Overhead transmission line airborne point cloud classification method, system and device and medium

Abstract

The invention relates to the technical field of intelligent inspection of electric power facilities, in particular to an overhead transmission line airborne point cloud classification method, system and device and a medium, point cloud data of a transmission line are collected, key features are extracted by adopting a statistical method, and data integrity is optimized in combination with a subspace extension strategy and an elevation statistical histogram; constructing a geometric surface by using a mathematical method to carry out directional filtering and curved surface fitting, and filtering terrain interference and unrelated vegetation point clouds; based on a particle swarm optimization algorithm, adaptively searching linearity, curvature and elevation threshold values, and realizing coarse classification of the transmission conductor and the tower; a subspace feature extension strategy is combined with a clustering analysis algorithm, horizontal extension tower data are dynamically discriminated, a vertical extension threshold value is optimized, and fine classification is completed; the problems of incomplete point cloud extraction, complex scene misjudgment and insufficient small sample classification precision are solved, and technical support is provided for intelligent inspection of the overhead transmission line.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power facility intelligent inspection, and in particular to an overhead power transmission line airborne point cloud classification method, system, device and medium. BACKGROUND

[0002] At present, airborne laser radar technology has become an important means of intelligent inspection of overhead power transmission lines due to its efficient three-dimensional data acquisition capability. The existing method mainly uses elevation statistical analysis and unsupervised clustering algorithm to realize preliminary extraction of power line point cloud, which significantly improves the efficiency and safety of traditional manual inspection. However, the existing technology has inherent defects in the elevation threshold segmentation process, especially in complex terrain areas, which easily leads to discontinuity of power transmission line point cloud data, and needs to rely on subsequent manual intervention for data completion, which seriously restricts the level of full-process automation.

[0003] On the other hand, the existing classification algorithm has high dependence on large-scale labeled data sets, while in actual engineering, the overhead power transmission line point cloud samples are scarce and unevenly distributed, resulting in low model training efficiency and insufficient generalization ability. Traditional methods need to repeatedly adjust multiple feature threshold parameters such as curvature and linearity when dealing with small sample scenarios, which not only consumes huge computing resources, but also is difficult to effectively solve the misclassification problem of the connection between towers and conductors. These problems have become the main obstacles to the large-scale application of intelligent inspection technology of power transmission lines. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an overhead power transmission line airborne point cloud classification method and system to solve the problems of discontinuity caused by point cloud rupture in complex terrain, difficulty in parameter optimization in small sample scenarios, misclassification in the connection area between towers and conductors, and excessive consumption of computing resources.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides an overhead power transmission line airborne point cloud classification method, comprising:

[0008] Collecting relevant point cloud data, using statistical methods to process and extract data, and retaining relevant features of the data;

[0009] Performing directional processing on the extracted data, and using mathematical methods to construct geometric surfaces;

[0010] Using the constructed geometric surfaces for fitting, and filtering through threshold values;

[0011] Using algorithms to adaptively find the required features, and realizing rough classification;

[0012] The fine classification is realized through a design strategy, and a data processing scheme is provided.

[0013] As a preferred scheme of the overhead transmission line airborne point cloud classification method, the related point cloud data is collected, the data is extracted and processed using a statistical method, and the data related features are retained, including:

[0014] The space is segmented through a space strategy, and the distribution characteristics are analyzed in combination with a statistical chart.

[0015] The point cloud data is determined and distinguished by setting a threshold value.

[0016] The classification range is dynamically adjusted by using a space strategy, and the key features are retained.

[0017] Through the synergistic effect of the space segmentation strategy and the statistical feature analysis, the distribution recognition ability of the point cloud data is effectively optimized; in combination with the dynamic threshold value for determining and distinguishing the key point cloud and the non-target interference data, the integrity and accuracy of the feature extraction are significantly improved; by using the space range self-adaptive adjustment mechanism, the classification stability under different terrains and complex scenes is ensured, so that the core features of the transmission line are retained while the environmental adaptability and reliability of the classification algorithm are enhanced.

