Power transmission line point cloud registration method and device, and computer readable storage medium

Abstract

The invention belongs to the technical field of power transmission line point cloud registration, and particularly relates to a power transmission line point cloud registration method and device and a computer readable storage medium. The method comprises the following steps: firstly, acquiring point cloud data sets of the same power transmission line, namely a source point cloud set and a target point cloud set, and extracting power lines and pole nodes in the source point cloud set and the target point cloud set; then, an optimal rotation matrix for registration is solved by utilizing the extracted power line, and an optimal translation vector for registration is solved by utilizing the extracted pole nodes; and finally, performing power transmission line point cloud registration by using the optimal rotation matrix and the optimal translation vector. According to the whole scheme, only the power line and tower nodes in the power transmission line point cloud need to be extracted, then the rotation matrix and the translation vector are calculated based on the extraction result, the calculation amount of the whole scheme is not large, but high precision can still be guaranteed, and point cloud registration in the complex scene of the power transmission line is achieved.

Description

Technical Field

[0001] This invention belongs to the field of transmission line point cloud registration technology, specifically relating to a transmission line point cloud registration method, equipment, and computer-readable storage medium. Background Technology

[0002] In the field of power transmission line inspection, airborne lidar systems can quickly acquire three-dimensional spatial information of power transmission lines, i.e., point cloud data. They offer advantages such as high efficiency, no loss, and non-contact operation, overcoming the shortcomings of traditional manual ground inspections and becoming a hot research topic in intelligent power transmission line inspection. However, lidar has a limited acquisition range, and laser light has occlusion characteristics. Therefore, when performing three-dimensional modeling of a power transmission line scene, point cloud registration technology is needed to fuse point cloud data collected from different routes and angles to obtain a comprehensive description of the power transmission line scene.

[0003] Iterative Closest Point (ICP) and its variants are the most commonly used registration algorithms. First, the closest points in two point clouds are selected as the closest point pair. Then, the rotation and translation transformation matrices are solved for all closest point pairs. The registration error between the two point clouds is reduced through continuous iteration until a pre-set threshold or number of iterations is met. While the ICP algorithm can meet general point cloud registration needs, it still has the following drawbacks in complex transmission line scenarios: when the initial transformation is poor, the algorithm is prone to getting trapped in local optima or failing to converge; the density, noise level, and uniformity of the point cloud distribution all affect the performance of the ICP algorithm.

[0004] Chinese invention patent application CN120912649A, published on November 7, 2025, discloses an automatic registration method for power grid line point clouds. This method is based on misaligned point cloud data of transmission channel morphology, offset point cloud data of power grid line transmission response, and hierarchical structure separation data of power grid line point clouds. It performs point cloud registration and pose error correction between multiple frames of power grid line point clouds, comprehensively utilizing various data types to fully leverage their advantages and achieve automatic registration of power grid line point clouds. This scheme requires combining various types of data and performing power topology analysis of power grid lines, electric field distribution and sensitivity analysis, and fiber-optic-electrical layout analysis to generate the misaligned point cloud data of transmission channel morphology, offset point cloud data of power grid line transmission response, and hierarchical structure separation data of power grid line point clouds. Undoubtedly, the calculations are complex and computationally intensive. Summary of the Invention

[0005] The purpose of this invention is to provide a method, device, and computer-readable storage medium for point cloud registration of transmission lines, so as to solve the problem of large computational load in point cloud registration in the prior art.

[0006] To address the aforementioned technical problems, this invention provides a technical solution for a point cloud registration method for transmission lines, as detailed below: A method for point cloud registration of transmission lines includes the following steps: 1) Obtain point cloud datasets of the same transmission line, namely source point cloud and target point cloud, and extract power line and tower nodes from the source point cloud and target point cloud; 2) The optimal rotation matrix for registration is obtained by using the extracted power lines, and the optimal translation vector for registration is obtained by using the extracted tower nodes; 3) Use the optimal rotation matrix and optimal translation vector to perform point cloud registration for transmission lines.

