Power transmission line 3D point cloud form segmentation method, device and system and storage medium

Abstract

The invention discloses a power transmission line 3D point cloud form segmentation method, device and system and a storage medium, and relates to the technical field of point cloud segmentation. The method comprises the following steps: acquiring original point cloud data and preprocessing the original point cloud data to obtain a three-dimensional point cloud space; extracting candidate tower poles based on geometric structure constraint conditions of the tower poles, performing coarse filtering to obtain coarse filtering tower poles, and performing instance segmentation on the coarse filtering tower poles based on a point cloud clustering algorithm to obtain segmented tower poles; carrying out fine filtering on the segmented tower pole to obtain a fine filtered tower pole, extracting an external connection line, obtaining a candidate line point cloud, carrying out straight line fitting and classification, and obtaining external connection lines in the same direction; diffusion is carried out in the growth direction of the external connection lines in the same direction until the lines coincide to obtain inter-tower lines; and curve fitting and parameterization representation are carried out on the lines between the towers, and 3D point cloud form segmentation of the power transmission line is completed. According to the embodiment of the invention, manual intervention is reduced, and the modeling efficiency and precision are improved.

Description

Technical Field

[0001] The present application belongs to the field of point cloud segmentation technology, and in particular, relates to a method, device, system and storage medium for 3D point cloud morphology segmentation of a power transmission line. Background Art

[0002] Tower line extraction and segmentation is absolutely necessary for large-scale autonomous inspections, and it also plays a key role in aerial inspection robot systems. Power towers are important power line infrastructure. The detection of power towers is the premise of automatic detection. Its core task is to identify and locate problems and determine the category and location information of power towers. At the same time, in order to ensure the construction efficiency of power facilities, the inspection of power towers can provide services for power companies to inspect the distribution of power towers.

[0003] With the rapid development of smart grids, the mileage of power lines and power equipment has increased rapidly. The contradiction between too many maintenance equipment and too few maintenance personnel has become increasingly prominent. The traditional manual inspection method cannot fully meet the inspection requirements of the development of the modern power industry, which brings great difficulties to the safety, management and maintenance of the transmission network, which will increase the risk of failure of the transmission system.

[0004] Point cloud segmentation solutions for power transmission lines mainly include two categories, based on deep learning and morphological geometry methods. Deep learning methods rely on massive amounts of data to train models, which makes data processing difficult, and model calculations rely on graphics cards, which are difficult to meet with lightweight edge computing boards on drones. The morphological geometry method only requires clear tower line features for segmentation, but the current geometric morphological transmission line segmentation algorithms at home and abroad are slow and cannot meet the requirements of real-time transmission line inspections by drones. Summary of the invention

[0005] The purpose of this application is to provide a method, device, system and storage medium for 3D point cloud morphological segmentation of transmission lines, so as to realize accurate three-dimensional modeling and analysis of transmission lines to support subsequent line maintenance, fault detection, risk assessment and other applications, while improving work efficiency and accuracy.

[0006] In order to achieve the above objectives, the solution of this application is:

[0007] In a first aspect, an embodiment of the present application provides a method for 3D point cloud morphology segmentation of a power transmission line, comprising:

[0008] The original point cloud data of the target area is obtained and preprocessed. The preprocessing includes intensity filtering. By performing intensity filtering on the original point cloud data of the target area, a three-dimensional point cloud space of the target area is obtained;

[0009] Extract candidate tower poles in the three-dimensional point cloud space based on the geometric structure constraints of the tower poles, perform rough filtering on the candidate tower poles to obtain rough filtered tower poles, and perform instance segmentation on the rough filtered tower poles based on the point cloud clustering algorithm to obtain segmented tower poles;

[0010] Fine filtering is performed on the segmented tower poles to obtain fine filtered tower poles, external lines in the fine filtered tower poles are extracted based on the point cloud clustering algorithm to obtain candidate line point clouds, straight line fitting is performed on the candidate line point clouds and classified to obtain external lines in the same direction;

[0011] Diffusion is performed in the growth direction of the external line in the same direction, and the point cloud of the external line in the same direction is extracted until it coincides with the external line of another tower pole to obtain the inter-tower line;

[0012] The lines between towers are curve fitted and parameterized to complete the 3D point cloud morphological segmentation of the transmission line.

[0013] According to the above method of the embodiment of the present application, the following additional technical features may also be provided:

[0014] Furthermore, the preprocessing also includes altitude restriction filtering, and by performing intensity filtering and altitude restriction filtering on the original point cloud data of the target area, a three-dimensional point cloud space of the target area excluding the ground is obtained.

[0015] Furthermore, the geometric constraints of the tower include:

[0016] The tower is continuous in elevation;

[0017] The cross-sectional area at the base of the tower is greater than that at the top;

[0018] There is at least one connecting line at the edge of the tower.

