Power line three-dimensional reconstruction method and device, computer device, medium and product

Abstract

The application relates to a power line three-dimensional reconstruction method and device, computer equipment, medium and product, and relates to the power grid technical field. The method comprises the following steps: in response to a three-dimensional reconstruction instruction for a power line in a target power grid area, three-dimensional point cloud data of the target power grid area is acquired; based on the three-dimensional point cloud data, the target power grid area is classified according to the terrain to obtain a terrain classification result of the target power grid area; a height threshold value matched with the terrain classification result is determined, and based on the height threshold value, candidate point cloud data matched with the power line is screened out from the three-dimensional point cloud data; and based on the candidate point cloud data, the power line is three-dimensionally reconstructed to obtain a three-dimensional model of the power line, wherein the three-dimensional model is a catenary model obtained after smoothing verification and reliability verification. The method can improve the power line reconstruction accuracy in a complex terrain scene.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to a method, apparatus, computer equipment, medium, and product for three-dimensional reconstruction of power lines. Background Technology

[0002] With the rapid development of modern power grids, the planning, construction, and maintenance of transmission lines have become increasingly important. The safety, stability, and operational efficiency of power lines are directly related to the reliability of the power system. Therefore, accurate three-dimensional reconstruction and monitoring of power lines has become one of the key aspects of smart grid construction.

[0003] Traditional power line reconstruction methods mostly rely on rule-based approaches, such as OPTICS (Ordering Points To Identify the Clustering Structure) density clustering and RANSAC (Random Sample Consensus) fitting. While these methods perform well in regular scenarios, they still have significant limitations in complex scenarios, namely, the accuracy of power line reconstruction in complex terrain is currently low. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for three-dimensional power line reconstruction that can improve the accuracy of power line reconstruction in complex terrain scenarios, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for three-dimensional reconstruction of power lines, comprising: in response to a three-dimensional reconstruction command for power lines within a target power grid area, acquiring three-dimensional point cloud data of the target power grid area; classifying the terrain of the target power grid area based on the three-dimensional point cloud data to obtain a terrain classification result for the target power grid area; determining an elevation threshold that matches the terrain classification result; and, based on the elevation threshold, selecting candidate point cloud data that matches the power lines from the three-dimensional point cloud data; and performing three-dimensional reconstruction of the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0006] In one embodiment, determining an elevation threshold that matches the terrain classification result includes: segmenting the three-dimensional point cloud data into point clouds based on the terrain classification result to obtain at least one three-dimensional point cloud block; acquiring the elevation distribution characteristics of the three-dimensional point cloud set for each three-dimensional point cloud block; and determining an elevation threshold that matches the three-dimensional point cloud block based on the elevation distribution characteristics.

[0007] In one embodiment, candidate point cloud data matching power lines is selected from three-dimensional point cloud data based on an elevation threshold. This includes: performing point cloud clustering on the three-dimensional point cloud data based on the elevation threshold to obtain at least one three-dimensional point cloud class; analyzing the spatial distribution characteristics of each three-dimensional point cloud class; and using the three-dimensional point cloud data of the three-dimensional point cloud class as candidate point cloud data matching power lines if the spatial distribution characteristics meet the power line distribution characteristics.

[0008] In one embodiment, a three-dimensional reconstruction of the power line is performed based on candidate point cloud data to obtain a three-dimensional model of the power line. This includes: extracting multi-scale features from the candidate point cloud data to obtain point cloud data features at different scales; fusing the features of each point cloud data based on an attention mechanism to obtain fused features; denoising the candidate point cloud data according to the fused features to obtain denoised target point cloud data; and performing three-dimensional reconstruction of the power line based on the target point cloud data to obtain a three-dimensional model of the power line.

[0009] In one embodiment, the power line is reconstructed in three dimensions based on candidate point cloud data to obtain a three-dimensional model of the power line, including: constructing a catenary model of the power line based on the candidate point cloud data; performing a smoothing verification on the catenary model to obtain a smoothing verification result of the catenary model; performing a reliability verification on the catenary model to obtain a reliability verification result of the catenary model; and using the catenary model as the three-dimensional model of the power line if both the smoothing verification result and the reliability verification result indicate that the verification has passed.

[0010] In one embodiment, the three-dimensional reconstruction method for power lines further includes: searching for candidate points within a preset neighborhood of each of the two endpoints of the catenary model, starting from both endpoints respectively; connecting the endpoints and the candidate points if the candidate points pass the curvature check, and updating the catenary model; searching for candidate points again starting from the candidate points until the search stopping condition is met, and using the current catenary model as the three-dimensional model of the power line.

[0011] Secondly, this application also provides a three-dimensional reconstruction device for power lines, comprising: a point cloud data acquisition module, used to acquire three-dimensional point cloud data of the target power grid area in response to a three-dimensional reconstruction command for power lines within a target power grid area; a terrain classification module, used to classify the terrain of the target power grid area based on the three-dimensional point cloud data to obtain the terrain classification result of the target power grid area; a point cloud data filtering module, used to determine an elevation threshold that matches the terrain classification result, and filter candidate point cloud data that matches the power lines from the three-dimensional point cloud data based on the elevation threshold; and a three-dimensional reconstruction module for power lines, used to perform three-dimensional reconstruction of the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing verification and reliability verification.