[0018] As a preferred scheme of the overhead transmission line airborne point cloud classification method, the data extracted is subjected to directional processing, and a geometric surface is constructed using a mathematical method, including:

[0019] The impurity data is removed through directional processing by an algorithm.

[0020] A mathematical method is used to construct a covering surface geometric model.

[0021] The point cloud data is optimized based on the distribution characteristics.

[0022] Through directional processing, noise and outlier data in the point cloud are effectively removed, and data quality is improved; in combination with the geometric curved surface constructed by a mathematical method, the spatial structural features of the transmission line are accurately represented, providing a reliable basic model for subsequent classification; at the same time, the optimization strategy based on the distribution characteristics further strengthens the spatial distribution rationality of the point cloud data, significantly enhances the robustness of feature extraction in complex scenes, and finally realizes the synergistic improvement of classification accuracy and data processing efficiency.

[0023] As a preferred scheme of the overhead transmission line airborne point cloud classification method, the geometric surface constructed is used for fitting, and a threshold value is used for filtering, including:

[0024] The geometric surface constructed by a mathematical method is used to fit the target point cloud data, and invalid data is filtered out.

[0025] Based on the position characteristics, a threshold value is set to filter out irrelevant point cloud data.

[0026] By constructing a geometric surface by a mathematical method and setting a threshold value based on a position feature, the spatial distribution characteristics of the target point cloud can be accurately fitted, invalid data caused by terrain undulations or scanning interference can be effectively filtered out, and irrelevant point clouds such as vegetation can be removed through a dynamic threshold value, so that the integrity of the power transmission conductor and tower data is significantly improved. The spatial constraint capability of the point cloud data in a complex scene is strengthened, high-purity and high-consistency input data are provided for subsequent classification, and therefore the classification accuracy and algorithm robustness are overall improved.

[0027] As a preferred scheme of the overhead transmission line airborne point cloud classification method, the algorithm is used to adaptively find the required features and realize rough classification, including:

[0028] The algorithm is used to adaptively determine the related features to realize preliminary classification;

[0029] The algorithm is used to optimize parameters of point cloud data features, perform feature calculation, and complete rough classification;

[0030] Based on adaptive matching of dynamic threshold values and feature parameters, the classification accuracy is solved.

[0031] As a preferred scheme of the overhead transmission line airborne point cloud classification method, the algorithm is used to adaptively determine the related features, including:

[0032] The optimal threshold value parameters of the linearity, elevation and curvature features of the power transmission conductor are dynamically determined through an adaptive optimization algorithm;

[0033] The principal component analysis features of the neighborhood points and the particle swarm optimization strategy are combined to realize automatic preliminary classification of the power transmission conductor and tower point clouds.

[0034] As a preferred scheme of the overhead transmission line airborne point cloud classification method, the algorithm is used to adaptively determine the related features, including:

[0035] Through a subspace feature expansion strategy combined with a vertical extension threshold value judgment, the power transmission conductor and tower point clouds after rough classification are classified;

[0036] A clustering analysis algorithm is used to detect the horizontally extended tower data, and the classification attribution is dynamically adjusted through comparison of the point cloud coordinates in the subspace with the minimum coordinate difference;

[0037] Based on the subspace extension strategy and geometric feature judgment, the power transmission conductor and tower point clouds are separated.

[0038] In a second aspect, the present application provides an overhead transmission line airborne point cloud classification system, including:

[0039] Point cloud preprocessing module, collect relevant point cloud data, use statistical method to process and extract data, and retain data related features;

[0040] Geometric modeling module, directional processing of the extracted data, and mathematical method for constructing geometric surface;

[0041] Surface filtering module, fitting using the constructed geometric surface, and filtering through threshold value;

[0042] Parameter optimization module, using algorithm to adaptively find the required features, and realizing coarse classification;

[0043] Hierarchical decision module, realizing fine classification through designed strategy, and providing data processing scheme.

[0044] In a third aspect, the present application provides an electronic device, comprising:

[0045] Memory and processor;

[0046] The memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions, which realize the steps of the overhead transmission line airborne point cloud classification method.

