An overhead line point cloud detection method based on catenary surface hough transformation

Abstract

The application discloses a kind of overhead line point cloud detection methods based on catenary surface hough transformation, comprising the following steps: step one, data initialization processing, directional bounding box algorithm is adopted in initial point cloud data, and the size of point cloud coordinate value is reduced without changing point cloud data model;Step two, the filtering of ground point, adopts cloth filtering algorithm to filter ground point cloud;Step three, the extraction of power corridor point, the threshold value is set to the safety distance existing between power corridor and power corridor or between power corridor and other ground object point, and DBSCAN clustering is adopted, to select power corridor point cloud data with the feature that the distance of transmission line across is significantly greater than other ground object point;The application provides a kind of overhead line point cloud detection methods based on catenary surface hough transformation, solves the extraction of overhead line point cloud under the condition of large ground fluctuation and complex line trend, avoids the case that overhead line point cloud is mixed with interval rod point cloud and insulator point cloud in certain conditions.

Description

Technical Field

[0001] This invention relates to a method for extracting and reconstructing laser point clouds of overhead lines, and in particular to a method for detecting point clouds of overhead lines based on catenary Hough transform. Background Technology

[0002] As the country with the world's largest power grid, and with social development and economic growth, electricity consumption is constantly increasing, posing greater challenges to the stable operation of the power grid. Transmission lines, as a crucial component of the power grid system, are vital for the safety and stability of people's lives and production. Therefore, monitoring their operational status and assessing risks have always been key focuses of power grid management. Traditional inspection methods involve manual ground inspections. While this method can directly obtain real-time information about transmission lines, it suffers from drawbacks such as high inspection costs, low accuracy, low efficiency, and inconvenience in areas with complex natural environments. With the increasing number of ultra-high voltage, extra-high voltage, high-capacity, and long-distance power transmission lines, existing transmission line corridor resources are dwindling. Transmission lines are increasingly located further away from cities and major transportation routes, and the geographical and climatic environments along these corridors are becoming increasingly complex. As the application of high-resolution aerial sensors in aerial photography gains importance, drones are increasingly being used for power line inspections, significantly improving inspection efficiency. However, processing aerial data collected by drones in complex scenarios suffers from low accuracy, making it difficult to identify insulators, power lines, and towers in conditions such as heavy rain and fog. In recent years, airborne lidar measurement technology has developed rapidly and has been widely applied in power line inspections. Airborne lidar scans the power line corridor by emitting laser pulses, acquiring target surface images with multi-echo laser reflection data. Using specific algorithms, combined with GPS location coordinates, POS system laser direction information, and simultaneously captured high-resolution visible light photographs, high-precision, high-density three-dimensional laser point cloud data is generated. Inspection personnel can use the constructed power line model to analyze potential safety hazards in the line, locate obstacles in the power corridor, and prevent accidents. The electric power line can be analyzed by establishing a three-dimensional model of the electric power line.

[0003] Existing methods for extracting power line point clouds mostly combine the spatial characteristics of power lines. These methods generally include a preprocessing step: ground point cloud filtering and feature point cloud filtering. Ground point cloud filtering often utilizes elevation distribution features, segmenting the ground and feature point clouds using a given elevation threshold. This method is only effective in plains areas; in complex terrains such as hills and mountains, it often suffers from incomplete filtering of ground and feature point clouds. There are various methods for filtering feature point clouds, including those based on the angle of echo information carried in the point cloud data, density clustering, and feature values ​​from point cloud SVD decomposition. These methods are prone to causing loss of power line point clouds and incomplete filtering of other non-power line point clouds, such as those of split rods and insulators. Hough transform is often used in solving equations for straight lines. This method can effectively avoid the above-mentioned problems. Existing studies only combine the feature that the projection of the power line on the ground is a straight line with Hough transform to separate the points of each overhead line. This method only considers the straight line feature of the overhead line projected in the XZ plane and ignores the catenary feature projected in the XY plane. In the process of extracting overhead lines, other noise points will inevitably appear.