[0019] Furthermore, the calculation method of the cross-sectional area includes:

[0020] All point clouds within a 5-cm error range 5 meters upward from the bottom of the tower and all point clouds within a 5-cm error range 5 meters downward from the top of the tower are extracted to obtain the cross-sectional point cloud of the bottom of the tower and the cross-sectional point cloud of the top of the tower; based on the cross-sectional point cloud of the bottom of the tower and the cross-sectional point cloud of the top of the tower, the cross-sectional area of ​​the bottom of the tower and the cross-sectional area of ​​the top of the tower are calculated by the Gift Wrapping algorithm.

[0021] Furthermore, based on the point cloud clustering algorithm, the coarse filtered tower poles are segmented to obtain segmented tower poles, including:

[0022] Obtain the point cloud of the coarsely filtered tower pole, and use any point in it as a seed point, calculate the Euclidean distance from the seed point to all other points, and if the Euclidean distance between two points is less than a first preset threshold, they are considered to belong to the same cluster;

[0023] Add all points whose Euclidean distance to the seed point is less than the first preset threshold to the same cluster, and recursively perform the process until no new point satisfies the condition that the Euclidean distance to the seed point is less than the first preset threshold;

[0024] When a point in a cluster no longer meets the condition that the Euclidean distance with the seed point is less than the first preset threshold, the cluster is regarded as a new cluster, and a new seed point is selected for repeated calculation until all points in the point cloud of the coarsely filtered tower are assigned to a cluster, and a segmented tower is obtained.

[0025] Furthermore, the segmented tower poles are finely filtered to obtain finely filtered tower poles, and external lines in the finely filtered tower poles are extracted based on the point cloud clustering algorithm to obtain candidate line point clouds. Linear fitting and classification are performed on the candidate line point clouds to obtain external lines in the same direction, including:

[0026] By judging whether there is an external line in the split tower rod, the split tower rod is finely filtered. If there is an external line in the split tower rod, it is filtered as a fine filtering tower rod. If there is no external line in the split tower rod, it is not filtered as a fine filtering tower rod. By fine filtering the split tower rod, the false tower rods in the split tower rod are eliminated.

[0027] Extract the external lines in the fine-filtered tower poles through Euclidean clustering to obtain a candidate line point cloud; perform straight line fitting on the candidate line point cloud through the RANSAC algorithm to obtain a fitting line, calculate the variance from the candidate line point cloud to the fitting line, and if the variance is higher than a second preset threshold, remove the candidate line point cloud;

[0028] The candidate line point clouds are classified according to their slope and overlap rate, and the candidate line point clouds with the same slope and overlap rate greater than 90% are classified as a circumscribed line in the same direction.

[0029] Furthermore, curve fitting and parameterization are performed on the tower lines to complete the 3D point cloud morphological segmentation of the transmission line, including:

[0030] The inter-tower line is parameterized by quadratic curve fitting, and the quadratic curve fitting method includes the least square method.

[0031] In a second aspect, an embodiment of the present application provides a transmission line 3D point cloud morphology segmentation device, comprising:

[0032] A data acquisition module is configured to acquire the original point cloud data of the target area and perform preprocessing, wherein the preprocessing includes intensity filtering, and the three-dimensional point cloud space of the target area is obtained by performing intensity filtering on the original point cloud data of the target area;

[0033] A filtering and segmentation module is configured to extract candidate tower poles in a three-dimensional point cloud space based on geometric structural constraints of the tower poles, perform rough filtering on the candidate tower poles to obtain rough filtered tower poles, and perform instance segmentation on the rough filtered tower poles based on a point cloud clustering algorithm to obtain segmented tower poles;

[0034] The filtering and fitting module is configured to perform fine filtering on the segmented tower poles to obtain fine filtered tower poles, extract external lines in the fine filtered tower poles based on a point cloud clustering algorithm to obtain candidate line point clouds, perform straight line fitting on the candidate line point clouds and classify them to obtain external lines in the same direction;

[0035] The diffusion extraction module is configured to diffuse in the growth direction of the external lines in the same direction and extract the point cloud of the external lines in the same direction until they coincide with the external lines of another tower pole to obtain the inter-tower lines;

[0036] The parameterization module is configured to perform curve fitting and parameterization representation on the inter-tower lines and complete the 3D point cloud morphological segmentation of the transmission line.

[0037] In a third aspect, an embodiment of the present application provides a transmission line 3D point cloud morphological segmentation system, the system comprising a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the transmission line 3D point cloud morphological segmentation method provided in the first aspect of the embodiment of the present application.

[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the transmission line 3D point cloud morphological segmentation method as described in the first aspect of the embodiment of the present application.