[0007] The beneficial effects of the above technical solution are as follows: This invention utilizes the obvious characteristics of power transmission line scenarios, namely, the distinct linear features of power lines and the numerous intersecting nodes in the tower structure, to calculate rotation matrices and translation vectors. Specifically, the rotation matrix is ​​calculated based on the linear features of the power lines, and the translation vector is calculated based on the intersecting nodes of the towers, thereby obtaining the optimal rotation matrix and optimal translation vector. The calculated optimal rotation matrix and optimal translation vector are then used for point cloud registration of the power transmission line. The entire solution only requires extracting the power lines and tower nodes from the power transmission line point cloud, and then calculating the rotation matrix and translation vector based on the extraction results. The computational load of the entire solution is not large, yet it still ensures high accuracy, achieving point cloud registration in complex power transmission line scenarios.

[0008] Furthermore, the process of extracting electric field lines includes: repeating the following process until the number of iterations reaches a preset iteration threshold or the number of output electric field line models reaches a preset electric field line model output threshold: selecting sample points from the source point cloud and the target point cloud, performing line fitting using the selected sample points to obtain a line fitting model; calculating the distance from other sample points in the point cloud to the line simulation model, classifying points with a distance less than or equal to a preset distance threshold as interior points of the line fitting model, and classifying points with a distance greater than a preset distance threshold as exterior points of the line fitting model; and outputting the line fitting model with an interior point count greater than or equal to a preset electric field line point count threshold as the electric field line model.

[0009] Furthermore, the process of extracting tower nodes includes: extracting towers from the source point cloud and the target point cloud as tower structural component models for output, and then repeating the following process iteratively until the number of iterations reaches the preset first iteration threshold or the number of output tower nodes reaches the preset tower node number threshold: select two output tower structural component models, find the set of interior points belonging to the two selected tower structural component models, and if the set of points is not empty, traverse the sum of the squares of the distances from each interior point in the set of interior points to the two tower structural component models, and output the interior point corresponding to the smallest sum of squares as a tower node.

[0010] Furthermore, the process of extracting tower structural component models includes: repeating the following process iteratively until the number of iterations reaches a preset second iteration threshold or the number of output tower structural component models reaches a preset tower structural component output threshold: selecting sample points from the source point cloud and the target point cloud, using the selected sample points to perform linear fitting to obtain a linear fitting model; calculating the distance from other sample points in the point cloud to the linear simulation model, classifying points with a distance less than a preset distance threshold as interior points of the linear fitting model, and classifying points with a distance greater than or equal to the preset distance threshold as exterior points of the linear fitting model; outputting linear fitting models with an interior point count greater than or equal to a preset tower structural component point count threshold and less than a preset power line point count threshold as tower structural component models; the preset tower structural component point count threshold is less than the preset power line point count threshold.

[0011] Furthermore, the process of finding the optimal rotation matrix includes: setting an initial rotation matrix, and then repeating the following process iteratively until the number of iterations reaches a preset iteration threshold or the error between line pairs is less than a preset error threshold, thereby obtaining the optimal rotation matrix: finding the electric field lines extracted from the target point cloud that are closest to each electric field line extracted from the source point cloud, thereby obtaining a series of line pairs; transforming the source point cloud and the electric field lines in the source point cloud using the current rotation matrix, and calculating the error between all line pairs using the transformation result; adjusting the rotation matrix according to the error to reduce the error.

[0012] Furthermore, the process of finding the optimal translation vector includes: setting an initial translation vector, and then repeating the following process iteratively until the number of iterations reaches a preset iteration threshold or the error between the nearest point pairs is less than a preset error threshold, thereby obtaining the optimal translation vector: finding the tower nodes closest to each tower node extracted from the source point cloud from the tower nodes extracted from the target point cloud, thereby obtaining a series of nearest point pairs; transforming the source point cloud and the tower nodes in the source point cloud using the current translation vector, and calculating the error between all nearest point pairs using the transformation result; adjusting the translation vector according to the error to reduce the error.

[0013] Furthermore, the source point cloud and the target point cloud are point cloud data collected from different headings and different acquisition angles.

[0014] Furthermore, the rotation matrix is ​​adjusted based on the error using the least squares method.