[0012] Thirdly, this application also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: in response to a three-dimensional reconstruction instruction for power lines within a target power grid area, acquiring three-dimensional point cloud data of the target power grid area; classifying the terrain of the target power grid area based on the three-dimensional point cloud data to obtain a terrain classification result for the target power grid area; determining an elevation threshold that matches the terrain classification result; and filtering candidate point cloud data that matches the power lines from the three-dimensional point cloud data based on the elevation threshold; and performing three-dimensional reconstruction of the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0013] Fourthly, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, performs the following steps: in response to a three-dimensional reconstruction instruction for power lines within a target power grid area, acquires three-dimensional point cloud data of the target power grid area; based on the three-dimensional point cloud data, performs terrain classification on the target power grid area to obtain terrain classification results for the target power grid area; determines an elevation threshold that matches the terrain classification results, and based on the elevation threshold, filters candidate point cloud data that matches the power lines from the three-dimensional point cloud data; and based on the candidate point cloud data, performs three-dimensional reconstruction on the power lines to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0014] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps: in response to a three-dimensional reconstruction instruction for power lines within a target power grid area, acquires three-dimensional point cloud data of the target power grid area; based on the three-dimensional point cloud data, performs terrain classification on the target power grid area to obtain terrain classification results for the target power grid area; determines an elevation threshold that matches the terrain classification results; based on the elevation threshold, filters candidate point cloud data that matches the power lines from the three-dimensional point cloud data; and performs three-dimensional reconstruction on the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0015] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for 3D power line reconstruction respond to a 3D reconstruction command for power lines within a target power grid area by acquiring 3D point cloud data of the target power grid area. Based on the 3D point cloud data, the target power grid area is classified into different terrain types to obtain terrain classification results, and an elevation threshold matching the terrain classification results is determined. Candidate point cloud data matching the power lines is then selected from the 3D point cloud data based on the elevation threshold. Finally, based on the candidate point cloud data, the power lines are 3D reconstructed to obtain a 3D model of the power lines, which is a catenary model obtained after smoothing and reliability verification. Thus, this scheme dynamically matches the most suitable elevation threshold based on the terrain classification results, thereby selecting more accurate candidate point cloud data through the elevation threshold, and ultimately achieving 3D reconstruction of power lines in the target power grid area. This overcomes the problem of low accuracy in complex terrain scenarios of traditional rule-based power line reconstruction methods, significantly improving the accuracy and robustness of 3D power line reconstruction in complex terrain scenarios. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is an application environment diagram of the three-dimensional reconstruction method for power lines in one embodiment;

[0018] Figure 2 This is a flowchart illustrating a three-dimensional reconstruction method for power lines in one embodiment;

[0019] Figure 3 This is a flowchart illustrating the process of determining an elevation threshold in one embodiment;

[0020] Figure 4 This is a flowchart illustrating point cloud clustering in one embodiment;

[0021] Figure 5 This is a schematic diagram of the point cloud data denoising process in one embodiment;

[0022] Figure 6 This is a schematic diagram of a deep learning network model in one embodiment;

[0023] Figure 7 This is a flowchart illustrating the verification process for a catenary model in one embodiment.

[0024] Figure 8 This is a schematic diagram of the bidirectional growth process in one embodiment;

[0025] Figure 9 This is a flowchart illustrating a specific embodiment of a three-dimensional reconstruction method for power lines.

[0026] Figure 10 This is a schematic diagram of the process for reconstructing a three-dimensional model of a power line in a specific embodiment;

[0027] Figure 11 This is a structural block diagram of a three-dimensional power line reconstruction device in one embodiment;

[0028] Figure 12 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0030] It should be noted that the terms "comprising" and "having," and any variations thereof, as used in this application, are intended to cover non-exclusive inclusion. The term "multiple" as used in this application refers to two or more.

[0031] The three-dimensional reconstruction method for power lines provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, and IoT devices. Server 104 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server providing cloud computing services.

[0032] Specifically, in response to a user's 3D reconstruction command for power lines within a target power grid area triggered by terminal 102, server 104 acquires 3D point cloud data of the target power grid area; based on the 3D point cloud data, it performs terrain classification on the target power grid area to obtain terrain classification results; it determines an elevation threshold matching the terrain classification results, and based on the elevation threshold, it filters candidate point cloud data matching the power lines from the 3D point cloud data; based on the candidate point cloud data, it performs 3D reconstruction on the power lines to obtain a 3D model of the power lines, wherein the 3D model is a catenary model obtained after smoothing and reliability verification. Alternatively, server 104 can automatically generate 3D reconstruction commands for power lines within a target power grid area without requiring terminal 102 to trigger them.

[0033] In one exemplary embodiment, such as Figure 2 As shown, a method for three-dimensional reconstruction of power lines is provided, which can be applied to... Figure 1 Taking server 104 as an example, the following steps are included:

[0034] Step S202: In response to the three-dimensional reconstruction command for power lines within the target power grid area, acquire three-dimensional point cloud data of the target power grid area.

[0035] The target power grid area can refer to the area covered by power lines to be reconstructed in 3D, such as the area where high-voltage power lines are located in a mountainous region. Power lines refer to overhead transmission lines. 3D reconstruction of power lines can be understood as the process of building a 3D model of the power lines. Through 3D reconstruction, visualized monitoring of power lines can be achieved, improving operation and maintenance efficiency and reducing safety hazards. 3D point cloud data refers to a set of discrete points in 3D space. These discrete points can correspond to various land features such as terrain, buildings, vegetation, and power lines. 3D point cloud data typically includes, but is not limited to, at least one of the following: 3D coordinates of each point, laser reflection intensity, and timestamp. The specific details can be determined according to the actual situation, and this embodiment does not impose any limitations on this. In one example, the 3D point cloud data of the target power grid area can be represented as: ,in, This represents the three-dimensional spatial coordinates of point i, namely longitude, latitude, and elevation. The value represents the laser reflection intensity at point i, and N represents the total number of points in the three-dimensional point cloud. The point density typically ranges from 5 to 50 (points / square meter).

[0036] For example, in response to a 3D reconstruction command for power lines within a target power grid area, the server first needs to acquire 3D point cloud data of that target power grid area. Specifically, the server can acquire the 3D point cloud data of the target power grid area using at least one data acquisition device, such as a drone, airborne LiDAR, or ground-based 3D scanning equipment. That is, the server sends a command to the data acquisition device to acquire the 3D point cloud data of the target power grid area, and after receiving the command, the data acquisition device can send the 3D point cloud data of the target power grid area, which it has scanned in real time, to the server.

[0037] In a preferred example, UAV equipment is suitable for small-area scanning, while ground-based 3D scanning equipment is mainly used for high-precision mapping of fixed scenarios such as substations. Airborne LiDAR, on the other hand, can quickly acquire 3D spatial information of power lines and their surrounding environment over a large area. It obtains high-precision 3D point cloud data of the target power grid area through laser ranging technology and completes scanning of large-area 3D information in a short time. Compared with traditional optical imaging, airborne LiDAR can penetrate vegetation and obtain point cloud data in complex environments, making it particularly suitable for the identification and modeling of slender targets such as power lines. Therefore, in this embodiment, airborne LiDAR is preferred as the 3D point cloud data acquisition device.