[0047] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer executable instructions, which realize the steps of the overhead transmission line airborne point cloud classification method when executed by a processor.

[0048] Compared with the prior art, the present application has the following beneficial effects: the present application effectively optimizes the boundary accuracy of coarse classification results by combining the subspace feature expansion strategy with the vertical extension threshold value discrimination, significantly reduces the misjudgment of transmission conductor and tower point cloud; through the clustering analysis algorithm for detecting horizontal extension tower data and dynamically adjusting the classification attribution, the robustness of point cloud classification in complex scenes is enhanced, and the classification confusion problem caused by occlusion or dense area is solved; based on the deep fusion of the subspace extension strategy and the geometric feature discrimination, the high-precision separation of transmission conductor and tower point cloud is realized, and the data integrity and classification reliability are ensured. Overall, the method improves the classification efficiency and generalization ability in small sample environment, and provides efficient and stable data processing support for intelligent inspection of overhead transmission lines. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 Flowchart of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application.

[0051] Figure 2 Selected aerial power transmission line point cloud graph of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application.

[0052] Figure 3 Filtered point cloud graph of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application using the original local range point elevation statistical histogram method.

[0053] Figure 4 Filtered point cloud graph of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application using the local range point elevation statistical histogram method with added subspace extension strategy.

[0054] Figure 5 Roughly classified point cloud graph of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application.

[0055] Figure 6 Precisely classified point cloud result graph of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application using the subspace extension strategy.

[0056] Figure 7 Point cloud graph classified by combining DNSCAN and Mean Shift algorithms of the aerial power transmission line airborne point cloud classification method according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] To make the above objectives, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0058] Embodiment 1, refer to Figure 1 According to an embodiment of the present application, an aerial power transmission line airborne point cloud classification method is provided, comprising:

[0059] S1: Collect relevant point cloud data, use statistical methods to process and extract data, and retain data-related features;

[0060] S2: Perform directional processing on the extracted data, and use mathematical methods to construct geometric surfaces;

[0061] S3: fitting using the constructed geometric surface and filtering by threshold value;

[0062] S4: finding required features using an algorithm and realizing coarse classification;

[0063] S5: realizing fine classification by designing a strategy and providing a data processing scheme.

[0064] It should be noted that in the prior art, there are problems such as incomplete point cloud extraction of overhead transmission lines, insufficient classification accuracy in a small sample environment, and high misjudgment rate in a complex scene.

[0065] Therefore, in view of the above problems of poor point cloud data integrity, fuzzy classification boundary, and environmental interference leading to classification confusion, through steps S1-S5, key features are retained by using a subspace extension strategy and an elevation statistical histogram, data space distribution is optimized by directional geometric filtering and surface fitting, coarse classification efficiency is improved by adaptive threshold search, and fine classification is realized based on feature expansion and a dynamic discrimination strategy, thereby significantly improving classification accuracy and scene adaptability on the basis of ensuring data integrity.

[0066] Embodiment 2, with reference to Figures 1-7 For an embodiment of the present application, based on the above embodiment, an overhead transmission line airborne point cloud classification method is provided.

[0067] In the present application, in step S1, the relevant point cloud data is collected, and the data is processed and extracted using a statistical method. Through the subspace extension strategy combined with the elevation statistical histogram analysis of the local range points, the point cloud subspace is dynamically divided and the elevation distribution features are counted. According to the preset elevation interval threshold, the high-elevation and low-elevation interval point clouds are screened, and the key point cloud data of the transmission conductor and the tower are effectively distinguished and retained. By optimizing the subspace size and the elevation interval criterion, the feature extraction deviation problem caused by the sparse or dense point clouds in the terrain undulating area is solved, ensuring the high integrity of the key data in the subsequent classification process and providing a reliable input basis for fine classification.

[0068] In an alternative embodiment, in step S1, the relevant point cloud data is collected, and the data is processed and extracted using a statistical method. The fixed-size subspace division combined with global elevation distribution statistics can also be used to set a unified elevation interval threshold to screen high and low elevation point clouds, retain preliminary feature data of the transmission conductor and the tower, and optimize the directional features of the point cloud by principal component analysis to improve the stability of the elevation determination in the subspace.