[0015] To address the aforementioned technical problems, the present invention also provides a technical solution for a computer device, as detailed below: A computer device according to the present invention includes a processor, the processor being configured to execute a computer program to implement the steps of the following method: 1) Obtain point cloud datasets of the same transmission line, namely source point cloud and target point cloud, and extract power line and tower nodes from the source point cloud and target point cloud; 2) The optimal rotation matrix for registration is obtained by using the extracted power lines, and the optimal translation vector for registration is obtained by using the extracted tower nodes; 3) Use the optimal rotation matrix and optimal translation vector to perform point cloud registration for transmission lines.

[0016] The beneficial effects of the above technical solution are as follows: The computer equipment of this invention provides basic hardware support for realizing the point cloud registration method for transmission lines of this invention. The core of this method is based on the obvious features in the transmission line scene, namely, the obvious linear features of power lines and the large number of intersection nodes in the tower structure. These obvious features are used to calculate the rotation matrix and translation vector. Specifically, the rotation matrix is ​​calculated based on the linear features of the power lines, and the translation vector is calculated based on the intersection nodes of the towers, thereby calculating the optimal rotation matrix and optimal translation vector. The calculated optimal rotation matrix and optimal translation vector are then used for point cloud registration of the transmission line. The entire solution only requires extracting the power lines and tower nodes from the transmission line point cloud, and then calculating the rotation matrix and translation vector based on the extraction results. The computational load of the entire solution is not large, but it can still guarantee high accuracy, realizing point cloud registration in complex transmission line scenes.

[0017] Furthermore, the process of extracting electric field lines includes: repeating the following process until the number of iterations reaches a preset iteration threshold or the number of output electric field line models reaches a preset electric field line model output threshold: selecting sample points from the source point cloud and the target point cloud, performing line fitting using the selected sample points to obtain a line fitting model; calculating the distance from other sample points in the point cloud to the line simulation model, classifying points with a distance less than or equal to a preset distance threshold as interior points of the line fitting model, and classifying points with a distance greater than a preset distance threshold as exterior points of the line fitting model; and outputting the line fitting model with an interior point count greater than or equal to a preset electric field line point count threshold as the electric field line model.

[0018] Furthermore, the process of extracting tower nodes includes: extracting towers from the source point cloud and the target point cloud as tower structural component models for output, and then repeating the following process iteratively until the number of iterations reaches the preset first iteration threshold or the number of output tower nodes reaches the preset tower node number threshold: select two output tower structural component models, find the set of interior points belonging to the two selected tower structural component models, and if the set of points is not empty, traverse the sum of the squares of the distances from each interior point in the set of interior points to the two tower structural component models, and output the interior point corresponding to the smallest sum of squares as a tower node.

[0019] Furthermore, the process of extracting tower structural component models includes: repeating the following process iteratively until the number of iterations reaches a preset second iteration threshold or the number of output tower structural component models reaches a preset tower structural component output threshold: selecting sample points from the source point cloud and the target point cloud, using the selected sample points to perform linear fitting to obtain a linear fitting model; calculating the distance from other sample points in the point cloud to the linear simulation model, classifying points with a distance less than a preset distance threshold as interior points of the linear fitting model, and classifying points with a distance greater than or equal to the preset distance threshold as exterior points of the linear fitting model; outputting linear fitting models with an interior point count greater than or equal to a preset tower structural component point count threshold and less than a preset power line point count threshold as tower structural component models; the preset tower structural component point count threshold is less than the preset power line point count threshold.

[0020] Furthermore, the process of finding the optimal rotation matrix includes: setting an initial rotation matrix, and then repeating the following process iteratively until the number of iterations reaches a preset iteration threshold or the error between line pairs is less than a preset error threshold, thereby obtaining the optimal rotation matrix: finding the electric field lines extracted from the target point cloud that are closest to each electric field line extracted from the source point cloud, thereby obtaining a series of line pairs; transforming the source point cloud and the electric field lines in the source point cloud using the current rotation matrix, and calculating the error between all line pairs using the transformation result; adjusting the rotation matrix according to the error to reduce the error.

