classification task head to complete point-by-point point cloud classification; finally, the various types of wires to be extracted and the substation facilities connected to them are preliminarily marked and identified.

[0022] Preferably, in step S3, it specifically includes:

[0023] The input point cloud data is voxelized, that is, the three-dimensional space is divided into a uniform voxel grid, and the size of each voxel is defined as V x ×V y ×V z , where V x 、V y 、V z Represent the size of the voxel in the X, Y and Z directions respectively, and use the formula to convert each point p of the point cloud into i =(x i ,y i ,z i ) is mapped into a voxel grid:

[0024]

[0025] Among them, v x , v y , v zRepresents the corresponding voxel index; using the height Z value information of the regular component as the index, quickly cluster the wire voxels of the corresponding height in the adjacent neighborhood, and after clustering, divide the long-distance transmission line point cloud voxels, and extract the corresponding non-substation facility area point cloud;

[0026] The divided long-distance transmission lines are removed, and then the spatial direction analysis is performed on the soft wires connected to the substation facilities to determine the projection direction of the soft wires on the XY plane. The entire point cloud is cropped on the XY plane to obtain the point cloud area of ​​all substation facilities within the Z-axis height range.

[0027] Preferably, in step S4, it specifically includes:

[0028] Based on the cropped point cloud area obtained in step S3, the initial cluster center is determined by the local density peak clustering method, and the given point cloud data set P = {p i |i=1,2,...,N} defines two key indicators for each point: local density ρ i , distance δ i , where the local density ρ i Calculated by the following formula:

[0029]

[0030] Among them, d ij Represents point p i and point p j The Euclidean distance between them is then calculated by constructing a similarity matrix and sorting the distances in descending order, selecting the top 2% as the cutoff distance d c , is the indicator function, if d ij <d c , on the contrary Thus, the areas with lower local density are identified as the initial cluster centers;

[0031] After determining the initial cluster center, the geometric information of the wire is used for constraint, and the centerline position and direction vector of the wire are calculated, where the direction vector is denoted as v = (v x ,v y ,v z ); In the cluster growth process, by calculating the point p i With the neighboring point p j The relative position vector u ij To control the cluster growth direction:

[0032] u ij =(x i -x j ,y i -y j,z i -z j );

[0033] And by calculating the angle θ between the wire direction v and the relative position vector:

[0034]

[0035] Set the angle threshold θ max To ensure that the cluster grows only along the wire direction;

[0036] After the initial cluster center is determined and the geometric constraints are completed, the cluster growth will gradually expand under the constraints of local density and geometric direction. The cluster growth process is expressed as:

[0037] C(p j )=C(p i )if vu ij ≥cosθ max and ρ j < threshold ;

[0038] Among them, C(p i ) is point p i The cluster to which it belongs, C(p j ) is point p j The cluster to which it will be assigned;

[0039] Finally, the point cloud of substation facilities connected to the conductors is further refined and extracted through density peaks and geometric constraints.

[0040] Preferably, in step S5, it specifically includes: firstly, regularizing the projection of the extracted conductors and substation facilities respectively, projecting the extracted conductors along the XY plane, calculating the spacing distance between the conductors with each assigned ID, and obtaining the spatial distribution relationship between different conductors; then projecting the conductors along the Z-axis direction to obtain the distance between the conductors and the substation facilities and the distance between the conductors and the ground point.

[0041] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0042] The present invention provides a method for calculating the distance between power conductors and equipment by integrating the extraction of regular components with cluster growth. The method uses a deep learning model to accurately extract point clouds of cross-arm structure facilities and complex power facilities in point cloud data. The method uses classification and clustering algorithms to carefully analyze the conductor types and their spatial distribution, thereby achieving accurate calculation of the distance between power conductors and equipment. The method provides an efficient and accurate solution for the intelligent detection and automated maintenance of power transmission systems, greatly improving the accuracy of facility monitoring and the degree of automation of maintenance, thereby solving the problems of the existing methods that are difficult to accurately extract conductors and substation facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] 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.