[0038] Step S204: Based on the 3D point cloud data, perform terrain classification on the target power grid area to obtain the terrain classification result of the target power grid area.

[0039] Terrain classification can be understood as dividing the target power grid area into different terrain categories, such as flat areas, mountainous areas, and complex terrain areas where power lines cross roads, rivers, etc. The specific terrain category can be determined according to the actual geographical conditions, and this embodiment does not impose any restrictions on it. The terrain classification result may include, but is not limited to, at least one of the following: terrain category label of the target power grid area, classification timestamp, and regional identifier of the target power grid area.

[0040] For example, due to the massive volume of 3D point cloud data, directly using 3D point cloud data for terrain classification requires high computational overhead. Therefore, to improve computational efficiency and reduce computational overhead, in this embodiment, after acquiring the 3D point cloud data of the target power grid area, the server can first project the 3D point cloud data into a 2D space to obtain a 2D elevation map, where each element represents the elevation value of a pixel. The 2D elevation map is then used as input to the terrain classification model, outputting the terrain classification result of the target power grid area. The terrain classification model is a lightweight convolutional neural network model used for terrain classification, such as the ResNet model; this embodiment does not limit the specific type of model.

[0041] Specifically, the server meshes the 3D point cloud data on a horizontal plane at a fixed resolution. In this embodiment, the specific value of the fixed resolution is not limited; it can be 0.5 meters or other values. For each grid point after meshing, the maximum elevation value of its grid area is calculated using the following expression:

[0042]

[0043] Where H represents a two-dimensional elevation map, R represents the set of real numbers. The matrix has dimensions of M rows and M columns; in this implementation, M=256, resulting in a 256×256 resolution two-dimensional elevation map. u represents the horizontal coordinate, v represents the vertical coordinate, and grid(u,v) represents the position of each pixel in the image. Using the maximum elevation value preserves terrain contour features and ignores interference such as low-lying vegetation.

[0044] The obtained two-dimensional elevation map is used as input to the terrain classification model, and the output is the terrain category label, which is expressed as follows: Where f represents the terrain classification model, and i represents the terrain category. For example, when i=1, the terrain category is flat; when i=2, it is mountainous; and when i=3, it is an intersecting area. The terrain classification model can be trained using cross-entropy loss, expressed as: ,in, represents the cross-entropy loss value, used to measure the difference between the predicted probability distribution and the true distribution; the smaller the value, the more accurate the classification. K represents the number of terrain categories. This represents the probability of the k-th class predicted by the model. This represents the actual category label.

[0045] Step S206: Determine the elevation threshold that matches the terrain classification results. Based on the elevation threshold, filter out candidate point cloud data that matches the power lines from the 3D point cloud data.

[0046] The elevation threshold is used to filter 3D point cloud data within a specific height range. Specifically, the elevation threshold serves as the neighborhood radius for subsequent clustering of the target power grid area, determining the calculation range of the core distance and thus whether a sample point is a core point. Understandably, a smaller neighborhood radius leads to more detailed clustering results but may result in too many clusters, while a larger neighborhood radius reduces the number of clusters but may lead to poor clustering results. Therefore, to ensure clustering effectiveness, an optimal neighborhood radius, i.e., the elevation threshold, is needed. In this embodiment, a suitable elevation threshold is determined based on the terrain classification results, ensuring a good match between the elevation threshold and the actual terrain. This makes the final clustering result more closely resemble the actual geographical scene, laying the foundation for accurate 3D reconstruction of power lines.

[0047] For example, a target power grid area may have multiple terrain categories, meaning the terrain classification results obtained by the server can include multiple terrain category labels. For each power grid area corresponding to a terrain category label within the target power grid area, the server can determine an elevation threshold suitable for that power grid area. Based on this elevation threshold, the server can cluster the 3D point cloud data of the power grid area and filter out candidate point cloud data that match the power lines from the 3D point cloud data.

[0048] Step S208: Based on the candidate point cloud data, perform three-dimensional reconstruction of the power line to obtain a three-dimensional model of the power line. The three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0049] For example, after obtaining candidate point cloud data, the server can use the candidate point cloud data to construct a catenary model of the power line. After the catenary module undergoes smoothness and reliability verification, it can be used as the three-dimensional model of the power line. It can be understood that a catenary model is a mathematical model describing the curve shape formed by a flexible chain or chain suspended under gravity, with both ends fixed and of uniform mass. Since the catenary model best matches the power line model, this embodiment uses the catenary model as the three-dimensional model of the power line.

[0050] In this embodiment, in response to a 3D reconstruction command for power lines within a target power grid area, 3D point cloud data of the target power grid area is acquired. Based on the 3D point cloud data, the target power grid area is classified into terrain types to obtain terrain classification results, and an elevation threshold matching the terrain classification results is determined. Therefore, candidate point cloud data matching the power lines is filtered from the 3D point cloud data based on the elevation threshold. Finally, based on the candidate point cloud data, the power lines are reconstructed in 3D to obtain a 3D model of the power lines. This 3D model is a catenary model obtained after smoothing and reliability verification. Thus, this embodiment dynamically matches the most suitable elevation threshold based on the terrain classification results, thereby filtering more accurate candidate point cloud data through the elevation threshold, and ultimately achieving 3D reconstruction of power lines in the target power grid area. This overcomes the problem of low accuracy in complex terrain scenarios of traditional rule-based power line reconstruction methods, significantly improving the accuracy and robustness of 3D power line reconstruction in complex terrain scenarios.

[0051] In one exemplary embodiment, such as Figure 3 As shown, determining the elevation threshold that matches the terrain classification results includes:

[0052] Step S302: Based on the terrain classification results, the three-dimensional point cloud data is segmented to obtain at least one three-dimensional point cloud block.

[0053] Point cloud segmentation refers to the process of dividing 3D point cloud data into different parts according to the terrain classification results. 3D point cloud blocks are the subsets of point clouds obtained after segmentation. Each 3D point cloud block represents a region with similar terrain features, and there is a one-to-one correspondence between 3D point cloud blocks and terrain category labels.