[0004] Single-strand, single-core conductors are generally made of steel-cored aluminum stranded wire, which has a complex structure. The mechanical and physical characteristics of steel-cored aluminum stranded wire are studied in detail. The electric field line can be described by the catenary model in theoretical mechanics. The parabola can be regarded as an approximate expression of the catenary model. Since the complexity of the catenary equation is much higher than that of other equations, the study often uses the parabolic equation or polynomial equation to approximate the catenary for three-dimensional reconstruction of electric field lines. Summary of the Invention

[0005] The technical problem this invention aims to solve is to provide an overhead line point cloud detection method based on the Hough transform of the catenary plane. This method addresses the challenges of point cloud extraction under conditions of significant ground undulations and complex line routes. Under certain circumstances, it avoids the inclusion of spacer rod and insulator point clouds in the overhead line point cloud, providing a more accurate data source for subsequent scientific research on overhead lines. Furthermore, by employing the spatial catenary equation combining the catenary plane and its profile, the acquisition of the catenary point cloud no longer relies on manual parameter setting, reducing labor costs and automating the extraction process.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for detecting point clouds of overhead lines based on catenary Hough transform, comprising the following steps:

[0007] Step 1: Data initialization processing. In the initial point cloud data, the directed bounding box algorithm is used to reduce the size of the point cloud coordinate values ​​without changing the point cloud data model.

[0008] Step 2: Filtering out ground points. A cloth filtering algorithm is used to filter out ground point clouds.

[0009] Step 3: Extraction of power corridor points. Thresholds are set for the safe distance between power corridors or between power corridors and other ground features, and DBSCAN clustering is used to filter power corridor point cloud data based on the feature that the cross distance of the transmission line is significantly greater than that of other ground features.

[0010] Step 4: Coarse extraction of power line points. In the 3D point cloud data, the PCA principal component analysis algorithm based on SVD decomposition of the covariance matrix is ​​used to reduce the dimensionality of the original data. Linear filtering is used to set a threshold to achieve coarse extraction of overhead line point clouds.

[0011] Step 5: Project the coarsely extracted power line point cloud onto the BEV plane. For cases where the power lines appear as straight lines from the BEV perspective, perform hough line detection in the 2D projected view to obtain the straight line equation of the conductor projection.

[0012] Step 6: Solve for the catenary profile of each conductor in 3D space based on the straight line equation of each electric field line extracted from the BEV plane.

[0013] Step 7: Convert the coarse point cloud of the overhead line to polar coordinates, perform spatial hough transformation based on the catenary parameter equation, and detect the existing catenary surfaces.

[0014] Step 8: Solve for the intersection lines between each catenary profile and the detected catenary surface, and classify the point clouds of adjacent overhead lines using KD-Tree;

[0015] Step nine: Verify whether the obtained catenary equation is the actual equation of the overhead line.

[0016] Preferably, the bounding box in step one is an AABB bounding box, which has all edges parallel to the initial coordinate axes. Its definition principle is as follows:

[0017] x min ≤x≤x max

[0018] y min ≤y≤y max

[0019] z min ≤z≤z max

[0020] Where x min ,ymin ,z min x is the minimum value existing in the initial data along the X, Y, Z axes. max ,y max ,z max The initial data represents the maximum value on the X, Y, and Z axes, and the transformed data points lie between these extreme points.

[0021] Preferably, in step two, a fabric ground filtering method is adopted to simulate fabric falling to achieve the filtering of point clouds on non-flat ground.

[0022] Preferably, in step three, the point cloud of the power corridor is extracted using the DBSCAN clustering method. DBSCAN clustering requires specifying two parameters:

[0023] Epsilon(ε): The maximum radius of the region. Data points that are less than or equal to the specified epsilon will belong to the same category.

[0024] MinPts: The neighborhood of the number of minpts within a radius of a region is considered a cluster.

[0025] Preferably, in step five, after projecting the point cloud onto the BEV plane and performing a Hough transform to detect straight lines, the lines are represented parametrically as follows:

[0026] r = xcos(θ) + ysin(θ)

[0027] Where r is the distance from the origin to the line along a vector perpendicular to the line, and θ is the angle between the X-axis and the vector in degrees. By performing a Hough transformation and substituting the data points, the values ​​of r and θ are returned and formed into a parameter space matrix. The equation of the line is determined by selecting the set of r and θ values ​​that appear most frequently in the space matrix.