[0021] Furthermore, the process of finding the optimal translation vector includes: setting an initial translation vector, and then repeating the following process iteratively until the number of iterations reaches a preset iteration threshold or the error between the nearest point pairs is less than a preset error threshold, thereby obtaining the optimal translation vector: finding the tower nodes closest to each tower node extracted from the source point cloud from the tower nodes extracted from the target point cloud, thereby obtaining a series of nearest point pairs; transforming the source point cloud and the tower nodes in the source point cloud using the current translation vector, and calculating the error between all nearest point pairs using the transformation result; adjusting the translation vector according to the error to reduce the error.

[0022] Furthermore, the source point cloud and the target point cloud are point cloud data collected from different headings and different acquisition angles.

[0023] Furthermore, the rotation matrix is ​​adjusted based on the error using the least squares method.

[0024] To address the aforementioned technical problems, the present invention also provides a technical solution for a computer-readable storage medium, as detailed below: The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the following method: 1) Obtain point cloud datasets of the same transmission line, namely source point cloud and target point cloud, and extract power line and tower nodes from the source point cloud and target point cloud; 2) The optimal rotation matrix for registration is obtained by using the extracted power lines, and the optimal translation vector for registration is obtained by using the extracted tower nodes; 3) Use the optimal rotation matrix and optimal translation vector to perform point cloud registration for transmission lines.

[0025] The beneficial effects of the above technical solution are as follows: The computer-readable storage medium of this invention stores software code implementing the point cloud registration method for transmission lines, ensuring the reliable implementation of the method. The core of this method is based on the obvious features of the transmission line scenario, namely, the obvious linear features of power lines and the large number of intersection nodes in the tower structure. These obvious features are used to calculate the rotation matrix and translation vector. Specifically, the rotation matrix is ​​calculated based on the linear features of the power lines, and the translation vector is calculated based on the intersection nodes of the towers, thereby calculating the optimal rotation matrix and optimal translation vector. The calculated optimal rotation matrix and optimal translation vector are then used for point cloud registration of the transmission line. The entire solution only requires extracting the power lines and tower nodes from the transmission line point cloud, and then calculating the rotation matrix and translation vector based on the extraction results. The computational load of the entire solution is not large, but it still ensures high accuracy, achieving point cloud registration in complex transmission line scenarios.

[0026] Furthermore, the process of extracting electric field lines includes: repeating the following process until the number of iterations reaches a preset iteration threshold or the number of output electric field line models reaches a preset electric field line model output threshold: selecting sample points from the source point cloud and the target point cloud, performing line fitting using the selected sample points to obtain a line fitting model; calculating the distance from other sample points in the point cloud to the line simulation model, classifying points with a distance less than or equal to a preset distance threshold as interior points of the line fitting model, and classifying points with a distance greater than a preset distance threshold as exterior points of the line fitting model; and outputting the line fitting model with an interior point count greater than or equal to a preset electric field line point count threshold as the electric field line model.

[0027] Furthermore, the process of extracting tower nodes includes: extracting towers from the source point cloud and the target point cloud as tower structural component models for output, and then repeating the following process iteratively until the number of iterations reaches the preset first iteration threshold or the number of output tower nodes reaches the preset tower node number threshold: select two output tower structural component models, find the set of interior points belonging to the two selected tower structural component models, and if the set of points is not empty, traverse the sum of the squares of the distances from each interior point in the set of interior points to the two tower structural component models, and output the interior point corresponding to the smallest sum of squares as a tower node.

[0028] Furthermore, the process of extracting tower structural component models includes: repeating the following process iteratively until the number of iterations reaches a preset second iteration threshold or the number of output tower structural component models reaches a preset tower structural component output threshold: selecting sample points from the source point cloud and the target point cloud, using the selected sample points to perform linear fitting to obtain a linear fitting model; calculating the distance from other sample points in the point cloud to the linear simulation model, classifying points with a distance less than a preset distance threshold as interior points of the linear fitting model, and classifying points with a distance greater than or equal to the preset distance threshold as exterior points of the linear fitting model; outputting linear fitting models with an interior point count greater than or equal to a preset tower structural component point count threshold and less than a preset power line point count threshold as tower structural component models; the preset tower structural component point count threshold is less than the preset power line point count threshold.