[0044] Figure 1 A flow chart of a wire distance calculation method integrating regular component extraction and cluster growth according to the present invention;

[0045] Figure 2 A schematic diagram of a method provided in Embodiment 1 of the present invention;

[0046] Figure 3 A schematic diagram of the point cloud semantic segmentation network model design provided in the first embodiment of the present invention;

[0047] Figure 4 A schematic diagram of clustering and dividing conductors along the cross-arm structure direction provided in the first embodiment of the present invention;

[0048] Figure 5 A schematic diagram of the clustering results of substation facilities provided in the first embodiment of the present invention;

[0049] Figure 6 A schematic diagram of a wire distance calculation and analysis report provided in the first embodiment of the present invention; wherein: Figure 6 (a) is the conductor distance calculation diagram after projection. Figure 6 (b) is the distance calculation diagram between the conductor and the substation equipment. Figure 6 (c) is the conductor-to-ground measurement diagram. DETAILED DESCRIPTION

[0050] 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.

[0051] 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.

[0052] Embodiment 1

[0053] like Figure 1 and Figure 2As shown, the present invention provides a wire distance calculation method integrating regular component extraction and cluster growth, comprising the following steps:

[0054] S1. Obtaining substation point cloud data, and performing noise and outlier removal and ground point cloud culling processing on the point cloud data to obtain remaining point cloud data including substation facilities and various conductors;

[0055] S2. Use the deep learning model to extract components with regular structures from the remaining point cloud data, and preliminarily mark and identify various types of conductors and the substation facilities connected to them;

[0056] S3. Classify various types of conductors based on the extracted regular structural components, and cut out the point cloud area where the substation facilities are located according to the spatial directions of the classified conductors;

[0057] S4, based on step S3, applying a cluster growth algorithm to further refine and extract the substation facility point cloud connected to the conductor;

[0058] S5. After the extraction and classification of substation facilities and conductors, the conductor spacing, the distance between the conductors and the ground, and the distance between the conductors and substation facilities are calculated, and an analysis report is generated for the operation, maintenance and monitoring of substation transmission lines.

[0059] In step S1, the point cloud data is subjected to noise and outlier removal and ground point cloud culling processing, including:

[0060] Select a statistical filter to determine the outliers by calculating the distribution of neighboring points of each point, and set the point cloud set P = {p i |i=1,2,...,N}, each point p i In three-dimensional space, the coordinates are (x i ,y i ,z i ), select each point p i The k nearest points of the point p are calculated. i The average distance d to the k nearest points i :

[0061]

[0062] Among them, d i For point p i The average Euclidean distance between its k nearest neighbors;

[0063] By counting the average distance distribution of all points, the mean μ and standard deviation σ are calculated, and d i Points that exceed the following distance threshold conditions are considered as outliers and filtered out:

[0064] di >μ+ασ;

[0065] Among them, α is a user-defined parameter ranging from 1 to 3, so that outliers are filtered out of the point cloud in the overall scene of the substation area by statistical outlier filtering;

[0066] Then, the RANSAC random sampling consensus algorithm is used to fit and filter out the ground points. The random sampling consensus method estimates the plane model by iteratively selecting a subset of the point cloud. Three non-collinear points are randomly selected from the point cloud to form a candidate plane. At the same time, the distance from all points to the plane is calculated, and the number of inliers less than the set threshold ∈ is counted. After multiple iterations, the maximum inliers are obtained to form a fitting plane. After identifying the ground plane, it is filtered out, and the height of the ground points is retained for subsequent distance calculations.

[0067] Secondly, in step S2, based on the deep learning model, components with regular structures in power facilities are extracted, especially the cross-arm structure point cloud. Such facilities are usually regular and uniform, have high classification extraction accuracy and recall rate, and are connected to long-distance transmission lines, which can be further used for subsequent refined operations. At the same time, other components to be extracted are preliminarily labeled and identified, including various types of wires (hard wires, soft wires, long-distance transmission wires, etc.) and various types of substation facilities connected to the wires; specifically, the voxel-based Transformer network model is used to process the input point cloud data, and the point cloud data is voxelized to form a fixed-size grid; secondly, with reference to Figure 3 , the same multi-scale feature extraction strategy of Pointnet++ is used to extract local features of point clouds within voxels and output features of (H, C) dimensions, where H is the number of voxels and C is the feature dimension corresponding to the voxel; then the Transformer network is used to extract contextual features between voxels in the vertical and horizontal directions, and the corresponding V is obtained through the QKV weight matrix. query 、V key 、V value Feature representation, after which query and key pass through the self-attention mechanism and the feature is multiplied by the value feature and the output size is (H,C 1 ) voxel features; the output voxel features are resampled again to restore to point-by-point features, and connected with the classification task head to complete point-by-point point cloud classification; finally, the various types of wires to be extracted and the substation facilities connected to them are preliminarily marked and identified.