[0028] Preferably, in step six, the equations of the line and the BEV plane are:

[0029] A0x+B0y+C0z+D0=0

[0030]

[0031] The formula for calculating the equation of the catenary profile is:

[0032]

[0033]

[0034]

[0035] k1(x-x0)+k2(y-y0)+k3(z-z0)=0

[0036] in, Let be the coefficient vector of the equations perpendicular to the plane. Let be the direction vector of the straight line through which the catenary passes. This represents the first element of the product of vectors n and s. This represents the second element of the product of vectors n and s. The third element represents the product of vectors n and s, and (x0, y0, z0) is the coordinate of a point through which the line passes in space.

[0037] Preferably, in step seven, the parametric equation of the catenary is:

[0038]

[0039] The equation is expressed in cylindrical coordinates, -∞ < u < +∞, 0 ≤ v ≤ 2π.

[0040] Preferably, the catenary equation is subjected to a Hough transform to return the spatial parameter matrix of the catenary, the most frequently occurring catenary parameter is obtained, and a threshold is set to use KD-Tree to determine the affiliation of neighboring point clouds.

[0041] This invention provides a method for detecting point clouds of overhead lines based on catenary Hough transform, which has the following beneficial effects:

[0042] 1. This invention constructs a new spatial coordinate system for point cloud data, which reduces computational complexity and increases the speed of computer computation.

[0043] 2. The ground filtering method adopted in this invention is not sensitive to ground undulations. Assuming a piece of cloth falls from above under gravity, the cloth that falls can represent the current terrain and can adapt to ground filtering in various environments to a certain extent.

[0044] 3. This invention uses a spatial catenary combined with a catenary profile to solve the spatial equation of overhead lines, and uses Hough transformation to perform a voting inverse solution for the given parameters. Three judgment conditions are set to determine the accuracy of the results, and the results are closer to the actual overhead line equations.

[0045] 4. The method proposed in this invention for extracting overhead line point clouds based on the catenary spatial equation achieves automated extraction of overhead line point clouds solely by relying on the coordinates of the point cloud in the data, without requiring any other known parameters. Attached Figure Description

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0047] Figure 1 This is a schematic diagram of the structure of the present invention;

[0048] Figure 2 A schematic diagram illustrating the principle of coordinate transformation using the bounding box algorithm;

[0049] Figure 3 This diagram illustrates the relationship between the overhead line, the catenary plane, and the catenary cross-section.

[0050] Figure 4 A visualization of the proposed catenary surface;

[0051] Figure 5 This is a schematic diagram of the intersection of the catenary plane and the catenary line section. Detailed Implementation

[0052] like Figure 1 As shown, an overhead line point cloud detection method based on catenary Hough transform includes the following steps:

[0053] Step 1: Data initialization processing. In the initial point cloud data, the directed bounding box algorithm is used to reduce the size of the point cloud coordinate values ​​without changing the point cloud data model.

[0054] Step 2: Filtering out ground points. A cloth filtering algorithm is used to filter out ground point clouds.

[0055] Step 3: Extraction of power corridor points. Thresholds are set for the safe distance between power corridors or between power corridors and other ground features, and DBSCAN clustering is used to filter power corridor point cloud data based on the feature that the cross distance of the transmission line is significantly greater than that of other ground features.

[0056] Step 4: Coarse extraction of power line points. In the 3D point cloud data, the PCA principal component analysis algorithm based on SVD decomposition of the covariance matrix is ​​used to reduce the dimensionality of the original data. Linear filtering is used to set a threshold to achieve coarse extraction of overhead line point clouds.

[0057] Step 5: Project the coarsely extracted power line point cloud onto the BEV plane. For cases where the power lines appear as straight lines from the BEV perspective, perform hough line detection in the 2D projected view to obtain the straight line equation of the conductor projection.

[0058] Step 6: Solve for the catenary profile of each conductor in 3D space based on the straight line equation of each electric field line extracted from the BEV plane.

[0059] Step 7: Convert the coarse point cloud of the overhead line to polar coordinates, perform spatial hough transformation based on the catenary parameter equation, and detect the existing catenary surfaces.

[0060] Step 8: Solve for the intersection lines between each catenary profile and the detected catenary surface, and classify the point clouds of adjacent overhead lines using KD-Tree;

[0061] Step nine: Verify whether the obtained catenary equation is the actual equation of the overhead line.