[0054] For example, after obtaining the terrain classification result, the server can perform point cloud segmentation on the original 3D point cloud data based on the terrain classification result, or based on the terrain category label in the terrain classification result, so as to obtain a 3D point cloud block that matches each terrain category label.

[0055] Step S304: For each 3D point cloud block, obtain the elevation distribution characteristics of the 3D point cloud set.

[0056] Here, a 3D point cloud set refers to the set of 3D point clouds composed of each 3D point cloud block. Elevation distribution characteristics can refer to the distribution characteristics of the 3D point cloud set with respect to elevation, and may include, but are not limited to, at least one of the following: elevation mean, elevation standard deviation, etc.

[0057] For example, after segmenting the 3D point cloud into blocks, each 3D point cloud block can have a corresponding 3D point cloud set. For each 3D point cloud set, the server can perform elevation statistical analysis, that is, analyze the elevation value of each point in the 3D point cloud set, and calculate the elevation mean and elevation standard deviation of the 3D point cloud set. The calculation expressions are as follows:

[0058]

[0059]

[0060] in, This represents the average elevation of the j-th 3D point cloud block. This represents the standard deviation of the elevation of the j-th 3D point cloud block. This represents the elevation value of the i-th 3D point cloud within the j-th 3D point cloud block. This represents the number of 3D point clouds in the j-th 3D point cloud block.

[0061] Step S306: Determine the elevation threshold that matches the 3D point cloud block based on the elevation distribution characteristics.

[0062] For example, after calculating the elevation distribution characteristics of each 3D point cloud block, the server can input these characteristics into a mapping function based on machine learning such as SVM (Support Vector Machine) or Random Forest. This function, based on the elevation distribution characteristics, can adaptively output an optimal elevation threshold as the optimal neighborhood radius for subsequent 3D point cloud clustering. In one example, the expression for calculating the elevation threshold is: ,in, Let f represent the elevation threshold, and f represent the mapping function. For example, in flat areas where power lines are densely packed, fine clustering is needed, and the elevation threshold can be reduced; while in mountainous areas and areas where power lines cross, the elevation threshold can be appropriately increased to avoid the terrain undulations disrupting the power lines. Of course, the specific strategy for determining the elevation threshold can be set according to the actual situation, and this implementation does not impose any restrictions on it.

[0063] In this embodiment, the elevation threshold is dynamically generated by machine learning algorithm, which can better adapt the elevation distribution characteristics of the three-dimensional point cloud of power lines under different terrain conditions, and lay a good data foundation for subsequent three-dimensional point cloud clustering.

[0064] In one exemplary embodiment, such as Figure 4 As shown, based on an elevation threshold, candidate point cloud data matching the power lines are selected from the 3D point cloud data, including:

[0065] Step S402: Based on the elevation threshold, perform point cloud clustering on the 3D point cloud data to obtain at least one 3D point cloud class.

[0066] Point cloud clustering refers to dividing 3D point cloud data into different categories or clusters. Each category or cluster is a 3D point cloud class, and the point cloud data in each 3D point cloud class have similar characteristics.

[0067] For example, after obtaining the elevation threshold, the server can use it to perform point cloud clustering on the 3D point cloud data, thereby obtaining multiple 3D point cloud classes with similar characteristics. Specifically, the server can perform point cloud clustering on the 3D point cloud set of each 3D point cloud block based on the elevation threshold matching each 3D point cloud block, thereby obtaining at least one 3D point cloud class in the 3D point cloud block, until all 3D point cloud blocks have completed point cloud clustering.

[0068] In one example, the server can use an improved OPTICS density clustering algorithm to cluster 3D point clouds. In this embodiment, the improvement lies in the neighborhood radius being dynamically determined based on the terrain classification results. Specifically, the OPTICS density clustering algorithm calculates the core distance and reachability distance for each point. The core distance is the minimum radius required for a point to become a core point, and the reachability distance is the density connectivity cost from a core point to other points. The reachability distance describes the density correlation between a point and other points in its neighborhood. By traversing all points and sorting them according to density reachability, a one-dimensional reachability map is generated. This map forms distinct "valleys" at points of abrupt density drops, which can be used to segment the cluster structure. Compared to traditional clustering methods, OPTICS does not rely on a fixed number of clusters and can better characterize the local density structure of point clouds, making it particularly suitable for processing data with non-uniform density distributions.

[0069] Step S404: For each 3D point cloud class, analyze the spatial distribution characteristics of the 3D point cloud class.

[0070] Among them, spatial distribution characteristics may include, but are not limited to, at least one of the following: whether the density of a three-dimensional point cloud is continuous, whether the distribution is elongated, or whether the elevation is prominent.

[0071] For example, for each 3D point cloud class, the server can further analyze the characteristics of the 3D point cloud class, such as whether the density is continuous, whether the distribution is elongated, and whether the elevation is prominent, so as to facilitate the subsequent screening of candidate point cloud data.

[0072] Step S406: If the spatial distribution characteristics meet the conditions of the power line distribution characteristics, the three-dimensional point cloud data of the three-dimensional point cloud class is used as candidate point cloud data to match the power line.

[0073] Among them, the power line distribution characteristic conditions can be rule conditions used to determine whether a three-dimensional point cloud belongs to a power line, such as at least one of the following: continuous density, slender distribution, prominent elevation, etc.

[0074] For example, after obtaining the spatial distribution features of each 3D point cloud class, the server can compare these features with preset power line distribution feature conditions. If the spatial distribution features indicate that the 3D point cloud class has continuous density, a slender distribution, and prominent elevation, then the spatial distribution features meet the power line distribution feature conditions. In this case, the 3D point cloud data of this class can be used as candidate point cloud data to match power lines. Conversely, if the spatial distribution features indicate that the 3D point cloud class has sparse density, a non-slender distribution, or an abnormal elevation at least once, then the spatial distribution features do not meet the power line distribution feature conditions. In this case, the 3D point cloud data of this class can be identified as outliers and removed.