[0069] In another alternative embodiment, the relevant point cloud data collected in step S1 is processed using statistical methods to extract data, and a sliding window strategy can be used to dynamically adjust the coverage of the subspace, combined with point cloud density weighted elevation distribution histogram analysis, to adaptively match the terrain undulation characteristics, filter high and low elevation interval point clouds, and introduce curvature to preliminarily distinguish the local morphological differences between the conductor and the tower, thereby providing multi-dimensional feature support for subsequent classification.

[0070] In the embodiments of the present application, the relevant point cloud data collected in step S1 is processed using statistical methods to extract data and retain data-related features, and further includes:

[0071] The local range point elevation statistical histogram method with the addition of the subspace extension strategy is used to extract high-elevation and low-elevation interval threshold point clouds. The overhead power transmission line point cloud is subspace-ized, and the size of the subspace will affect the extraction effect of the power line point cloud. If the size is too large, part of the power transmission conductor will be classified as a non-power transmission conductor, especially in areas with large terrain undulations. If the size is too small, the number of point clouds in each subspace is too small to statistically analyze the elevation distribution characteristics.

[0072] After determining the size of the subspace, the elevation distribution in the subspace needs to be statistically analyzed at a certain interval. The size of the interval will affect the elevation discontinuity of the point cloud. A high-elevation interval threshold H1 is set, which represents the elevation difference between the power transmission conductor and the ground point. The merged elevation interval value is subjected to threshold judgment: if H≤H1, all points in the subspace are determined to be non-power transmission conductor points; if H>H1, the minimum elevation value M1 of the elevation interval is recorded; the z-coordinate value M of all points in the subspace is subjected to threshold judgment: if M>M1, the point is a power transmission conductor point; otherwise, it is a non-power transmission conductor point.

[0073] The z-coordinate value D of the lowest point in the subspace is recorded, and the difference ΔD between the z-coordinate value D1 of the point in the subspace and the z-coordinate value D of the lowest point is calculated. The difference ΔD is subjected to threshold judgment, and e is a set extension threshold. If ΔD≤e, the point is a power transmission conductor point; otherwise, it is a non-power transmission conductor point.

[0074] A large-elevation interval threshold H x and a small-elevation interval threshold H n are set, respectively, to obtain a large-elevation interval threshold point cloud A and a small-elevation interval threshold point cloud B.

[0075] In the embodiment of the present application, the extracted data is processed in step S2, and the geometric surface is constructed by using a mathematical method. The extracted point cloud data is processed by using a density-based spatial clustering algorithm to remove outliers and determine the main direction to optimize the spatial distribution. Then, the point cloud is divided into multiple subspaces, the lowest elevation reference points in each subspace are extracted, the Delaunay triangulation algorithm is used to construct a continuous geometric surface, and the spatial topological structure of the power transmission line and the tower is accurately described. Finally, combined with the surface fitting and the vertical space threshold filtering, the noise points outside the surface and the interference data of the vegetation below are filtered out to ensure the spatial consistency and integrity of the subsequent classified data, thereby laying a foundation for the refined classification.

[0076] In an alternative embodiment, the extracted data is processed in step S2, and the geometric surface is constructed by using a mathematical method. The mean shift clustering algorithm can be used to locate the main direction and remove outliers of the point cloud. Then, the dynamic subspaces are divided based on the region growing method, the elevation distribution feature points in each subspace are extracted, and the local geometric surface is constructed by using the constrained triangulation algorithm. Combined with the curvature analysis of the surface, the unstructured noise points are filtered out to enhance the expression accuracy of the spatial topology.

[0077] In another alternative embodiment, the extracted data is processed in step S2, and the geometric surface is constructed by using a mathematical method. The hierarchical clustering algorithm can be used to identify the density distribution characteristics of the point cloud, and the principal component analysis is used to determine the optimal subspace division direction. The non-uniform B-spline surface fitting technology is used to reconstruct the geometric surface, and the multi-scale space projection strategy is used to screen the point cloud data inside and outside the surface to improve the fitting degree of the geometric surface to the real structure of the power transmission line.