[0062] Since most point cloud data uses the WGS 84 (World Geodetic System) coordinate system, which uses the Earth's central mass as the origin and comprises a reference ellipsoid, standard coordinate system, height data, and geoid, it contains a larger amount of coordinate data. This can easily lead to larger errors when fitting equations to it. Furthermore, computers tend to cancel out relatively small values ​​in the same equations during calculation. To avoid this phenomenon, this method employs the Axis-aligned bounding box (AABB) algorithm for coordinate system transformation of the point cloud data. An AABB bounding box encloses an object in a cuboid, with each face of the cuboid parallel to a specific coordinate axis plane. The bounding box is determined by recording the minimum and maximum values ​​representing the bounding box on each coordinate axis. Data points processed by the bounding box algorithm must meet the following conditions:

[0063] x min ≤x≤x max

[0064] y min ≤y≤y max

[0065] z min ≤z≤z max

[0066] Where x min ,y min ,z min x is the minimum value existing in the initial data along the X, Y, Z axes. max ,y max ,z max The initial data points represent the maximum values ​​on the X, Y, and Z axes. The transformed data points lie between these extreme points. (See attached image.) Figure 2 As shown, the AABB bounding box algorithm is used to transform the coordinate system by taking one vertex as the origin of the point cloud coordinate system.

[0067] Preferably, in step two, a fabric ground filtering method is adopted to simulate fabric falling to achieve the filtering of point clouds on non-flat ground.

[0068] The initial point cloud data contains a large number of points, including towers, overhead lines, buildings, trees, and ground features. Ground point cloud data is the most extensive, often accounting for half or more of the entire point cloud data. In our study of overhead lines, we first need to remove ground point cloud data to increase the speed of computer calculations and reduce errors in equation acquisition. This method uses cloth filtering to filter out ground point clouds. Traditional airborne point cloud filtering algorithms mostly consider differences in slope and elevation changes to distinguish between ground features and ground points. The cloth filtering algorithm, however, uses a completely new approach. First, the point cloud is flipped. Then, assuming a piece of cloth falls from above under gravity, the falling cloth represents the current terrain. The cloth filtering defines a basic formula:

[0069]

[0070] These respectively represent the external driving factors (gravity, collision, etc.) and internal driving factors (internal connection factors between particles) of the "cloth".

[0071] The specific implementation process is as follows:

[0072] 1) Initial removal of noise points and inversion of point cloud data.

[0073] 2) Initialize the fabric grid and set its size. The initial "fabric" position is usually above the highest point.

[0074] 3) Project all LiDAR points and grid particles onto the same horizontal plane, find the nearest neighbor of each particle, and record its elevation before projection.

[0075] 4) For each movable grid "particle", calculate its displacement caused by gravity and compare it with the IHV of the current particle's corresponding cp point. If the particle's height is lower than or equal to IHV, set the particle's height to IHV and set it as an immovable point.

[0076] 5) For each grid "particle", calculate its displacement caused by the influence of internal driving factors.

[0077] 6) Repeat steps 4) and 5) above until the maximum high commitment change of all particles is small enough or the number of iterations reaches the user's preset threshold, then stop the simulation process.

[0078] 7) Calculate the height difference between the LiDAR point cloud and the grid particles to distinguish between ground points and non-ground points. If the distance between the LiDAR point and the simulated particle is less than the preset threshold hcc, it is considered a ground point; otherwise, it is considered a non-ground point.

[0079] The point cloud data after ground filtering will contain point clouds of power corridors to be studied, power corridors not to be studied, buildings, and other ground features. We need to extract the power corridors to be studied, which means filtering out the point clouds of other ground features and power corridors not to be studied. Here, the traditional DBSCAN clustering algorithm is used. Density-based noise applied spatial clustering (DBSCAN) is an unsupervised ML clustering algorithm. Unsupervised means that it does not use pre-labeled targets to cluster data points. Clustering refers to attempting to group similar data points into manually determined groups or clusters. Compared with traditional K-means and hierarchical clustering algorithms, it is better suited for extracting power corridors. The DBSCAN clustering algorithm divides data points into three categories: core points, boundary points, and noise points. The DBSCAN clustering algorithm defines the maximum radius Epsilon and the minimum point minpts, which are specifically represented as follows:

[0080] Epsilon(ε): The maximum radius of the region. Data points that are less than or equal to the specified epsilon will belong to the same category.

[0081] MinPts: The neighborhood of the number of minpts within a radius of a region is considered a cluster.

[0082] Based on the safety distance specified in the national standard for overhead transmission line design, appropriate maximum radius Epsilon and minimum point minpts are selected to extract the power corridor under study.