[0039] Compared with the prior art, the 3D point cloud morphological segmentation method for power transmission lines provided by the embodiment of the present application has the following beneficial technical effects:

[0040] The embodiment of the present application effectively removes noise and irrelevant information in the original point cloud data, such as ground point cloud, through preprocessing steps such as intensity filtering and altitude limit filtering, thereby obtaining a purer and more accurate three-dimensional point cloud space of the target area; based on the geometric structure constraints of the tower, candidate towers are extracted from the three-dimensional point cloud space, and through a series of steps such as coarse filtering, instance segmentation, and fine filtering, the tower part of the transmission line and its external wiring are accurately identified, thereby reducing manual intervention and improving the efficiency and accuracy of modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram showing a flow chart of a method for segmenting a 3D point cloud morphology of a power transmission line according to an embodiment of the present application is shown;

[0042] Figure 2 A structural block diagram of a transmission line 3D point cloud morphological segmentation device according to an embodiment of the present application is shown;

[0043] Figure 3 A structural block diagram of a computer device according to an embodiment of the present application is shown;

[0044] Figure 4 A schematic diagram of a three-dimensional point cloud space of a target area in an embodiment of the present application is shown;

[0045] Figure 5 A schematic diagram of a split tower pole after coloring in an embodiment of the present application is shown;

[0046] Figure 6 A schematic diagram of external wiring after coloring in an embodiment of the present application is shown;

[0047] Figure 7 A schematic diagram of a three-dimensional point cloud space model according to an embodiment of the present application is shown;

[0048] Figure 8 A schematic diagram showing the comparison between the segmentation effect of the embodiment of the present application and the original effect diagram of the model is shown;

[0049] in, Figure 8 (a) to Figure 8 (e) are schematic diagrams of the original effects of the models from Center01 to Center05. Figure 8 (f) to Figure 8 (j) are schematic diagrams of the segmentation effects from Center01 to Center05. Figure 8 (k) to Figure 8 (o) are schematic diagrams of the original model effects of Seg01 to Seg05, Figure 8 (p) to Figure 8 (t) Schematic diagrams of the segmentation effects from Seg01 to Seg05. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some structures related to the present application are shown in the accompanying drawings, rather than all structures. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0051] The terms "including" and "having" and any variations thereof in this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0052] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0053] like Figure 1 As shown, the embodiment of the present application provides a method for 3D point cloud morphology segmentation of a power transmission line, comprising the following steps:

[0054] Step 101, obtaining the original point cloud data of the target area and performing preprocessing, wherein the preprocessing includes intensity filtering, and obtaining the three-dimensional point cloud space of the target area by performing intensity filtering on the original point cloud data of the target area.

[0055] This step is used to obtain the original point cloud data of the target area and perform preprocessing, which is the basis for subsequent processing. Usually, the original point cloud data of the target area (such as the transmission line area) is obtained through LiDAR, drone photogrammetry or other 3D scanning technologies. These data contain a large number of 3D coordinate points, each of which represents a physical location in the target area.

[0056] Preprocessing is to remove noise and irrelevant information from the original point cloud data to obtain a purer and more accurate three-dimensional point cloud space of the target area. The preprocessing process mainly includes two steps: intensity filtering and altitude limit filtering.

[0057] Intensity filtering is the process of filtering based on the intensity information of each point in the point cloud. During the LiDAR scanning process, each point cloud data usually contains reflection intensity information in addition to the three-dimensional coordinates. By setting a suitable intensity threshold, points with too low or too high intensity values ​​can be removed. These points are often caused by noise, reflection interference or instability of the scanning device. The point cloud data after intensity filtering will be purer, which is conducive to subsequent processing and analysis.

[0058] Altitude limit filtering is the process of filtering based on the altitude information of each point in the point cloud. Since ground point clouds usually have lower altitudes and are unnecessary for the morphological segmentation of transmission lines, these ground point clouds can be removed by setting a suitable altitude threshold. In this way, the resulting point cloud data will only contain the three-dimensional point cloud space of the target area (such as the transmission line area), without including irrelevant information such as the ground.

[0059] like Figure 4 As shown, the embodiment of the present application obtains a purer and more accurate three-dimensional point cloud space of the target area by pre-processing the original point cloud data of the target area, including intensity filtering and altitude limit filtering, which provides strong data support for subsequent processing and analysis and helps to achieve fast and accurate segmentation of transmission lines.

[0060] Step 102 , extracting candidate tower poles in the three-dimensional point cloud space based on the geometric structure constraints of the tower poles, roughly filtering the candidate tower poles to obtain roughly filtered tower poles, and performing instance segmentation on the roughly filtered tower poles based on a point cloud clustering algorithm to obtain segmented tower poles.

[0061] This step involves the extraction of candidate towers, coarse filtering, and instance segmentation based on a point cloud clustering algorithm.

[0062] The extraction of candidate towers is based on the geometric structural constraints of the towers, which include altitude continuity: the towers should be continuous in altitude, which helps to distinguish the towers from other features in the point cloud data; cross-sectional area difference: the cross-sectional area at the bottom of the tower is usually larger than the cross-sectional area at the top, which is determined by the structural characteristics of the tower. This feature can serve as an important basis for identifying towers; edge connection existence: there is at least one direction of connection at the edge of the tower, which helps to outline the outline of the tower in the point cloud data. Based on the above constraints, candidate towers can be extracted from the preprocessed three-dimensional point cloud space.

[0063] The purpose of coarse filtering is to remove those candidate towers that do not meet the tower characteristics to reduce the amount of calculation for subsequent processing. Coarse filtering can be based on some simple geometric features and statistical information, such as: height range restriction: set a reasonable height range to remove those candidate towers that are too high or too low; point cloud density analysis: analyze the point cloud density of the candidate towers and remove those candidate towers with too low or too high point cloud density, because these may be noise or irrelevant objects; after coarse filtering, a series of candidate towers that are closer to the real towers are obtained, and these candidate towers will be used for subsequent instance segmentation.