[0078] In the embodiment of the present application, the extracted data is processed in step S2, and the geometric surface is constructed by using a mathematical method, which further includes:

[0079] The large-elevation-threshold point cloud is directionally geometrically filtered to remove outliers, and a geometric surface is constructed by using the Delaunay triangulation algorithm.

[0080] The large-elevation-threshold point cloud is clustered by using a density-based spatial clustering algorithm with noise application (DBSCAN) to ensure the integrity of the main direction clustering and locate the main direction. The core formula of the DBSCAN clustering algorithm is represented as:

[0081]

[0082] Wherein, A is a large-elevation-interval-threshold point cloud data set, p is a given point, q is a point in the point cloud data set A, and dist(p, q) represents the distance between the points p and q.

[0083] The point cloud data after directional geometric filtering is divided into many subspaces, the point with the lowest z coordinate in each subspace is extracted, and a surface containing all transmission conductor data and part of tower data is constructed by using Delaunay triangulation;

[0084] For a triangle formed by three points b(x1, y1), c(x2, y2), and d(x3, y3), and a(x, y) being any other point in the point cloud A with a large elevation interval threshold, the formula for judging whether a triangle belongs to Delaunay triangulation is:

[0085]

[0086] If det(E) > 0, the point a is not in the circumcircle of the triangle determined by points b, c, and d, which means that the triangle bcd is a Delaunay triangle.

[0087] In the embodiment of the application, the geometric surface constructed in step S3 is used for fitting, and filtering is performed through a threshold. The small-elevation-threshold point cloud data is spatially fitted based on the geometric surface generated by Delaunay triangulation, and the effective point cloud in the vertical space (including a specific range above and below) of the surface is extracted. By setting a position feature threshold, vegetation point cloud irrelevant to the transmission line below the surface is dynamically filtered out, and the interference of terrain undulations or scanning noise is eliminated. This step effectively retains the key point cloud data of the transmission conductor and tower by combining geometric constraints and threshold screening, reduces the influence of invalid information on subsequent classification, provides a high-purity and high-consistency data basis for coarse classification and fine classification, and significantly improves the accuracy and environmental adaptability of the classification process.

[0088] In an alternative embodiment, the geometric surface constructed in step S3 is used for fitting, and filtering is performed through a threshold. A continuous geometric surface can also be generated by a local polynomial surface fitting algorithm, the small-elevation-threshold point cloud is subjected to spatial constraint analysis, and the effective point cloud data in a specific extension range in the normal direction of the surface is extracted. At the same time, a dynamic threshold is set in combination with the regional density distribution characteristics to filter out low-density discrete points irrelevant to the transmission line and tower below the surface, and further suppress the interference of terrain noise on the classification process.

[0089] In another alternative embodiment, the geometric surface constructed in step S3 is used for fitting, and filtering is performed through a threshold. A multi-level spatial grid division model can also be constructed to divide the geometric surface into a plurality of sub-grid units. According to the point cloud distribution characteristics and elevation gradient variation law in each unit, the vertical space filtering threshold is adaptively adjusted to filter out vegetation and ground scattered point clouds, retain the topological structure consistency of the key point cloud of the transmission line, and ensure the stability of the subsequent classification input.

[0090] In the embodiment of the present application, the fitting in step S3 uses the constructed geometric surface and is filtered by a threshold, and further includes:

[0091] The small-elevation-threshold point cloud data is filtered out of the vertical space by surface fitting, and the vegetation point cloud under the surface is filtered out by a threshold.

[0092] According to the coordinate information of the eight outermost points of the surface, the corresponding points of the small-elevation-threshold point cloud data are found, and the fitting of the small-elevation-threshold point cloud data is completed by the surface. Then, the points in the vertical space of the surface are extracted, that is, the points in each subspace located directly above the surface and having a z coordinate greater than the highest point of the surface, and the points in each subspace located directly below the surface and having a z coordinate less than the lowest point of the surface.