[0083] The point cloud of a power corridor contains towers, insulators, spacers, and overhead lines. The overhead line point cloud is characterized by its elongated, linear extension and length significantly greater than its width. On elevation planes containing towers, the tower point cloud is far larger than the overhead line point cloud, posing a significant challenge to subsequent overhead line equation extraction. To address this, a PCA algorithm based on SVD covariance matrix decomposition is employed to reduce the number of tower point clouds. PCA (Principal Component Analysis) is used for point cloud preprocessing, plane detection, normal vector solving, dimensionality reduction, classification, and decompression (dimensionality increase). PCA is used to classify points in the point cloud, such as ground points, wall points, and points on objects, and is a data dimensionality reduction technique used for data preprocessing. PCA projects a 3D object onto a surface to identify its dominant orientation. The surface selection criterion is to maximize the variance of the point distribution; a more dispersed surface with higher variance contains more information. In the point cloud covariance eigenvalues ​​of a straight line feature, one eigenvalue is always significantly larger than the other two. The direction of the line is given by the eigenvector corresponding to the eigenvalue with the largest point cloud covariance. By solving for the mean, centering, covariance matrix, and SVD decomposition of the point cloud data, the n eigenvalues ​​of matrix A are obtained and sorted in descending order. The formula for SVD decomposition is:

[0084] A=W∑W -1

[0085] Where W is an n*n dimensional matrix composed of n eigenvectors, and ∑ is an n*n dimensional matrix with these n eigenvalues ​​as the main diagonal. The final result includes point clouds of overhead lines, some spacer point clouds, and some tower point clouds, collectively referred to as the overhead line coarse point cloud.

[0086] Preferably, in step five, after projecting the point cloud onto the BEV plane and performing a Hough transform to detect straight lines, the lines are represented parametrically as follows:

[0087] r = xcos(θ) + ysin(θ)

[0088] Where r is the distance from the origin to the line along a vector perpendicular to the line, and θ is the angle between the X-axis and the vector in degrees. By performing a Hough transformation and substituting the data points, the values ​​of r and θ are returned and formed into a parameter space matrix. The equation of the line is determined by selecting the set of r and θ values ​​that appear most frequently in the space matrix.

[0089] Because the span of an overhead transmission line is much larger than the cross-sectional dimension of the overhead line—meaning the length of the entire span is much greater than its diameter—and because overhead lines are mostly made of stranded wires composed of multiple thin metal strands, the rigidity of the overhead line has little impact on the shape of its suspension curve. Therefore, it can be assumed that the overhead line is a flexible chain without rigidity. Based on this assumption, the overhead line can only withstand tensile force and not bending moment; furthermore, it can be assumed that the load acting on the overhead line is uniformly distributed along its length. Based on these two assumptions, the overhead line suspended between two towers has a catenary shape. The derivation shows that the general integral form of the catenary equation for the overhead line is:

[0090]

[0091] Where C1 and C2 are integration constants, their values ​​depending on the position of the origin of the coordinate system. σ0 is the axial stress at the lowest point of the sag. γ is the uniformly distributed load acting along the length of the overhead line. Let... The equation can then be written as:

[0092]

[0093] The derivation of the overhead catenary model is performed in a two-dimensional plane; therefore, it needs to be transformed into a spatial coordinate system to describe the mathematical model of the overhead catenary point cloud. A spatial curve can be represented as the intersection of a curved surface and a plane. Based on solid geometry, it can be deduced that a catenary can be represented by the intersection of the curved surface (catenary surface) formed by rotating the catenary around its horizontal directrix and the catenary profile. The coarse point cloud of the overhead catenary is projected onto the BEV plane (the projection of the point cloud onto a plane perpendicular to the height direction). Utilizing the property that the overhead catenary point cloud is a straight line on the BEV plane, and using the ideal condition that the overhead catenary profile is a plane passing through its projected straight line and perpendicular to the BEV plane, the profile equation of each overhead catenary is obtained.

[0094] The point cloud projected onto the BEV plane includes partial projection point clouds of the towers, spacers, and the required overhead lines. A Hough transform is applied in the plane, and the n most-voted line equations are selected through a voting process, where n is the number of overhead lines. The core idea of ​​the Hough transform here is:

[0095] Initial equations:

[0096] y=kx+b

[0097] After Hough transform:

[0098] b = -xk + y

[0099] By repeatedly selecting two random points (x1, y1) and (x2, y2) and substituting them into the equation after Hough transform, the values ​​of k and b are calculated. The groups with the same values ​​of k and b are counted, and the group with the most frequent occurrences is the optimal expression parameter for the desired equation. This process is repeated until the profile equation of the BEV projection for each overhead line is obtained. The relationship between the catenary profile, the BEV projection line, and the catenary is shown in the attached figure. Figure 3 As shown.