[0064] The purpose of instance segmentation is to further segment the candidate towers after coarse filtering into independent individuals, that is, segment the towers. This is usually achieved through a point cloud clustering algorithm. Commonly used point cloud clustering algorithms include Euclidean clustering, DBSCAN clustering, etc. In the embodiment of the present application, the Euclidean clustering algorithm is used.

[0065] The clustering process includes the selection of seed points: randomly selecting a point from the point cloud of the coarse filtering tower pole as the seed point; distance calculation and cluster generation: calculating the Euclidean distance from the seed point to all other points. If the distance between two points is less than the first preset threshold, they are considered to belong to the same cluster. All points whose distance to the seed point is less than the threshold are added to the same cluster, and this process is recursively performed until no new points meet the condition that the distance is less than the threshold; generation of new clusters: when a point in a cluster no longer meets the condition that the distance to the seed point is less than the threshold, the cluster is regarded as a new cluster, and a new seed point is selected for repeated calculation. This process will continue until all points in the point cloud of the coarse filtering tower pole are assigned to a cluster; acquisition of segmented tower poles: after the above clustering process, a series of independent clusters are obtained, each cluster represents a segmented tower pole.

[0066] In the process of extracting and roughly filtering candidate towers, the calculation of cross-sectional area is an important auxiliary means. The calculation method of cross-sectional area includes point cloud extraction: extracting all point clouds within a 5 cm error range 5 meters above the bottom of the tower and all point clouds within a 5 cm error range 5 meters below the top of the tower, and obtaining the cross-sectional point cloud at the bottom of the tower and the cross-sectional point cloud at the top of the tower respectively; algorithm application: based on the extracted cross-sectional point cloud, the cross-sectional area at the bottom of the tower and the cross-sectional area at the top of the tower are calculated by the Gift Wrapping algorithm. This step helps to further verify the authenticity of the candidate tower and provide important parameters for subsequent processing.

[0067] Specifically, the polygon area calculation method is used to calculate the convex hull area to estimate the cross-sectional area. A reference point is selected from the point set of the cross-sectional point cloud. Usually, the point with the smallest ordinate (if there are multiple points with the same ordinate, the leftmost point is selected) is selected as the reference point. The remaining points are sorted according to the polar angle relative to the reference point. The result of the sorting is that, starting from the reference point, the other points are arranged counterclockwise according to the polar angle. Consider the sorted points in turn and add them to the convex hull. At each step, check whether the current point can form a convex hull. If not, pop out the point added to the convex hull last time until the current point can be added to the convex hull. Repeat the above process until all points have been considered, and the final convex hull is what is required.

[0068] Sort the points on the convex hull according to the polar angles to ensure that the points on the convex hull are arranged in order. Use the points on the convex hull to calculate the area of ​​the convex hull using the area calculation formula. The formula is as follows:

[0069]

[0070] Among them, Area represents the area of ​​the convex hull, (x i ,y i ) and (x i+1 ,y i+1 ) represents the coordinates of the points on the convex hull, and n is the number of points on the convex hull.

[0071] like Figure 5 As shown, through the processing of step 102, a series of independent segmented tower poles can be obtained, which provides strong data support for subsequent processing and analysis.

[0072] Step 103, fine filtering is performed on the segmented tower poles to obtain fine filtered tower poles, external lines in the fine filtered tower poles are extracted based on a point cloud clustering algorithm to obtain candidate line point clouds, straight line fitting is performed on the candidate line point clouds and classification is performed to obtain external lines in the same direction.

[0073] like Figure 6 As shown in the figure, this step performs fine filtering on the segmented tower poles and extracts external wires. After obtaining the segmented tower poles, they need to be further fine filtered to eliminate possible false tower poles. The basis for fine filtering is to determine whether there are external wires in the segmented tower poles. External wires usually refer to the parts used to connect wires between tower poles. In 3D point cloud data, these external wires are often expressed as a series of continuous point clouds.

[0074] An algorithm is used to determine whether there is such an external line on the split tower pole. If there is an external line on the split tower pole, it is regarded as a real tower pole, that is, a fine filtering tower pole; if there is no external line on the split tower pole, it may be a false tower pole and should be eliminated.

[0075] After the fine-filtered towers are determined, the next step is to extract the circumference lines of these towers. The extraction of circumference lines is usually based on point cloud clustering algorithms, such as Euclidean clustering. The point clouds in the fine-filtered towers are clustered using the Euclidean clustering algorithm to extract the circumference lines. These circumference lines appear as a series of continuous point clouds in 3D space, connecting different towers. The extracted circumference line point clouds are candidate line point clouds, which will be used in subsequent straight line fitting and classification steps.