[0075] In this embodiment, point cloud clustering is performed on the 3D point cloud data, and the spatial distribution characteristics of each 3D point cloud class are analyzed. If the spatial distribution characteristics meet the conditions for power line distribution characteristics, the 3D point cloud data of that class is retained as candidate point cloud data. This significantly reduces the amount of point cloud data and lowers the computational complexity. Simultaneously, it effectively improves the accuracy and robustness of candidate point cloud data selection, providing more accurate and reliable data support for subsequent 3D reconstruction and significantly enhancing the accuracy of power line reconstruction in complex environments.

[0076] In one exemplary embodiment, such as Figure 5 As shown, based on candidate point cloud data, a 3D reconstruction of the power line is performed to obtain a 3D model of the power line, including:

[0077] Step S502: Perform multi-scale feature extraction on the candidate point cloud data to obtain the point cloud data features of the candidate point cloud data at different scales.

[0078] Multi-scale feature extraction refers to extracting local feature vectors of the same point cloud at different scales. Different scales can correspond to different neighborhood search radii. Point cloud data features are mathematical expressions used to describe the local geometric properties of point cloud data, such as electric field line thickness, sag amplitude, curvature, density gradient, etc. Each scale can output an independent set of features.

[0079] For example, after filtering out candidate point cloud data, the server can extract features from the candidate point cloud data to obtain the point cloud data features at different scales. Specifically, in order to further model the multi-scale differences in the local ensemble structure of the point cloud, this embodiment can introduce a three-branch dynamic graph convolution module, such as... Figure 6 As shown, each dynamic graph convolutional module can be set with a different neighborhood search radius. For example, the neighborhood search radius of dynamic graph convolutional module 1 can be set to 8, the neighborhood search radius of dynamic graph convolutional module 2 can be set to 16, and the neighborhood search radius of dynamic graph convolutional module 3 can be set to 32. In each branch, for each point cloud... In its k nearest neighbor Constructing local features:

[0080] e ij (k) = ϕ (k) ([ a i , a j - a i ]),j∈ N k i

[0081] in, For branch k, a shared MLP (Multilayer Perceptron). Represents a relative position vector. Let k represent the k nearest neighbors of point i. Max pooling is used to aggregate the features of each neighborhood to obtain enhanced representations at various scales:

[0082]

[0083] Finally, the features from the three scales are concatenated, and then multi-level features are aggregated through an MLP layer: , These are the features of point cloud data.

[0084] In one example, such as Figure 6 As shown, before multi-scale feature extraction, the server can also extract global basic features of the candidate point cloud data. Specifically, the server can input the candidate point cloud data into a feature extractor, such as the PointNet++ feature extractor, to perform global feature extraction, thereby obtaining the global basic features of the candidate point cloud data for subsequent enhancement.

[0085] Step S504: Based on the attention mechanism, feature fusion is performed on the features of each point cloud data to obtain fused features.

[0086] Among them, fusion features refer to the features obtained by fusing point cloud data features at different scales based on the attention mechanism.

[0087] For example, to further enhance the global dependency modeling and redundancy compression of point cloud data features, this embodiment designs an attention mechanism to fuse the features of each point cloud data. Specifically, firstly, the point cloud data features are mapped using three independent shared MLPs to obtain query, key, and value vectors, where:

[0088]

[0089] vector Transpose and merge vectors Perform a matrix dot product and divide by the scaling factor. We then perform the softmax function calculation to construct the attention weight matrix:

[0090]

[0091] in, This represents the degree of dependence of point i on point j. Then, weighted aggregation is performed to obtain the fused features:

[0092]

[0093] Step S506: Based on the fusion features, the candidate point cloud data is denoised to obtain the denoised target point cloud data.

[0094] Among them, target point cloud data refers to 3D point cloud data after noise reduction processing.

[0095] For example, after obtaining the fused features, a residual structure can be used to fuse the fused features. With original features Adding together, that is: This preserves original local features, avoids information loss, and improves information stability and gradient propagation effectiveness. Furthermore, the features... The input is a two-layer MLP for denoising point coordinate prediction, i.e.: Output the denoised point cloud coordinates This achieves point cloud noise reduction.

[0096] In one example, feature extraction, feature fusion, and feature denoising can all be deployed in the same deep learning network model.

[0097] Step S508: Based on the target point cloud data, perform three-dimensional reconstruction of the power line to obtain a three-dimensional model of the power line.

[0098] For example, after obtaining the denoised target point cloud data, the target point cloud data can be used to perform three-dimensional reconstruction of the power line, thereby obtaining a three-dimensional model of the power line.

[0099] In this embodiment, multi-scale feature extraction is performed on candidate point cloud data, and feature fusion is performed based on an attention mechanism. Finally, the candidate point cloud data is denoised based on the fused features, which realizes high-precision feature representation and robust denoising processing of power line candidate point cloud data, improves the accuracy of point cloud data, and lays a data foundation for subsequent three-dimensional reconstruction of power lines.

[0100] In one exemplary embodiment, such as Figure 7As shown, based on candidate point cloud data, a 3D reconstruction of the power line is performed to obtain a 3D model of the power line, including:

[0101] Step S702: Based on the candidate point cloud data, construct the catenary model of the power line.

[0102] For example, the server can use the RANSAC (Random Sample Consensus) fitting method to fit the geometry of the power line, thereby constructing a catenary model. Specifically, the server can randomly select two points from the candidate point cloud data as initial sample points, calculate the initial parameters of the catenary model, and thus construct the corresponding catenary model. The initial parameters determine the steepness of the catenary model, and their expression is:

[0103]

[0104] in, Indicates the horizontal distance between two points. This represents the elevation difference between two points. () represents the inverse hyperbolic sine function. Using initial parameters... This will generate the corresponding electric field line catenary model:

[0105]

[0106] Step S704: Perform a smoothing verification on the catenary model to obtain the smoothing verification result of the catenary model.

[0107] Among them, smoothing verification can be understood as the process of verifying the continuity of the catenary curvature. The smoothing verification result is used to characterize whether the catenary curvature is continuous. It can include smoothing verification passing and smoothing verification failing. Smoothing verification passing means that the catenary curvature is continuous, while smoothing verification failing means that the catenary curvature is discontinuous and may have abrupt change points or broken segments.