[0029] Furthermore, the process of finding the optimal rotation matrix includes: setting an initial rotation matrix, and then repeating the following process iteratively until the number of iterations reaches a preset iteration threshold or the error between line pairs is less than a preset error threshold, thereby obtaining the optimal rotation matrix: finding the electric field lines extracted from the target point cloud that are closest to each electric field line extracted from the source point cloud, thereby obtaining a series of line pairs; transforming the source point cloud and the electric field lines in the source point cloud using the current rotation matrix, and calculating the error between all line pairs using the transformation result; adjusting the rotation matrix according to the error to reduce the error.

[0030] Furthermore, the process of finding the optimal translation vector includes: setting an initial translation vector, and then repeating the following process iteratively until the number of iterations reaches a preset iteration threshold or the error between the nearest point pairs is less than a preset error threshold, thereby obtaining the optimal translation vector: finding the tower nodes closest to each tower node extracted from the source point cloud from the tower nodes extracted from the target point cloud, thereby obtaining a series of nearest point pairs; transforming the source point cloud and the tower nodes in the source point cloud using the current translation vector, and calculating the error between all nearest point pairs using the transformation result; adjusting the translation vector according to the error to reduce the error.

[0031] Furthermore, the source point cloud and the target point cloud are point cloud data collected from different headings and different acquisition angles.

[0032] Furthermore, the rotation matrix is ​​adjusted based on the error using the least squares method. Attached Figure Description

[0033] Figure 1 This is a flowchart of the transmission line point cloud registration method of the present invention. Detailed Implementation

[0034] The core concept of this invention lies in extracting the intersection points of power lines and towers from the point cloud of a transmission line, calculating the rotation matrix based on the line features of the power lines, calculating the translation vector based on the intersection points of the towers, and iteratively solving for the optimal rotation matrix and translation vector to achieve point cloud registration in complex scenarios of transmission lines. Based on this concept, a transmission line point cloud registration method, a computer device, and a computer-readable storage medium can be implemented. To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings.

[0035] An implementation method for point cloud registration of transmission lines: The overall process of a point cloud registration method for transmission lines is as follows: Figure 1 As shown, the details are as follows: Step 1: Collect point cloud data of the transmission line to obtain the source point cloud set P and the target point cloud set Q.

[0036] Using drones equipped with radar scanning devices, a large-scale coverage scan is conducted along the power transmission line corridor to acquire a power transmission line point cloud dataset. During the data acquisition, different models of acquisition devices can be used, or different flight paths and acquisition angles can be used to acquire the same power transmission line corridor, obtaining the source point cloud P and the target point cloud Q.

[0037] Step 2: Preprocess the acquired point cloud data. Perform point cloud preprocessing on the source point cloud set P and the target point cloud set Q, including extraction of power lines and tower structural components, and extraction of tower nodes.

[0038] It should be noted that this invention extracts line features and intersection features based on the structural characteristics of power lines and towers in a transmission line scenario. If the extracted features do not belong to power lines and towers, it will not affect the subsequent registration results. In addition, power line extraction only needs to extract the straight sections of the power lines to simplify the calculation model and speed up the extraction process.

[0039] The process of extracting power line and tower structural components is as follows: 1) Randomly select a certain number of sample points from both the source point cloud and the target point cloud.

[0040] 2) Fit a straight line model based on the selected sample points.

[0041] 3) Calculate the distance from other points to the linear fitting model, set a preset distance threshold, and classify points whose calculated distance is less than or equal to the preset distance threshold as interior points of the linear fitting model, and points whose calculated distance is greater than the preset distance threshold as exterior points.

[0042] 4) Set thresholds for the number of power line points and the number of points for pole / tower structural components, with the power line point threshold being greater than the pole / tower structural component point threshold. Output models with an interior point count greater than or equal to the pole / tower structural component point threshold but less than the power line point threshold as pole / tower structural component models, and output models with an interior point count greater than or equal to the power line point threshold as power line models. The final output pole / tower structural component model has fewer interior points than the power line model because transmission lines are relatively long, resulting in a longer extractable straight section and thus more interior points.