[0076] After obtaining the candidate line point cloud, straight line fitting and classification are required to obtain the external line in the same direction. The embodiment of the present application uses the RANSAC algorithm to perform straight line fitting on the candidate line point cloud. The RANSAC algorithm is a robust model fitting algorithm that can accurately fit the model in data containing a large amount of noise and outliers. The basic principle process of straight line fitting is as follows:

[0077] 1) Random sampling: Randomly select two points from the point cloud as the starting point and end point of the line.

[0078] 2) Model fitting: Use the two selected points to fit the straight line model. Here, the least squares method (LeastSquares) is usually used to fit a straight line.

[0079] 3) Inlier check: Calculate the distance from all points to the fitted line and determine which points are considered to be inliers. This step uses a predefined threshold, usually a distance threshold, which is a preset empirical value.

[0080] 4) Evaluate the model: Calculate the quality of the fitted model, i.e. the number of points considered as inliers. The quality of the model can be measured by the number of inliers, the more inliers, the better the model.

[0081] 5) Repeated iteration: Repeat the above steps multiple times (the number of RANSAC iterations is a pre-set parameter), and each time a fitting model and corresponding inliers will be obtained.

[0082] 6) Optimal model selection: Among all iterations, the model with the largest number of inliers is selected as the final fitting result.

[0083] Through the RANSAC algorithm, a fitting straight line can be obtained, which represents the main direction of the candidate line point cloud. The variance from the candidate line point cloud to the fitting straight line is calculated. If the variance is higher than the second preset threshold, it means that the deviation between the candidate line point cloud and the fitting straight line is large, which may be noise or an outlier and should be eliminated.

[0084] Candidate line point clouds are classified according to their slope and overlap rate. Candidate line point clouds with the same slope may belong to the same external line. At the same time, if the overlap rate of two candidate line point clouds is greater than 90%, they can also be considered to belong to the same external line. Through classification, external lines in the same direction can be obtained, which will be used for the subsequent 3D point cloud morphological segmentation of the transmission line.

[0085] Through the processing of step 103, a more accurate and reliable transmission line 3D point cloud morphological segmentation result can be obtained.

[0086] Step 104 , diffuse in the growth direction of the external lines in the same direction, and extract the point cloud of the external lines in the same direction until they coincide with the external lines of another tower pole, thereby obtaining the inter-tower lines.

[0087] In the above steps, through a series of processing, such as the acquisition and preprocessing of the original point cloud data, the extraction and coarse filtering of candidate towers, the acquisition of segmented towers, the extraction of fine filtered towers, and the linear fitting and classification of the external lines, the external lines in the same direction are obtained. These external lines represent the possible connection paths between towers.

[0088] The growth direction of the circumferential line in the same direction is determined based on the result of the straight line fitting. In the straight line fitting process, the embodiment of the present application obtains the direction vector of the fitting line, and this direction vector represents the growth direction of the circumferential line.

[0089] After the growth direction is determined, diffusion is performed along this direction. The diffusion process is actually to find points adjacent to the current external line in the point cloud data and add these points to the point cloud of the external line. This process needs to be iterated continuously until a certain stop condition is met.

[0090] During the diffusion process, it is important to note that the diffusion range needs to be moderate, neither too large nor too small. A range that is too large may lead to the introduction of unnecessary noise points, while a range that is too small may miss some important connection points; the number of iterations also needs to be controlled to avoid infinite loops. Usually, we can set a maximum number of iterations, or stop the iteration when it diffuses to a certain area (such as the external line of another tower); during the diffusion process, the quality of the added point clouds needs to be checked to ensure that they meet our requirements. For example, check the coordinates, intensity and other properties of the points to eliminate those points that do not meet the requirements.

[0091] During the diffusion process, when the outer line meets the outer line of another tower, they are considered to have overlapped. At this time, the two outer lines are connected to form a complete inter-tower line.

[0092] Specifically, take the side of the external wiring of the transmission tower close to the tower pole as the starting point and the other side as the growth direction, and extend the fitting straight line of the external wiring by 5 meters in the growth direction; search for candidate point clouds: for the straight line obtained by growth, traverse and search all point clouds within 2 meters of the straight line, and mark these point clouds as candidate point clouds of the transmission line of the external wiring; update the fitted straight line: according to the newly added transmission line point cloud, fit the new straight line equation. In order to achieve faster speed, you only need to update the endpoints, intercept and slope of the straight line equation. The intercept and slope can be directly obtained using the two-point formula. Iterate for each external wiring until there is no subsequent point cloud.

[0093] Through the above steps, the inter-tower line is obtained. The inter-tower line represents the actual connection path between the tower poles and is of great significance in the 3D point cloud morphological segmentation of the transmission line. After obtaining the inter-tower line, it needs to be processed later, such as curve fitting and parameterized representation. These processing steps can further improve the accuracy and reliability of the inter-tower line and provide strong support for subsequent analysis and application.

[0094] Step 105, curve fitting and parameterization are performed on the lines between towers to complete the 3D point cloud morphological segmentation of the transmission line.

[0095] Step 105 aims to further smooth and optimize the inter-tower lines obtained through a series of processing, so as to more accurately describe the shape of the transmission line.