[0108] For example, after constructing the catenary model, the server needs to validate the model. Smoothness validation is one such validation, used to verify whether the curvature of the catenary is continuous. Specifically, the server can sample the catenary at preset intervals, such as 0.2 meters, and calculate the curvature change of adjacent points. If the curvature change is greater than a preset curvature threshold, the line segment formed by the adjacent points is considered not smooth, and a new catenary model needs to be constructed by re-selecting sample points. If the curvature change is less than or equal to the preset curvature threshold, the line segment formed by the adjacent points is considered smooth, i.e., the validation passes. The preset curvature threshold is a pre-set curvature change threshold, such as 0.15, which can be set according to actual conditions. This embodiment does not limit the specific value of the preset curvature threshold.

[0109] In one example, the expression for calculating the curvature change is as follows:

[0110]

[0111] in, This represents the change in curvature between adjacent points. Denotes the curvature at point i+1. Denotes the curvature at point i. Indicates the sampling interval.

[0112] Step S706: Perform reliability verification on the catenary model to obtain the reliability verification result of the catenary model.

[0113] Reliability verification can be understood as the process of verifying the reliability of the catenary model, specifically verifying whether the catenary model conforms to the spatial layout and morphological characteristics of power lines. The reliability verification result is used to characterize whether the catenary model is reliable, and it can include reliability verification passing or failing.

[0114] For example, reliability verification can be achieved using a 3D-CNN (3D Convolutional Neural Network) model. Unlike traditional 2D CNNs, 3D-CNNs process 3D data using 3D convolutional kernels, capturing changes in data over time for better analysis of dynamic processes. The 3D-CNN network structure includes 3D convolutional layers, max-pooling layers, and fully connected layers. Specifically, the server inputs the catenary model into the 3D-CNN model. The 3D-CNN model normalizes the point cloud count, elevation, and laser reflection intensity of the corresponding voxel mesh, obtaining density, elevation, and laser reflection intensity features. The fully connected layer outputs a reliability score for the catenary model based on these features. In one example, the reliability score ranges from [0, 1]. If the reliability score exceeds a specified value, the catenary model is considered to have passed reliability verification.

[0115] It is understandable that this implementation does not impose any restrictions on the execution order of smoothness verification and reliability verification.

[0116] Step S708: If both the smoothness verification result and the reliability verification result indicate that the verification has passed, the catenary model is used as the three-dimensional model of the power line.

[0117] For example, if both the smoothing verification result and the reliability verification result indicate that the verification passed, the catenary model can be used as the three-dimensional model of the power line. Conversely, if at least one of the smoothing verification result or the reliability verification result indicates that the verification failed, it means that the two randomly selected initial sample points cannot fit the catenary model well, and the sample points need to be selected again.

[0118] In one embodiment, the server can also set an initial number of iterations and calculate the proportion of interior points of the catenary model in real time, using a dynamic adjustment strategy to optimize the number of fitting iterations. For example, setting an initial number of iterations... Real-time calculation of the interior point ratio of the catenary model: ,in, This indicates the number of point clouds that passed the verification. This represents the total number of candidate point clouds. The formula for adjusting the number of iterations is as follows:

[0119]

[0120] in, This represents the number of iterations after optimization. Thus, by adaptively adjusting the number of iterations based on the current fit, a stable and accurate catenary model can be output within a finite time.

[0121] In this embodiment, by performing smoothing and reliability checks on the catenary model, the accuracy of the catenary model can be effectively improved.

[0122] In one exemplary embodiment, such as Figure 8 As shown, the three-dimensional reconstruction method for power lines also includes:

[0123] Step S802: Starting from the two endpoints of the catenary model, search for candidate points within the preset neighborhood range of each endpoint.

[0124] It should be noted that after obtaining a suitable catenary model, some sample points may be lost. Therefore, this embodiment introduces a bidirectional growth strategy for catenary extension. That is, starting from the current catenary model, it extends along both the left and right ends, thereby automatically realizing the completion, structural optimization, and high-precision modeling of the power line point cloud.

[0125] Here, the endpoints refer to the left and right endpoints of the catenary model, which are the starting points for growth and expansion. The preset neighborhood range is a predefined search range, such as a spherical neighborhood with a radius of 0.3 meters, used to find candidate points. Candidate points are point clouds existing within the preset neighborhood range, which may be extensions of power lines or noise points.

[0126] For example, after obtaining the catenary model, the server can further initiate the extension of the power line from both the left and right endpoints of the catenary model simultaneously. Specifically, starting from the left endpoint of the catenary model, it searches for possible new candidate points within its preset neighborhood. Simultaneously, starting from the right endpoint of the catenary model, it searches for possible new candidate points within its preset neighborhood.

[0127] Step S804: If the candidate point passes the curvature check, connect the endpoint and the candidate point, and update the catenary model.

[0128] For example, for each candidate point, the server calculates its local curvature and compares it with the curvature of the current catenary model. If the curvature difference is less than a preset curvature threshold, the difference between the local curvature and the current catenary model curvature is considered small, the curvature verification passes, and the point can be included in the catenary model, i.e., the endpoint and the candidate point are connected, and the catenary model is updated. Conversely, if the curvature difference is greater than or equal to the preset curvature threshold, the difference between the local curvature and the current catenary model curvature is considered large, the curvature verification fails, and the curvature verification continues for the next candidate point. The preset curvature threshold is a baseline value pre-set for the difference between the local curvature and the curvature of the current catenary model, such as 0.1 meters. Of course, other values ​​can be set according to actual conditions, and this embodiment does not impose any restrictions on this.

[0129] Step S806: Starting from the candidate point, perform the candidate point search again until the search stopping condition is met, and use the current catenary model as the three-dimensional model of the electric power line.

[0130] Among them, the search stopping condition refers to the precondition for stopping the search for candidate points, such as no candidate points within the preset neighborhood, or all candidate points within the preset neighborhood have failed the curvature verification.

[0131] For example, after finding a suitable candidate point, the server can continue to use the candidate point as the starting point for a new round of candidate point search until there are no candidate points in the preset neighborhood, or when all candidate points in the preset neighborhood fail the curvature verification, the search stops and the current catenary model is output as the final three-dimensional model of the power line.