[0100] The equations of the straight line and the BEV plane are:

[0101] A0x+B0y+C0z+D0=0

[0102]

[0103] The formula for calculating the equation of the catenary profile is:

[0104]

[0105]

[0106]

[0107] k1(x-x0)+k2(y-y0)+k3(z-z0)=0

[0108] in, Let be the coefficient vector of the equations perpendicular to the plane. Let be the direction vector of the straight line through which the catenary passes. This represents the first element of the product of vectors n and s. This represents the second element of the product of vectors n and s. The third element represents the product of vectors n and s, and (x0, y0, z0) is the coordinate of a point through which the line passes in space.

[0109] The formula for calculating the vector product is:

[0110]

[0111] Preferably, in step seven, the parametric equation of the catenary is:

[0112]

[0113] The equation is expressed in cylindrical coordinates, -∞ < u < +∞, 0 ≤ v ≤ 2π.

[0114] The mathematical equation for a catenary in space must satisfy the following conditions:

[0115] Condition 1: All points are on this catenary profile;

[0116] Condition 2: All points are on the catenary surface formed after the catenary is rotated.

[0117] Condition 3: The intersection of the catenary profile and the catenary plane should include two parts with upward openings and downward openings, and the overhead line catenary line should open upwards in the Cartesian coordinate system.

[0118] Based on the catenary parameter model, a Hough transform strategy is applied to the coarse point cloud of the overhead line to estimate the catenary model, satisfying condition one. The intersection line between the catenary and the catenary profile is then solved, as shown in the attached figure. Figure 5 As shown, condition two is satisfied. A discrimination mechanism is added to the intersection of the catenary plane and the catenary profile; the equation of the upward-opening curve satisfies condition three, and is the desired catenary model for the overhead line.

[0119] Preferably, the catenary equation is subjected to a Hough transform to return the spatial parameter matrix of the catenary, the most frequently occurring catenary parameter is obtained, and a threshold is set to use KD-Tree to determine the affiliation of neighboring point clouds.

[0120] The point clouds of adjacent overhead lines are classified according to the equation of a single catenary. The obtained catenary model in polar coordinates is converted into a Cartesian coordinate system model. A reasonable threshold is set, and a KD-tree index is established for the points on the single catenary. The neighboring point clouds within a distance less than the Euclidean distance R are selected as the point clouds of that overhead line. The point clouds of each overhead line are obtained in this way.

[0121] KD-Tree, or K-dimensional tree for short, is a spatial partitioning data structure that is often used for searching in high-dimensional spaces. Here, we use KD-Tree for nearest neighbor search.

[0122] Point clouds are data sets composed of points, possessing rich geometric and depth information. They exhibit irregular structure and disordered data arrangement. Traditional point cloud feature extraction methods mostly rely on information such as curvature, normal vector, and density of local point clouds, combined with methods such as Gaussian models, support vector machines, and random forests, to manually describe and extract point cloud features.

[0123] Extracting power line point clouds requires understanding the characteristics of power lines and their distribution. The spatial characteristics of power lines can be summarized as follows: they extend linearly in a narrow shape with a length much greater than their width; there are no other obvious ground features nearby; individual power lines are parallel to each other and maintain a certain distance; power line points are discontinuous in the vertical direction but closely connected in the horizontal direction with a certain density. The length and sag of overhead power lines are functions of their specific load and stress. When meteorological conditions change, these parameters also change. Temperature fluctuations cause thermal expansion and contraction of overhead power lines, resulting in corresponding changes in length, sag, and stress. Strong winds and icing increase the specific load and stress of overhead power lines, leading to an increase in line length due to elastic deformation. The equations revealing the relationships between parameters of an overhead power line under varying weather conditions are called the overhead power line state equations. These equations can be determined using a power line point cloud model. Currently, most methods for 3D reconstruction of overhead power lines require initial extraction of the overhead power line point cloud. After extraction, the 3D model equations of individual power lines are obtained by projecting them onto the extracted overhead power lines. (See attached image) Figure 1 As shown, this method involves continuously assigning values ​​to points based on the given state equation expression of the overhead line under 3D projection and then voting to select the optimal parameters of the overhead line state equation. Due to the special geographical location of overhead lines, such as in mountainous or hilly areas, and the interference from trees, houses, and other industrial buildings, this method improves the extraction step of overhead line point clouds in complex terrain during the process of obtaining the overhead line equation.