[0043] 5) Set iteration count thresholds and output quantity thresholds, and repeat the above steps multiple times until the number of repetitions reaches the iteration count threshold or the number of output models reaches the output quantity threshold. The output quantity thresholds here include two different thresholds, one for poles and one for power lines. Since there are generally only 5 power lines between two poles without splitting, while a single pole has many structural components, the number of pole structural components is much larger than that of power lines. The specific thresholds need to be set empirically. Furthermore, the termination condition here is to satisfy the two different thresholds, one for poles and one for power lines. It should be noted that the above process extracts both power lines and pole structural components in one process. This process can extract both power line models and pole structural component models, and the corresponding iteration termination condition is to satisfy the two different thresholds, one for poles and one for power lines. Alternatively, the process can be performed in two separate processes. The first process extracts only the power line model and does not extract the tower structure model. The corresponding iteration termination condition is that the number of iterations reaches the corresponding threshold or the number of output power line models reaches the preset power line model output threshold. The second process extracts only the tower structure model and does not extract the power line model. The corresponding iteration termination condition is that the number of iterations reaches the corresponding threshold or the number of output tower structure models reaches the preset tower structure output threshold.

[0044] The process of extracting tower nodes is as follows: 1) Randomly select two output tower structure models L1 and L2.

[0045] 2) Find the set of internal points that belong to both L1 and L2. If the set is not empty, proceed to the next step; otherwise, start over from step 1).

[0046] 3) Iterate through and calculate the sum of squares S of the distances from each inner point to L1 and L2, and take the inner point corresponding to the minimum S as the output tower node.

[0047] 4) Set the iteration number threshold and the output tower node number threshold, and repeat the above steps multiple times until the iteration number reaches the iteration number threshold or the output tower node number reaches the tower node number threshold.

[0048] Step 3: Calculate the rotation matrix based on the line characteristics of the electric field lines to obtain the optimal rotation matrix.

[0049] The rotation matrix is ​​calculated based on the linear characteristics of electric field lines. (From the formula...) (θ represents the angle between two lines,) and Given that the unit direction vectors of two straight lines are equal, we know that the sine of the angle between two straight lines in three-dimensional space is equal to the magnitude of the cross product of their unit direction vectors. Furthermore, when two electric field lines are parallel, the magnitude of their cross product should be 0. Therefore, to achieve optimal data matching after rotating two point cloud images, i.e., ensuring that the straight lines are essentially parallel, we only need to iteratively calculate using optimization methods to minimize the magnitude of the cross product of the unit direction vectors of the straight line pairs. That is, we need to solve for the rotation matrix R to minimize the following result: In the formula, M( R () indicates the error of the straight line pair. N This indicates the number of line pairs.

[0050] The process of solving the rotation matrix is ​​as follows: 1) Initial transformation: Set an initial rotation matrix R, which can be the identity matrix (representing no rotation) or other estimates based on prior knowledge.

[0051] 2) Finding pairs of lines: For each electric field line extracted from the source point set P, find the electric field line that is closest to it in the electric field line extracted from the target point set Q, that is, the magnitude of the cross product of the unit direction vectors of the two lines is closest to 0, and form a series of pairs of lines.

[0052] 3) Error Calculation: After transforming the source point cloud P and its electric field lines based on the current rotation matrix R, calculate the error M between all pairs of straight lines. R ).

[0053] 4) Optimize the rotation matrix: Adjust the rotation matrix according to the error to minimize the error after transformation. As a preferred approach, the least squares method can be used to adjust the rotation matrix based on the error.

[0054] 5) Iterative update: Repeat the above steps until the error is less than the preset error threshold or the number of iterations exceeds the preset maximum number of iterations.

[0055] Step 4: Calculate the translation vector based on the intersection of the towers to obtain the optimal translation vector.

[0056] 1) Initial transformation: Set an initial translation vector, which can be a zero vector (indicating no translation) or other estimates based on prior knowledge.

[0057] 2) Find the nearest point pair: For each tower node extracted from the source point set P, find the tower node with the closest distance from the tower node extracted from the target point set Q, forming a series of nearest point pairs.

[0058] 3) Error calculation: After transforming the source point cloud P and tower nodes based on the current translation vector, calculate the error between all nearest point pairs, which is the sum of squares of the Euclidean distance.

[0059] 4) Optimize the transformation matrix: Use the least squares method to adjust the translation vector based on the error, making the error smaller after transformation. As a better approach, the least squares method can be used here to adjust the translation vector based on the error.