[0132] In one embodiment, after completing the bidirectional growth of the point cloud, the expanded 3D point cloud data can be remodeled structurally to update the catenary parameters. That is, by optimizing the initial parameters... This allows the expanded 3D point cloud to match the ideal catenary model as closely as possible, thus achieving physical consistency modeling of electric power lines in 3D space. The fitting objective function is:

[0133]

[0134] in The natural hanging shape of the catenary represents a typical structure of electric field lines under gravity. This optimization process can be viewed as a least-squares fit of the catenary model parameters, ensuring that the model remains physically consistent with the sampled data, thus enhancing its physical interpretability and fitting accuracy.

[0135] In this embodiment, by extending the catenary model bidirectionally, automatic completion, structural optimization, and high-precision modeling of the power line point cloud are achieved.

[0136] Figure 9 A flowchart of a specific embodiment of 3D power line reconstruction is shown. In this embodiment, in response to a 3D reconstruction command for power lines within a target power grid area, the server first acquires 3D point cloud data of the target power grid area. Then, the 3D point cloud data is projected onto a 2D space to obtain a 2D elevation map. This 2D elevation map is used as input to a terrain classification model, outputting a terrain classification result. Based on the terrain classification result, the original 3D point cloud data is segmented into multiple 3D point cloud blocks. The elevation mean and standard deviation of the 3D point cloud set for each block are calculated. The elevation mean and standard deviation are input into a mapping function, outputting an elevation threshold, which serves as the optimal neighborhood radius for subsequent 3D point cloud clustering.

[0137] Based on an elevation threshold, the OPTICS density clustering algorithm is used to cluster the 3D point cloud data, resulting in at least one 3D point cloud class. The spatial distribution characteristics of each 3D point cloud class are analyzed. If the spatial distribution characteristics meet the power line distribution characteristics, the 3D point cloud data of the 3D point cloud class is used as candidate point cloud data that matches the power lines. Points whose spatial distribution characteristics do not meet the power line distribution characteristics are considered outliers and are removed.

[0138] After obtaining the candidate point cloud data, global feature extraction and multi-scale feature extraction are further performed on the candidate point cloud data to obtain the point cloud data features of the candidate point cloud data at different scales; based on the attention mechanism, feature fusion is performed on each point cloud data feature to obtain fused features; based on the fused features, the candidate point cloud data is denoised to obtain the denoised target point cloud data.

[0139] Figure 10 A schematic diagram illustrating the process of reconstructing a 3D model of power lines is shown. Combined with... Figure 9 and Figure 10As can be seen, this embodiment uses the RANSAC fitting method to analyze the target point cloud data to construct a catenary model of the power line. Specifically, firstly, two points are randomly selected from the target point cloud data as initial sample points to construct the corresponding catenary model. Next, a smoothing verification is performed on the catenary model to obtain the smoothing verification result. Then, a reliability verification is performed on the catenary model to obtain the reliability verification result. The order of the smoothing verification and reliability verification is not restricted. If both the smoothing verification result and the reliability verification result indicate that the verification has passed, the catenary model is updated. If the current iteration count has been reached, the loop stops, and the point cloud data is output; if the iteration count has not been reached, the point cloud data is updated, and the catenary model is resampled to construct it. If at least one of the smoothing verification result or the reliability verification result indicates that the verification has failed, the sample points are resampled.

[0140] After obtaining a suitable catenary model, candidate points are searched within the preset neighborhood range of each end point, starting from both ends of the catenary model. If a candidate point passes the curvature check, the end points and the candidate point are connected, and the catenary model is updated. The candidate point search is performed again starting from the candidate point until the search stops, and the current catenary model is used as the three-dimensional model of the electric power line.

[0141] In this embodiment, in response to a 3D reconstruction command for power lines within a target power grid area, 3D point cloud data of the target power grid area is acquired. Based on the 3D point cloud data, the target power grid area is classified into terrain types to obtain terrain classification results, and an elevation threshold matching the terrain classification results is determined. Therefore, candidate point cloud data matching the power lines is filtered from the 3D point cloud data based on the elevation threshold. Finally, based on the candidate point cloud data, the power lines are reconstructed in 3D to obtain a 3D model of the power lines. This 3D model is a catenary model obtained after smoothing and reliability verification. Thus, this embodiment dynamically matches the most suitable elevation threshold based on the terrain classification results, thereby filtering more accurate candidate point cloud data through the elevation threshold, and ultimately achieving 3D reconstruction of power lines in the target power grid area. This overcomes the problem of low accuracy in complex terrain scenarios of traditional rule-based power line reconstruction methods, significantly improving the accuracy and robustness of 3D power line reconstruction in complex terrain scenarios.

[0142] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0143] Based on the same inventive concept, this application also provides a three-dimensional power line reconstruction apparatus for implementing the aforementioned three-dimensional power line reconstruction method. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more embodiments of the three-dimensional power line reconstruction apparatus provided below can be found in the limitations of the three-dimensional power line reconstruction method described above, and will not be repeated here.

[0144] In one exemplary embodiment, such as Figure 11 As shown, a three-dimensional reconstruction device for power lines is provided, comprising: a point cloud data acquisition module 1102, used to acquire three-dimensional point cloud data of the target power grid area in response to a three-dimensional reconstruction command for power lines within a target power grid area; a terrain classification module 1104, used to classify the terrain of the target power grid area based on the three-dimensional point cloud data to obtain the terrain classification result of the target power grid area; a point cloud data filtering module 1106, used to determine an elevation threshold that matches the terrain classification result, and based on the elevation threshold, filter out candidate point cloud data that matches the power lines from the three-dimensional point cloud data; and a three-dimensional reconstruction module 1108, used to perform three-dimensional reconstruction of the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing verification and reliability verification.

[0145] In one embodiment, the point cloud data filtering module 1106 is further configured to: perform point cloud segmentation on the three-dimensional point cloud data according to the terrain classification results to obtain at least one three-dimensional point cloud block; for each three-dimensional point cloud block, obtain the elevation distribution characteristics of the three-dimensional point cloud set; and determine the elevation threshold that matches the three-dimensional point cloud block based on the elevation distribution characteristics.