[0068] This model is used to extract different types of equipment in substation scenarios. Since the vertical pole-shaped cross-arm structure is extremely regular and stable, good results in accuracy and recall are achieved in all categories of projects. It is also connected to long-distance transmission lines at higher altitudes. Therefore, the cross-arm structure is used as a guide to divide the long-distance transmission lines, hard lines, and equipment connection lines in the line category. At the same time, the hard lines, equipment connection lines extracted by the alternative model and the rough results of the model-extracted equipment are used for refined division in subsequent steps.

[0069] Again, in step S3, it specifically includes:

[0070] The input point cloud data is voxelized, that is, the three-dimensional space is divided into a uniform voxel grid, and the size of each voxel is defined as V x ×V y ×V z , where V x 、V y 、V z Represent the size of the voxel in the X, Y and Z directions respectively, and use the formula to convert each point p of the point cloud into i =(x i ,y i ,z i ) is mapped into a voxel grid:

[0071]

[0072] Among them, v x , v y , v z Represents the corresponding voxel index; using the Z value information of the crossarm height as the index, the wire voxels of the corresponding height in the adjacent neighborhood are quickly clustered, such as Figure 4 As shown in the figure, after clustering is completed, the point cloud voxels of the long-distance transmission line are divided, and then the corresponding point cloud of the non-substation facility area is obtained.

[0073] The divided long-distance transmission lines are removed, and then the soft wires connected to the substation facilities are analyzed in spatial direction to determine the projection direction of the soft wires on the XY plane. The entire point cloud is cut on the XY plane to obtain the substation facility point cloud area within the Z-axis height range, including:

[0074] Determine the projection direction of the soft wire on the XY plane, and directly cut the entire point cloud on the XY plane to obtain the point cloud area of ​​all substation facilities within the Z-axis height range. The cutting process is as follows: First, based on the starting and ending positions of the soft wire, determine the range of the wire on the XY plane, that is, (x min ,y min ) to (x max ,y max), and then cut all point cloud data within the corresponding range, and the points that meet the conditions within the rectangular range belong to the possible area of ​​the point cloud. This step reduces the amount of subsequent point cloud processing data, reduces the calculation complexity, and speeds up the device's refined extraction process.

[0075] Again, in step S4, it specifically includes:

[0076] Based on the cropped point cloud area obtained in step S3, the initial cluster center is determined by the local density peak clustering method, and the given point cloud data set P = {p i |i=1,2,...,N} defines two key indicators for each point: local density ρ i , distance δ i , where the local density ρ i Calculated by the following formula:

[0077]

[0078] Among them, d ij Represents point p i and point p j The Euclidean distance between them is then calculated by constructing a similarity matrix and sorting the distances in descending order, selecting the top 2% as the cutoff distance d c , is the indicator function, if d ij <d c , on the contrary Thus, the area with low local density is identified as the initial cluster center; through the above local density formula, the area with low local density, such as the wire area, can be identified. Since the points in the wire area are usually sparsely distributed and the local density is low, these areas can be used as the initial cluster center.