[0093] After the points in the vertical space of the surface are extracted, since there are still some vegetation point cloud data under the surface, a threshold is set to filter out the vegetation point cloud under the surface, that is, the vegetation point cloud in each subspace having an x coordinate within the range [x min ,x max ] and located below the surface is deleted.

[0094] In the embodiment of the present application, the algorithm is used to adaptively find the required features in step S4, and a rough classification is realized, including:

[0095] The particle swarm optimization algorithm (PSO) is used to adaptively find the optimal threshold of the linearity, elevation information and curvature of the power transmission conductor;

[0096] For each point, its k nearest neighbors are found, the neighborhood of each point is traversed, and the covariance matrix H is calculated for each neighborhood using principal component analysis (PCA), and the covariance matrix H is decomposed into eigenvalues, as follows:

[0097] H = UΛU T

[0098] wherein Λ = diag(λ0, λ1, λ2), and the eigenvalues satisfy λ0≤ λ1≤ λ2.

[0099] The formula is used to calculate the curvature C e of each neighborhood, the formula is used to calculate the linearity L y of each neighborhood, and the vertical height difference Z of each neighborhood is obtained by calculating the difference between the maximum value Z max and the minimum value Z min of the z coordinates of the neighborhood points.

[0100] The threshold search ranges of the curvature, linearity and vertical height difference are respectively set, and the PSO algorithm is used to adaptively find the optimal thresholds of the curvature, linearity and vertical height difference, so as to preliminarily realize the coarse classification of the power transmission conductor and the tower.

[0101] In the embodiment of the application, the fine classification is realized by the design strategy in step S5, and a data processing scheme is provided, which comprises the following steps:

[0102] The fine classification of the power transmission conductor and the tower is realized by the subspace extension strategy; the z coordinate threshold z d = 58.5 is set, and it is judged whether the tower data of each subspace with a z coordinate greater than the threshold z d is extended downward, the difference between the z coordinate of each point in the current subspace and the minimum z coordinate of the subspace is calculated; if the difference is less than the vertical extension threshold h1 = 0.4, the point is reclassified as a power transmission conductor, otherwise, the original classification is retained.

[0103] The tower data with an x coordinate outside the range [x min , x max ] is screened out; in the vicinity of the tower, the power transmission conductor data and the tower data are mixed together, and a local dense area may be formed due to occlusion or scanning angle, so the DBSCAN clustering algorithm needs to be used to detect the horizontally extended tower; the difference between the z coordinate of each tower point in the current subspace and the minimum z coordinate of the subspace is calculated; if the difference is less than the vertical extension threshold h1 = 0.5, the point is reclassified as a power transmission conductor, otherwise, the original classification is retained.

[0104] The above algorithm evaluates its performance by using a classification index, and the evaluation index of the classification algorithm is the recall rate, and the definition of the recall rate is as follows:

[0105]

[0106] Wherein, N c N c represents the number of point clouds correctly classified as the target category; N t N t represents the true number of point clouds of the target category; the higher the value of the index is, the better, and the best result is 1.

[0107] The basic flow of the airborne overhead power transmission line laser point cloud classification method proposed in the embodiment is as shown in Figure 1 ; the point cloud data of the overhead power transmission line scanned and collected by a laser radar system carried by a certain unmanned aerial vehicle is selected; the selected overhead power transmission line contains three-dimensional coordinate information of 2585074 points, contains 2 tower data and 5 power transmission conductor data, as shown in Figure 2 .

[0108] In order to visually show the results of the two point cloud filtering methods, the results of the two point cloud filtering methods are visualized. In order to compare fairly, the size of the subspace is set to 2 meters, and the elevation interval threshold is set to 2 meters and 6 meters respectively. Red represents overhead power line point cloud, and gray represents vegetation point cloud. Figure 3 is the point cloud filtered using the original local range point based elevation histogram method. It can be observed that in the position where the right power conductor changes gently, there are many power conductor point clouds that are not correctly classified. Figure 4 is the point cloud filtered using the local range point based elevation histogram method with the added subspace extension strategy. Although there are more vegetation point clouds misclassified as overhead power line point clouds than the original filtering algorithm, the improved elevation histogram method can preserve power conductor and tower point clouds as much as possible.