[0096] As one of the results of the 3D point cloud morphological segmentation of the transmission line, the inter-tower line represents the actual connection path between the tower poles. However, due to the discreteness and noise of the point cloud data, there may be some irregular and non-smooth parts in the inter-tower line. Therefore, through curve fitting, the embodiment of the present application smoothes the inter-tower line to make it more consistent with the actual morphology of the transmission line. In addition, curve fitting can also provide more accurate and reliable input data for subsequent applications such as three-dimensional modeling and path planning of transmission lines.

[0097] When performing curve fitting on the inter-tower line, the present application embodiment selects a quadratic curve fitting method, which includes the least squares method. The least squares method is a mathematical optimization technique that finds the best function match for the data by minimizing the sum of squares of the errors. The principle steps of quadratic curve fitting are as follows.

[0098] 1) Data preparation: Select points containing quadratic curve shapes from the point cloud.

[0099] 2) Parametric model: Parameterize the quadratic curve model y = ax 2 +bx+c, where a, b, c are the parameters that need to be fitted.

[0100] 3) Least squares optimization: Use the least squares method to optimize the parameters and minimize the sum of squares of the residuals from the observation point to the fitting curve. The residual is the deviation of each observation point in the vertical direction.

[0101] 4) Iteration: Optimize parameters continuously through iteration until the convergence condition is met or the predetermined number of iterations is reached.

[0102] 5) Model evaluation: Evaluate the quality of the fitted model, usually measured by the residual sum of squares or other indicators. A smaller residual sum of squares indicates a better fit.

[0103] Before fitting, the present embodiment also organizes the point cloud data of the tower line into a series of ordered coordinate points. These coordinate points will be used as input data for the fitting model. After obtaining the coefficients of the fitting model, they will be substituted into the model formula to calculate the coordinates of any point on the fitting curve. At the same time, the fitting results need to be evaluated to determine whether they meet our accuracy requirements.

[0104] After completing the curve fitting, the fitting curve needs to be parametrically represented. Parametric representation means describing the shape and position of the curve with a set of parameters. These parameters can be geometric quantities such as the length and curvature of the curve, or other physical quantities related to the shape of the curve.

[0105] For the fitting curve obtained by quadratic curve fitting, select appropriate parameterization methods, such as arc length parameterization, curvature parameterization, etc., to parametrically represent it. This not only facilitates further description and analysis of the shape and position of the curve, but also provides more flexible and diverse input data for subsequent applications.

[0106] Specifically, the embodiment of the present application uses an embedded development board to perform performance testing, and the performance parameter information of the core board used is shown in the following table:

[0107] Table 1

[0108]

[0109] The data was collected by DJI M300-RTK equipped with Zenmuse L1 LiDAR payload on a power transmission line in a certain place. The point cloud was post-processed by DJI Zhitu to generate a point cloud model. The point cloud model has a data volume of about 200 million points. Due to its large volume, the point cloud model is divided into 5 segments with a radius of 24 meters around the tower, that is, the visible range of the LiDAR when the drone is on the top of the transmission tower, and 5 segments of point cloud data between the two towers, a total of 10 groups of point cloud models. Some data models are as follows Figure 7 shown.

[0110] The performance indicators mainly include the algorithm's running time, transmission tower segmentation accuracy, transmission line segmentation accuracy, and transmission line fitting accuracy.

[0111] 1) Algorithm running time: the total time consumed by the entire process of the transmission line segmentation algorithm, in milliseconds (ms);

[0112] 2) Transmission tower segmentation accuracy: The algorithm can correctly segment the real transmission towers without missing any segmentation or segmenting any false towers. The results are either accurate or inaccurate.

[0113] 3) Transmission line segmentation accuracy: the algorithm can correctly instance segment the real transmission line, and the wire bundle matches the actual wire bundle, and the results are accurate and inaccurate;

[0114] 4) Transmission line fitting accuracy: reflects the error between the fitting curve and the actual transmission line point cloud, in meters (m), and is calculated as follows: the sum of the distances between each transmission line laser point and the fitting curve divided by the total number of transmission line laser points.

[0115] On the RK3588 platform, by writing 10 point cloud model file paths in the algorithm configuration file, running the algorithm in turn to perform power line segmentation tests and evaluate performance indicators. The final results at room temperature of 20 degrees Celsius are shown in the following table:

[0116] Table 2

[0117]

[0118]

[0119] The segmentation effects of each model and the original model are shown in the figure below. Figure 8 (a) to Figure 8 (t) shown.

[0120] From Table 2 and Figure 8 It can be observed that the transmission line segmentation algorithm has a relatively fast overall running speed. Even in the case of a point cloud of 50 million, the segmentation algorithm can complete the instance segmentation of the transmission line and the tower within 4 seconds. In addition, the segmentation accuracy of the tower line is relatively high, and the curve fitting error of the transmission line is controlled within 20 cm. This shows that the algorithm not only has an advantage in speed, but also shows good performance in instance segmentation and accuracy.