[0060] 5) Iterative update: Repeat steps 2) to 4) until the error is less than the preset error threshold or the number of iterations exceeds the preset maximum number of iterations.

[0061] Step 5: Perform point cloud registration for transmission lines using the optimal rotation matrix and optimal translation vector.

[0062] One implementation method for a computer device: One embodiment of the computer device according to the present invention includes a memory, a processor, an internal bus, and a computer program stored in the memory. The processor and the memory communicate and interact with each other via the internal bus. The processor executes the computer program to implement the steps of the transmission line point cloud registration method of the present invention. The processor can be a microprocessor (MCU), a programmable logic device (FPGA), or other processing devices; the memory can be various types of memory that store information using electrical energy, such as RAM, ROM, etc., or other types of memory. In specific implementations, the computer device can be a terminal device or a cloud-based device.

[0063] The core idea and process of the power transmission line point cloud registration method implemented by this computer device are as follows: 1) Collect point cloud data of transmission lines to obtain source point clouds and target point clouds. Specifically, different models of acquisition equipment can be used, or the same transmission line channel can be acquired from different flight paths and acquisition angles to obtain source point clouds and target point clouds.

[0064] 2) Preprocessing of the acquired point cloud data: Point cloud preprocessing is performed on both the source point cloud and the target point cloud, including extraction of power lines and tower structural components, and extraction of tower nodes. Specifically, sample points are selected to fit a linear fitting model, and the distance between the remaining points and the linear fitting model is used to determine whether they are classified as interior points of the linear fitting model. Then, based on the set thresholds for the number of power line points and the number of tower structural components, it is determined whether the linear fitting model is a power line model or a tower structural component model.

[0065] 3) The optimal rotation matrix is ​​obtained by solving the rotation matrix based on the line characteristics of the power line, and the optimal translation vector is obtained by solving the translation vector based on the intersection of the tower. The optimal rotation matrix and translation vector are solved by iteratively solving.

[0066] 4) Use the optimal rotation matrix and optimal translation vector to perform point cloud registration for transmission lines, so as to realize point cloud registration in complex scenarios of transmission lines.

[0067] For a detailed description of the above method, please refer to the implementation method of a point cloud registration method for transmission lines.

[0068] One embodiment of a computer-readable storage medium: A computer-readable storage medium stores computer-executable instructions / programs that, when invoked and executed by a processor, cause the processor to implement a transmission line point cloud registration method according to the present invention. The core idea and process of the implemented transmission line point cloud registration method are as follows: 1) Collect point cloud data of transmission lines to obtain source point clouds and target point clouds. Specifically, different models of acquisition equipment can be used, or the same transmission line channel can be acquired from different flight paths and acquisition angles to obtain source point clouds and target point clouds.

[0069] 2) Preprocessing of the acquired point cloud data: Point cloud preprocessing is performed on both the source point cloud and the target point cloud, including extraction of power lines and tower structural components, and extraction of tower nodes. Specifically, sample points are selected to fit a linear fitting model, and the distance between the remaining points and the linear fitting model is used to determine whether they are classified as interior points of the linear fitting model. Then, based on the set thresholds for the number of power line points and the number of tower structural components, it is determined whether the linear fitting model is a power line model or a tower structural component model.

[0070] 3) The optimal rotation matrix is ​​obtained by solving the rotation matrix based on the line characteristics of the power line, and the optimal translation vector is obtained by solving the translation vector based on the intersection of the tower. The optimal rotation matrix and translation vector are solved by iteratively solving.

[0071] 4) Use the optimal rotation matrix and optimal translation vector to perform point cloud registration for transmission lines, so as to realize point cloud registration in complex scenarios of transmission lines.

[0072] For a detailed description of the above method, please refer to the implementation method of a point cloud registration method for transmission lines.

[0073] In summary, this invention achieves point cloud registration in complex scenarios of power transmission lines by extracting the intersection nodes of power lines and towers in the point cloud of power transmission lines, calculating the rotation matrix based on the line features of the power lines, calculating the translation vector based on the intersection nodes of the towers, and iteratively solving for the optimal rotation matrix and translation vector.