[0109] The small elevation threshold point cloud data after filtering uses the linearity, elevation information and curvature information of the power conductor and the PSO algorithm to adaptively find the optimal threshold of linearity, elevation information and curvature of the power conductor, and preliminarily realizes the coarse classification of power conductor point cloud and tower point cloud. As shown in Figure 5 , red represents power conductor points, and green represents tower points. Although most of the point cloud data is correctly classified, there are still a small number of power conductor point cloud data misjudged as tower point cloud. Near the tower, there are many tower points misjudged as power conductor points, so further discrimination of the misclassified results is needed to achieve more accurate classification.

[0110] After fine classification, the point cloud basically realizes the correct classification of most points. As shown in Figure 6 , red represents power conductor points, and green represents tower points. The power conductor point cloud and tower point cloud after fine classification have obvious boundaries, and most of the misclassified point clouds in coarse classification are also correctly classified.

[0111] In order to further verify the effectiveness of the proposed method, the proposed method and the clustering algorithm combining DBSCAN and Mean Shift are compared, Figure 7 is the point cloud classified using the clustering algorithm combining DBSCAN and Mean Shift. Red represents power conductor points, green represents tower points, and black represents unclassified points. Both methods use the point cloud filtered using the local range point based elevation histogram method with the added subspace extension strategy, which contains 34692 points. In order to quantitatively analyze the accuracy of overhead power line point cloud separation and extraction, the results of power conductor point cloud and tower point cloud extracted manually using third-party point cloud data processing software are used as reference. By comparing these results, the accuracy of power conductor point cloud and tower point cloud is calculated, which is used as the main standard for evaluation.

[0112] To sum up, the application provides an overhead power transmission line airborne point cloud classification method, which dynamically divides point cloud subspaces by combining subspace extension strategy with elevation statistical histogram, accurately extracts key feature data of power transmission conductors and towers, filters out terrain noise and vegetation interference point clouds by using directional geometric filtering and Delaunay triangulation to construct geometric curved surfaces, ensures the integrity and consistency of data space distribution, adaptively searches linear degree, curvature and elevation threshold based on particle swarm optimization algorithm, realizes rough classification of power transmission conductors and towers, further dynamically adjusts classification attribution through subspace feature expansion strategy and vertical extension threshold judgment, and completes fine classification. The method effectively solves the problems of incomplete point cloud extraction, misjudgment in complex scenes and insufficient classification accuracy of small samples, significantly improves the classification boundary accuracy, environmental adaptability and data reliability, provides efficient and stable technical support for intelligent inspection of overhead power transmission lines, and promotes the intelligent and accurate development of power facility operation and maintenance.

[0113] In the embodiment 3, the above is a schematic scheme of an overhead power transmission line airborne point cloud classification method. It should be noted that the technical scheme of the overhead power transmission line airborne point cloud classification system belongs to the same concept as the technical scheme of the overhead power transmission line airborne point cloud classification method described above. The technical scheme of the overhead power transmission line airborne point cloud classification system in this embodiment is not described in detail, and the description of the technical scheme of the overhead power transmission line airborne point cloud classification method can be referred to.

[0114] The embodiment also provides an overhead power transmission line airborne point cloud classification system, which comprises:

[0115] A point cloud preprocessing module collects relevant point cloud data, processes and extracts the data using a statistical method, and retains relevant features of the data;

[0116] A geometric modeling module performs directional processing on the extracted data and constructs a geometric surface using a mathematical method;

[0117] A curved surface filtering module uses the constructed geometric surface for fitting and filtering through a threshold value;

[0118] A parameter optimization module uses an algorithm to adaptively find the required features and realize rough classification;

[0119] A hierarchical decision module realizes fine classification by designing a strategy and provides a data processing scheme.

[0120] The embodiment also provides an electronic device suitable for the case of overhead power transmission line airborne point cloud classification, which comprises a memory and a processor. The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the overhead power transmission line airborne point cloud classification method proposed in the above embodiment.