[0121] like Figure 2 As shown, the embodiment of the present application provides a transmission line 3D point cloud morphological segmentation device, including a data acquisition module 201, a filtering segmentation module 202, a filtering fitting module 203, a diffusion extraction module 204 and a parameterization module 205, wherein:

[0122] The data acquisition module 201 is configured to acquire the original point cloud data of the target area and perform preprocessing, wherein the preprocessing includes intensity filtering, and the three-dimensional point cloud space of the target area is obtained by performing intensity filtering on the original point cloud data of the target area;

[0123] The filtering and segmentation module 202 is configured to extract candidate tower poles in the three-dimensional point cloud space based on the geometric structure constraints of the tower poles, perform rough filtering on the candidate tower poles to obtain rough filtered tower poles, and perform instance segmentation on the rough filtered tower poles based on the point cloud clustering algorithm to obtain segmented tower poles;

[0124] The filtering and fitting module 203 is configured to perform fine filtering on the segmented tower poles to obtain fine filtered tower poles, extract external lines in the fine filtered tower poles based on a point cloud clustering algorithm to obtain candidate line point clouds, perform straight line fitting on the candidate line point clouds and classify them to obtain external lines in the same direction;

[0125] The diffusion extraction module 204 is configured to diffuse in the growth direction of the external lines in the same direction and extract the point cloud of the external lines in the same direction until they coincide with the external lines of another tower pole to obtain the inter-tower lines;

[0126] The parameterization module 205 is configured to perform curve fitting and parameterization on the tower lines to complete the 3D point cloud morphological segmentation of the transmission line.

[0127] The transmission line 3D point cloud morphological segmentation device of the embodiment of the present application can be a computer device, or a component in the computer device, such as an integrated circuit or a chip. The computer device can be a terminal, or a device other than a terminal. Exemplarily, the computer device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted computer device, a mobile Internet device (Mobile Internet Device, MID), an ultra-mobile personal computer (Ultra-Mobile Personal Computer, UMPC), a netbook or a personal digital assistant (Personal Digital Assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (Personal Computer, PC), etc., which is not specifically limited in the embodiment of the present application.

[0128] The 3D point cloud morphology segmentation device for power transmission lines provided in the embodiment of the present application can achieve Figure 1 To avoid repetition, the various processes implemented in the embodiment of the transmission line 3D point cloud morphological segmentation method are not repeated here.

[0129] The present application also provides a computer device, such as Figure 3 As shown, the computer device includes a processor 301 and a memory 302. The memory 302 stores programs or instructions that can be run on the processor 301. When the program or instructions are executed by the processor 301, the various steps of the above-mentioned transmission line 3D point cloud morphological segmentation method are implemented, and the same technical effect can be achieved. To avoid repetition, they are not repeated here.

[0130] It should be noted that the computer device in the embodiment of the present application includes the above-mentioned mobile computer device and non-mobile computer device.

[0131] The memory 302 can be used to store software programs and various data. The memory 302 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory 302 may include a volatile memory or a non-volatile memory, or the memory 302 may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM). The memory 302 of the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

[0132] The processor 301 may include one or more processing units; optionally, the processor 301 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to an operating system, a user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor 301.

[0133] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned embodiment of the 3D point cloud morphological segmentation method for transmission lines is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0134] An embodiment of the present application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned transmission line 3D point cloud morphological segmentation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0135] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0136] An embodiment of the present application also provides a computer program product, which is stored in a storage medium. The program product is executed by at least one processor to implement the various processes of the above-mentioned transmission line 3D point cloud morphological segmentation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0137] It should be noted that, in the present application, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises one..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be noted that the scope of the method and device in the embodiment of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0138] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A 3D point cloud morphological segmentation method for power transmission lines, characterized in that: The method comprises: Acquire the original point cloud data of the target area and perform preprocessing, wherein the preprocessing includes intensity filtering, and obtain the three-dimensional point cloud space of the target area by performing intensity filtering on the original point cloud data of the target area; Extracting candidate tower poles in the three-dimensional point cloud space based on geometric structural constraints of the tower poles, roughly filtering the candidate tower poles to obtain roughly filtered tower poles, and performing instance segmentation on the roughly filtered tower poles based on a point cloud clustering algorithm to obtain segmented tower poles; Fine filtering is performed on the segmented tower poles to obtain fine filtered tower poles, external lines in the fine filtered tower poles are extracted based on a point cloud clustering algorithm to obtain candidate line point clouds, and straight line fitting and classification are performed on the candidate line point clouds to obtain external lines in the same direction; Diffusion is performed in the growth direction of the external line in the same direction, and the point cloud of the external line in the same direction is extracted until it coincides with the external line of another tower pole, so as to obtain the inter-tower line; The inter-tower lines are subjected to curve fitting and parameterized representation to complete the 3D point cloud morphological segmentation of the transmission line.

2. The method for 3D point cloud morphology segmentation of a power transmission line according to claim 1, characterized in that: The preprocessing also includes altitude restriction filtering, and the three-dimensional point cloud space of the target area excluding the ground is obtained by performing intensity filtering and altitude restriction filtering on the original point cloud data of the target area.