[0079] After determining the initial cluster center, the geometric information of the wire is used for constraint, and the centerline position and direction vector of the wire are calculated, where the direction vector is denoted as v = (v x ,v y ,v z ); In the cluster growth process, by calculating the point p i With the neighboring point p j The relative position vector u ij To control the cluster growth direction:

[0080] u ij =(x i -x j ,y i -y j ,z i -z j );

[0081] And by calculating the angle θ between the wire direction v and the relative position vector:

[0082]

[0083] Set the angle threshold θ max To ensure that the cluster grows only along the wire direction, for cosθ>cosθ max If there are points, they will be added to the same cluster to control the scope and speed of clustering;

[0084] After the initial cluster center is determined and the geometric constraints are completed, the cluster growth will gradually expand under the constraints of local density and geometric direction. The cluster growth process is expressed as:

[0085] C(p j )=C(p i )if vu ij ≥cosθ max andρ j < threshold ;

[0086] Among them, C(p i ) is point p i The cluster to which it belongs, C(p j ) is point p j The cluster to which it will be assigned;

[0087] The above method uses density peaks and geometric constraints to quickly extract the point cloud of substation facilities connected to the conductors, and completes the substation facilities extracted by the previous model to further improve the extraction of power facilities. The results are as follows: Figure 5 shown.

[0088] Finally, in step S5, the following steps are specifically performed: first, regularized projection is performed on the extracted conductors and substation facilities, the extracted conductors are projected along the XY plane, and the spacing distance between the conductors of each assigned ID is calculated. The result is as follows: Figure 6 As shown in (a) in the figure, the spatial distribution relationship between different conductors is obtained; then the conductor is projected along the Z axis to obtain the distance between the conductor and the substation facility and the distance between the conductor and the ground point. The result is as follows Figure 6 As shown in (b) and (c) in .

[0089] Therefore, the present invention adopts the above-mentioned method for calculating the distance between conductors by integrating regular component extraction and cluster growth, uses a deep learning model to accurately extract point clouds of cross-arm structure facilities and complex power facilities in point cloud data, and carefully analyzes the conductor type and its spatial distribution through classification and clustering algorithms, thereby realizing accurate calculation of the distance between power conductors and equipment, providing an efficient and accurate solution for intelligent detection and automated maintenance of power transmission systems, greatly improving the accuracy of facility monitoring and the degree of automation of maintenance, and thus solving the problems of high difficulty and inability to accurately extract conductors and substation facilities in existing methods.

[0090] 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 wire distance calculation method integrating regular component extraction and cluster growth, characterized in that: The following steps are involved: S1. Obtaining substation point cloud data, and performing noise and outlier removal and ground point cloud culling processing on the point cloud data to obtain remaining point cloud data including substation facilities and various conductors; S2. Use the deep learning model to extract components with regular structures from the remaining point cloud data, and preliminarily mark and identify various types of conductors and the substation facilities connected to them; S3. Classify various types of conductors based on the extracted regular structural components, and cut out the point cloud area where the substation facilities are located according to the spatial directions of the classified conductors; S4, based on step S3, applying a cluster growth algorithm to further refine and extract the substation facility point cloud connected to the conductor; S5. After the extraction and classification of substation facilities and conductors, the conductor spacing, the distance between the conductors and the ground, and the distance between the conductors and substation facilities are calculated, and an analysis report is generated for the operation, maintenance and monitoring of substation transmission lines.

2. The wire distance calculation method integrating regular component extraction and cluster growth according to claim 1 is characterized in that: In step S1, the point cloud data is subjected to noise and outlier removal and ground point cloud culling processing, including: Select a statistical filter to determine the outliers by calculating the distribution of neighboring points of each point, and set the point cloud set P = {p i |i=1,2,...,N},each point p i In three-dimensional space, the coordinates are (x i ,y i , z i ), select each point p i The k nearest points of the point p are calculated. i The average distance d to the k nearest points i : Among them, d i For point p i The average Euclidean distance between its k nearest neighbors; By counting the average distance distribution of all points, the mean μ and standard deviation σ are calculated, and d i Points that exceed the following distance threshold conditions are considered as outliers and filtered out: d i >m+as; Among them, α is a user-defined parameter ranging from 1 to 3, so that outliers are filtered out of the point cloud in the overall scene of the substation area by statistical outlier filtering; Then, the RANSAC random sampling consensus algorithm is used to fit and filter out the ground points, and three non-collinear points are randomly selected to form a candidate plane. The maximum inner point set is iteratively obtained to form a fitting plane, thereby identifying the ground plane and retaining the height information of the ground points.