[0121] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for classifying the point cloud of the overhead power transmission line.

[0122] The storage medium proposed by the embodiment belongs to the same inventive concept as the method for classifying the point cloud of the overhead power transmission line proposed by the above embodiment, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH memory, a hard disk, or an optical disc, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of various embodiments of the present application.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for classifying airborne point clouds of overhead transmission lines, characterized in that: include: Collect relevant point cloud data, use statistical methods to process and extract data, and retain relevant features of the data; Directional processing is performed on the extracted data, and geometric surfaces are constructed using mathematical methods; Use the constructed geometric surface for fitting and filter by threshold; Use algorithms to adaptively find required features and achieve coarse classification; Achieve precise classification through design strategies and provide data processing solutions.

2. The method for classifying airborne point clouds of overhead power transmission lines according to claim 1, wherein: Collect relevant point cloud data, use statistical methods to process and extract data, and retain relevant data features, including: Use spatial strategies to segment the space and analyze distribution characteristics using statistical graphs; By setting the threshold, the point cloud data is judged and distinguished; A spatial strategy is used to dynamically adjust the classification range and retain key features.

3. The method for classifying airborne point clouds of overhead power transmission lines according to claim 2, wherein: Directional processing is performed on the extracted data, and geometric surfaces are constructed using mathematical methods, including: Perform directional processing through algorithms to remove impure data; Use mathematical methods to construct the coverage surface geometric model; Optimize point cloud data based on distribution characteristics.

4. The method for classifying airborne point clouds of overhead power transmission lines according to claim 3, wherein: Fitting is done using constructed geometric faces and filtering is done using thresholds, including: The geometric surface constructed by mathematical methods is used to fit the target point cloud data and filter out invalid data; Based on location features, a threshold is set to filter out irrelevant point cloud data.

5. The method for classifying airborne point clouds of overhead power transmission lines according to claim 4, wherein: Use algorithms to adaptively find required features and implement rough classification, including: Adopting algorithms to adaptively determine relevant features and achieve preliminary classification; The algorithm is used to optimize the parameters of the point cloud data features, perform feature calculations, and complete rough classification; Based on the adaptive matching of dynamic threshold and feature parameters, the classification accuracy is improved.

6. The method for classifying airborne point clouds of overhead power transmission lines according to claim 5, wherein: The method of adaptively determining relevant features using an algorithm includes: Dynamically determine the optimal threshold parameters for the linearity, elevation, and curvature characteristics of transmission lines through an adaptive optimization algorithm; Combining the principal component analysis characteristics of neighborhood points with particle swarm optimization strategy, the automatic preliminary classification of transmission line and tower point clouds is achieved.

7. The method for classifying airborne point clouds of overhead power transmission lines according to claim 6, wherein: Achieve refined classification through design strategies and provide data processing solutions, including: The subspace feature expansion strategy is combined with the vertical extension threshold discrimination to finely classify the transmission line and tower point clouds after coarse classification. A cluster analysis algorithm is used to detect horizontally extended tower data, and the classification is dynamically adjusted by comparing the point cloud coordinates in the subspace with the minimum coordinate difference; Based on the subspace extension strategy and geometric feature discrimination, the point cloud of transmission lines and towers is separated.

8. An airborne point cloud classification system for overhead transmission lines, applying the method according to any one of claims 1 to 7, characterized in that: include: Point cloud preprocessing module collects relevant point cloud data, uses statistical methods to process and extract data, and retains data-related features; The geometric modeling module performs directional processing on the extracted data and constructs geometric surfaces using mathematical methods; Surface filtering module, which uses the constructed geometric surface for fitting and filters it through the threshold; The parameter optimization module uses an algorithm to adaptively find the required features and achieve coarse classification; The hierarchical decision-making module achieves precise classification through design strategies and provides data processing solutions.

9. An electronic device comprising: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the overhead transmission line airborne point cloud classification method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the overhead transmission line airborne point cloud classification method according to any one of claims 1 to 7.

Images