Claims

1. A method for point cloud registration of transmission lines, characterized in that, Includes the following steps: 1) Obtain point cloud datasets of the same transmission line, namely source point cloud and target point cloud, and extract power line and tower nodes from the source point cloud and target point cloud; 2) The optimal rotation matrix for registration is obtained by using the extracted power lines, and the optimal translation vector for registration is obtained by using the extracted tower nodes; 3) Use the optimal rotation matrix and optimal translation vector to perform point cloud registration for transmission lines.

2. The point cloud registration method for transmission lines according to claim 1, characterized in that, The process of extracting the power line includes: Repeat the following process until the number of iterations reaches a preset threshold or the number of output electric line models reaches a preset threshold: Sample points are selected from the source point cloud and the target point cloud. Lines are fitted using the selected sample points to obtain a line fitting model. The distances from other sample points in the point cloud to the line simulation model are calculated. Points with a distance less than or equal to a preset distance threshold are classified as interior points of the line fitting model, and points with a distance greater than the preset distance threshold are classified as exterior points of the line fitting model. The line fitting model with an interior point count greater than or equal to a preset power line point count threshold is output as the power line model.

3. The point cloud registration method for transmission lines according to claim 1, characterized in that, The process of extracting tower nodes includes: Extract the towers from the source point cloud and the target point cloud as tower structural component models for output. Then repeat the following process iteratively until the number of iterations reaches the preset first iteration threshold or the number of output tower nodes reaches the preset tower node number threshold: Select two output tower structure models, find the set of interior points belonging to the two selected tower structure models, and if the set of interior points is not empty, traverse the sum of the squares of the distances from each interior point in the set to the two tower structure models, and output the interior point corresponding to the smallest sum of squares as the tower node.

4. The point cloud registration method for transmission lines according to claim 3, characterized in that, The process of extracting the tower structure model includes: Repeat the following process iteratively until the number of iterations reaches the preset second iteration threshold or the number of output tower structure models reaches the preset tower structure output threshold: Sample points are selected from the source point cloud and the target point cloud. Linear fitting is performed using the selected sample points to obtain a linear fitting model. The distance from other sample points in the point cloud to the linear simulation model is calculated. Points with a distance less than a preset distance threshold are classified as interior points of the linear fitting model, and points with a distance greater than or equal to the preset distance threshold are classified as exterior points of the linear fitting model. The linear fitting model with an interior point count greater than or equal to a preset threshold for the number of pole / tower structural components and less than a preset threshold for the number of power line points is output as the pole / tower structural component model. The preset threshold for the number of pole / tower structural components is less than the preset threshold for the number of power line points.

5. The point cloud registration method for transmission lines according to claim 1, characterized in that, The process of finding the optimal rotation matrix includes: Set an initial rotation matrix, and then repeat the following process iteratively until the number of iterations reaches a preset upper limit or the error between line pairs is less than a preset error threshold, thereby obtaining the optimal rotation matrix: Find the electric field lines extracted from the target point cloud that are closest to each electric field line extracted from the source point cloud, thus obtaining a series of line pairs; use the current rotation matrix to transform the source point cloud and the electric field lines in the source point cloud, and use the transformation result to calculate the error between all line pairs; adjust the rotation matrix according to the error to reduce the error.

6. The point cloud registration method for transmission lines according to claim 1, characterized in that, The process of finding the optimal translation vector includes: Set an initial translation vector, and then repeat the following process iteratively until the number of iterations reaches a preset upper limit or the error between the nearest point pairs is less than a preset error threshold, thus obtaining the optimal translation vector: Find the tower nodes that are closest to each tower node extracted from the source point cloud from the tower node set extracted from the target point cloud, thus obtaining a series of closest point pairs; use the current translation vector to transform the source point cloud and the tower nodes in the source point cloud, and use the transformation result to calculate the error between all closest point pairs; adjust the translation vector according to the error to reduce the error.

7. The transmission line point cloud registration method according to any one of claims 1 to 6, characterized in that, The source point cloud and the target point cloud are point cloud data collected from different headings and different acquisition angles.

8. The point cloud registration method for transmission lines according to claim 5, characterized in that, The rotation matrix is ​​adjusted based on the error using the least squares method.

9. A computer device, comprising a processor, characterized in that, The processor is used to execute a computer program to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images