[0124] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for detecting point clouds of overhead power lines based on catenary Hough transform, characterized in that, Includes the following steps: Step 1: Data initialization processing. In the initial point cloud data, the directed bounding box algorithm is used to reduce the size of the point cloud coordinate values ​​without changing the point cloud data model. Step 2: Filtering out ground points. A cloth filtering algorithm is used to filter out ground point clouds. Step 3: Extraction of power corridor points. Thresholds are set for the safe distance between power corridors or between power corridors and other ground features, and DBSCAN clustering is used to filter power corridor point cloud data based on the feature that the cross distance of the transmission line is significantly greater than that of other ground features. Step 4: Coarse extraction of power line points. In the 3D point cloud data, the PCA principal component analysis algorithm based on SVD decomposition of the covariance matrix is ​​used to reduce the dimensionality of the original data. Linear filtering is used to set a threshold to achieve coarse extraction of overhead line point clouds. Step 5: Project the coarsely extracted overhead line point cloud onto the BEV plane. For cases where the power lines appear as straight lines from the BEV perspective, perform hough line detection in the projected 2D view to obtain the straight line equations of each conductor projection. Step 6: Solve the catenary profile of each conductor in 3D space based on the straight line equation of each overhead line extracted from the BEV plane. Step 7: Convert the coarse point cloud of the overhead line to polar coordinates, perform spatial hough transformation based on the catenary parameter equation, and detect the existing catenary surfaces. Step 8: Solve for the intersection lines between each catenary profile and the detected catenary surface, and classify the point clouds of adjacent overhead lines using KD-Tree; Step nine: Verify whether the obtained catenary equation is the actual equation of the overhead line.

2. The overhead line point cloud detection method based on catenary Hough transform according to claim 1, characterized in that, In step one, the bounding box has all edges parallel to the initial coordinate axes, and its definition principle is as follows: x min ≤x≤x max and min ≤y≤y max With min ≤z≤z max Where x min ,y min ,z min x is the minimum value existing in the initial data along the X, Y, Z axes. max ,y max ,z max The initial data represents the maximum value on the X, Y, and Z axes, and the transformed data points lie between these extreme points.

3. The overhead line point cloud detection method based on catenary Hough transform according to claim 1, characterized in that, In step five, the point cloud is projected onto the BEV plane and a Hough transform is performed to detect straight lines. The lines are then represented parametrically as follows: r = xcos(θ) + ysin(θ) Where r is the distance from the origin to the line along a vector perpendicular to the line, and θ is the angle between the X-axis and the vector in degrees. By performing a Hough transformation and substituting the data points, the values ​​of r and θ are returned and formed into a parameter space matrix. The equation of the line is determined by selecting the set of r and θ values ​​that appear most frequently in the space matrix.

4. The overhead line point cloud detection method based on catenary Hough transform according to claim 1, characterized in that, In step six, the equations of the line and the BEV plane are: A0x+B0y+C0z+D0=0 The formula for calculating the equation of the catenary profile is: k1(x-x0)+k2(y-y0)+k3(z-z0)=0 in, Let be the coefficient vector of the equations perpendicular to the plane. Let be the direction vector of the straight line through which the catenary passes. This represents the first element of the product of vectors n and s. This represents the second element of the product of vectors n and s. The third element represents the product of vectors n and s, and (x0, y0, z0) is the coordinate of a point through which the line passes in space.

5. The overhead line point cloud detection method based on catenary Hough transform according to claim 1, characterized in that, In step seven, the parametric equations for the catenary are: The equation is expressed in cylindrical coordinates, -∞ <u∞+<,0≤v≤2π。 6. The overhead line point cloud detection method based on catenary Hough transform according to claim 5, characterized in that, The catenary equation is subjected to a Hough transform to return the spatial parameter matrix of the catenary. The catenary parameter with the most occurrences is obtained, and a threshold is set to use KD-Tree to determine the affiliation of neighboring point clouds.

Images