3. The wire distance calculation method integrating regular component extraction and cluster growth according to claim 1 is characterized in that: In step S2, it specifically includes: first, using a voxel-based Transformer network model to process the input point cloud data, voxelize the point cloud data, and form a fixed-size grid; secondly, using a multi-scale feature extraction strategy to extract local features of the point cloud in the voxel and output features of dimensions (H, C), where H is the number of voxels and C is the feature dimension corresponding to the voxel; then, using the Transformer network to extract context features between voxels in the vertical and horizontal directions, obtain corresponding feature representations through the QKV weight matrix, and perform self-attention mechanism calculations to output voxel features of size (H, C1); resampling the output voxel features again to restore them to point-by-point features, and connecting them with the classification task head to complete point-by-point point cloud classification; finally, preliminarily marking and identifying the various types of wires to be extracted and the substation facilities connected to them.

4. The wire distance calculation method integrating regular component extraction and cluster growth according to claim 1 is characterized in that: In step S3, it specifically includes: The input point cloud data is voxelized, that is, the three-dimensional space is divided into a uniform voxel grid, and the size of each voxel is defined as V x ×V y ×V z , where V x 、V y 、V z Represent the size of the voxel in the X, Y and Z directions respectively, and use the formula to convert each point p of the point cloud into i =(x i ,y i ,z i ) is mapped into a voxel grid: Among them, v x , v y , v z Represents the corresponding voxel index; using the height Z value information of the regular component as the index, quickly cluster the wire voxels of the corresponding height in the adjacent neighborhood, and after clustering, divide the long-distance transmission line point cloud voxels, and extract the corresponding non-substation facility area point cloud; The divided long-distance transmission lines are removed, and then the spatial direction analysis is performed on the soft wires connected to the substation facilities to determine the projection direction of the soft wires on the XY plane. The entire point cloud is cropped on the XY plane to obtain the point cloud area of ​​all substation facilities within the Z-axis height range.

5. The wire distance calculation method integrating regular component extraction and cluster growth according to claim 1 is characterized in that: In step S4, it specifically includes: Based on the cropped point cloud area obtained in step S3, the initial cluster center is determined by the local density peak clustering method, and the given point cloud data set P = {p i |i=1,2,...,N} defines two key indicators for each point: local density ρ i , distance δ i , where the local density ρ i Calculated by the following formula: Among them, d ij Represents point p i and point p j The Euclidean distance between them is then calculated by constructing a similarity matrix and sorting the distances in descending order, selecting the top 2% as the cutoff distance d c , is the indicator function, if d ij <d c , on the contrary Thus, the areas with lower local density are identified as the initial cluster centers; After determining the initial cluster center, the geometric information of the wire is used for constraint, and the centerline position and direction vector of the wire are calculated, where the direction vector is denoted as v = (v x , v y , v z ); In the cluster growth process, by calculating the point p i With the neighboring point p j The relative position vector u ij To control the cluster growth direction: u ij =(x i -x j ,y i -y j ,z i -z j ); And by calculating the angle θ between the wire direction v and the relative position vector: Set the angle threshold θ max To ensure that the cluster grows only along the wire direction; After the initial cluster center is determined and the geometric constraints are completed, the cluster growth will gradually expand under the constraints of local density and geometric direction. The cluster growth process is expressed as: C(p j )=C(p i )if v.u ij ≥cosθ max and ρ j <ρ threshold ; Among them, C(p i ) is point p i The cluster to which it belongs, C(p j ) is point p j The cluster to which it will be assigned; Finally, the point cloud of substation facilities connected to the conductors is further refined and extracted through density peaks and geometric constraints.

6. The wire distance calculation method integrating regular component extraction and cluster growth according to claim 1 is characterized in that: In step S5, it specifically includes: firstly, regularizing the projection of the extracted conductors and substation facilities respectively, projecting the extracted conductors along the XY plane, calculating the spacing distance between the conductors with each assigned ID, and obtaining the spatial distribution relationship between different conductors; then projecting the conductors along the Z-axis direction to obtain the distance between the conductors and the substation facilities and the distance between the conductors and the ground point.

Images