3. The method for 3D point cloud morphology segmentation of a power transmission line according to claim 1, characterized in that: The geometric constraints of the tower include: The tower is continuous in elevation; The cross-sectional area at the base of the tower is greater than that at the top; There is at least one connecting line at the edge of the tower.

4. The method for 3D point cloud morphology segmentation of a power transmission line according to claim 3, characterized in that: The calculation method of the cross-sectional area includes: All point clouds within a 5-centimeter error range 5 meters above the bottom of the tower pole and all point clouds within a 5-centimeter error range 5 meters below the top of the tower pole are extracted to obtain the cross-sectional point cloud of the bottom of the tower pole and the cross-sectional point cloud of the top of the tower pole; based on the cross-sectional point cloud of the bottom of the tower pole and the cross-sectional point cloud of the top of the tower pole, the cross-sectional area of ​​the bottom of the tower pole and the cross-sectional area of ​​the top of the tower pole are calculated by the Gift Wrapping algorithm.

5. The method for 3D point cloud morphology segmentation of a power transmission line according to claim 1, characterized in that: The method of performing instance segmentation on the coarse filtering tower pole based on the point cloud clustering algorithm to obtain segmented tower poles includes: Obtaining a point cloud of a coarsely filtered tower pole, and taking any point therein as a seed point, calculating the Euclidean distance from the seed point to all other points, and if the Euclidean distance between two points is less than a first preset threshold, it is considered that they belong to the same cluster; Add all points whose Euclidean distance from the seed point is less than a first preset threshold to the same cluster, and recursively perform the process until no new point satisfies the condition that the Euclidean distance from the seed point is less than the first preset threshold; When a point in a cluster no longer satisfies the condition that the Euclidean distance with the seed point is less than the first preset threshold, the cluster is regarded as a new cluster, and a new seed point is selected for repeated calculation until all points in the point cloud of the coarsely filtered tower are assigned to a cluster to obtain a segmented tower.

6. The method for 3D point cloud morphology segmentation of a power transmission line according to claim 1, characterized in that: The method of finely filtering the segmented tower poles to obtain finely filtered tower poles, extracting external lines in the finely filtered tower poles based on a point cloud clustering algorithm to obtain candidate line point clouds, and performing straight line fitting and classification on the candidate line point clouds to obtain external lines in the same direction includes: By judging whether the split tower rod has an external line, the split tower rod is finely filtered. If the split tower rod has an external line, it is filtered as a fine filtering tower rod. If the split tower rod does not have an external line, it is not filtered as a fine filtering tower rod. By fine filtering the split tower rod, false tower rods existing in the split tower rod are eliminated. Extracting the external lines in the fine filtering tower pole by Euclidean clustering to obtain a candidate line point cloud; performing straight line fitting on the candidate line point cloud by RANSAC algorithm to obtain a fitting line, calculating the variance from the candidate line point cloud to the fitting line, and if the variance is higher than a second preset threshold, eliminating the candidate line point cloud; The candidate line point clouds are classified according to their slopes and overlap rates, and candidate line point clouds with the same slope and an overlap rate greater than 90% are classified as a circumscribed line in the same direction.

7. The method for 3D point cloud morphology segmentation of a power transmission line according to claim 1, characterized in that: The curve fitting and parameterized representation of the inter-tower line to complete the 3D point cloud morphological segmentation of the transmission line includes: The inter-tower line is parameterized by quadratic curve fitting, and the quadratic curve fitting method includes a least squares method.

8. A transmission line 3D point cloud morphology segmentation device, characterized in that: The device comprises: A data acquisition module is configured to acquire the original point cloud data of the target area and perform preprocessing, wherein the preprocessing includes intensity filtering, and the three-dimensional point cloud space of the target area is obtained by performing intensity filtering on the original point cloud data of the target area; A filtering and segmentation module is configured to extract candidate tower poles in the three-dimensional point cloud space based on geometric structural constraints of the tower poles, perform coarse filtering on the candidate tower poles to obtain coarse filtered tower poles, and perform instance segmentation on the coarse filtered tower poles based on a point cloud clustering algorithm to obtain segmented tower poles; A filtering and fitting module is configured to perform fine filtering on the segmented tower poles to obtain fine filtered tower poles, extract external lines in the fine filtered tower poles based on a point cloud clustering algorithm to obtain candidate line point clouds, perform straight line fitting and classification on the candidate line point clouds to obtain external lines in the same direction; A diffusion extraction module is configured to diffuse in the growth direction of the external line in the same direction and extract the point cloud of the external line in the same direction until it coincides with the external line of another tower pole to obtain the inter-tower line; The parameterization module is configured to perform curve fitting and parameterization on the tower-tower line to complete the 3D point cloud morphological segmentation of the transmission line.

9. A transmission line 3D point cloud morphological segmentation system, the system comprising a processor and a memory, the memory storing a computer program, characterized in that: The computer program is loaded and executed by the processor to implement the transmission line 3D point cloud morphological segmentation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, wherein a computer program is stored in the storage medium, characterized in that: When the computer program is executed by a processor, it is used to implement the transmission line 3D point cloud morphological segmentation method according to any one of claims 1 to 7.

Images