[0146] In one embodiment, the point cloud data filtering module 1106 is further configured to: perform point cloud clustering on the three-dimensional point cloud data based on an elevation threshold to obtain at least one three-dimensional point cloud class; analyze the spatial distribution characteristics of each three-dimensional point cloud class; and, if the spatial distribution characteristics meet the power line distribution characteristics, use the three-dimensional point cloud data of the three-dimensional point cloud class as candidate point cloud data that matches the power line.

[0147] In one embodiment, the power line 3D reconstruction module 1108 is further configured to: extract multi-scale features from candidate point cloud data to obtain point cloud data features at different scales; fuse the features of each point cloud data based on an attention mechanism to obtain fused features; denoise the candidate point cloud data according to the fused features to obtain denoised target point cloud data; and reconstruct the power line in 3D based on the target point cloud data to obtain a 3D model of the power line.

[0148] In one embodiment, the power line 3D reconstruction module 1108 is further configured to: construct a catenary model of the power line based on candidate point cloud data; perform a smoothing verification on the catenary model to obtain a smoothing verification result of the catenary model; perform a reliability verification on the catenary model to obtain a reliability verification result of the catenary model; and, if both the smoothing verification result and the reliability verification result indicate that the verification has passed, use the catenary model as the 3D model of the power line.

[0149] In one embodiment, the power line 3D reconstruction device is further configured to: search for candidate points within a preset neighborhood of each of the two endpoints of the catenary model, starting from both endpoints respectively; connect the endpoints and the candidate points if the candidate points pass the curvature check, and update the catenary model; and search for candidate points again starting from the candidate points until the search stops, and use the current catenary model as the 3D model of the power line.

[0150] Each module in the aforementioned three-dimensional power line reconstruction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0151] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 12As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores three-dimensional power line reconstruction data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a three-dimensional power line reconstruction method.

[0152] Those skilled in the art will understand that Figure 12 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0153] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: in response to a three-dimensional reconstruction instruction for power lines within a target power grid area, acquiring three-dimensional point cloud data of the target power grid area; classifying the terrain of the target power grid area based on the three-dimensional point cloud data to obtain a terrain classification result for the target power grid area; determining an elevation threshold that matches the terrain classification result; and filtering candidate point cloud data that matches the power lines from the three-dimensional point cloud data based on the elevation threshold; and performing three-dimensional reconstruction of the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0154] In one embodiment, when the processor executes the computer program, it further performs the following steps: performing point cloud segmentation on the three-dimensional point cloud data according to the terrain classification results to obtain at least one three-dimensional point cloud block; for each three-dimensional point cloud block, obtaining the elevation distribution characteristics of the three-dimensional point cloud set; and determining the elevation threshold that matches the three-dimensional point cloud block based on the elevation distribution characteristics.

[0155] In one embodiment, when the processor executes the computer program, it further performs the following steps: clustering the three-dimensional point cloud data based on an elevation threshold to obtain at least one three-dimensional point cloud class; analyzing the spatial distribution characteristics of each three-dimensional point cloud class; and, if the spatial distribution characteristics meet the power line distribution characteristics, using the three-dimensional point cloud data of the three-dimensional point cloud class as candidate point cloud data that matches the power line.

[0156] In one embodiment, when the processor executes the computer program, it further performs the following steps: multi-scale feature extraction on the candidate point cloud data to obtain point cloud data features of the candidate point cloud data at different scales; feature fusion of each point cloud data feature based on an attention mechanism to obtain fused features; denoising the candidate point cloud data according to the fused features to obtain denoised target point cloud data; and three-dimensional reconstruction of the power line based on the target point cloud data to obtain a three-dimensional model of the power line.

[0157] In one embodiment, when the processor executes the computer program, it further performs the following steps: constructing a catenary model of the power line based on candidate point cloud data; performing a smoothing verification on the catenary model to obtain a smoothing verification result of the catenary model; performing a reliability verification on the catenary model to obtain a reliability verification result of the catenary model; and using the catenary model as a three-dimensional model of the power line if both the smoothing verification result and the reliability verification result indicate that the verification has passed.

[0158] In one embodiment, when the processor executes the computer program, it further performs the following steps: starting from the two endpoints of the catenary model, searching for candidate points within the preset neighborhood range of each endpoint; if the candidate point passes the curvature check, connecting the endpoints and the candidate point, and updating the catenary model; starting from the candidate point, searching for candidate points again until the search stopping condition is met, and using the current catenary model as the three-dimensional model of the electric power line.

[0159] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, performs the following steps: in response to a three-dimensional reconstruction instruction for power lines within a target power grid area, acquiring three-dimensional point cloud data of the target power grid area; classifying the terrain of the target power grid area based on the three-dimensional point cloud data to obtain a terrain classification result for the target power grid area; determining an elevation threshold that matches the terrain classification result; and, based on the elevation threshold, filtering candidate point cloud data that matches the power lines from the three-dimensional point cloud data; and reconstructing the power lines in three dimensions based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0160] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: segmenting the three-dimensional point cloud data into point clouds based on the terrain classification results to obtain at least one three-dimensional point cloud block; acquiring the elevation distribution characteristics of the three-dimensional point cloud set for each three-dimensional point cloud block; and determining an elevation threshold that matches the three-dimensional point cloud block based on the elevation distribution characteristics.

[0161] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: clustering the three-dimensional point cloud data based on an elevation threshold to obtain at least one three-dimensional point cloud class; analyzing the spatial distribution characteristics of each three-dimensional point cloud class; and, if the spatial distribution characteristics satisfy the power line distribution characteristics, using the three-dimensional point cloud data of the three-dimensional point cloud class as candidate point cloud data that matches the power line.

[0162] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: multi-scale feature extraction on the candidate point cloud data to obtain point cloud data features of the candidate point cloud data at different scales; feature fusion of each point cloud data feature based on an attention mechanism to obtain fused features; denoising the candidate point cloud data according to the fused features to obtain denoised target point cloud data; and three-dimensional reconstruction of the power line based on the target point cloud data to obtain a three-dimensional model of the power line.

[0163] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: constructing a catenary model of the power line based on candidate point cloud data; performing a smoothing verification on the catenary model to obtain a smoothing verification result of the catenary model; performing a reliability verification on the catenary model to obtain a reliability verification result of the catenary model; and using the catenary model as a three-dimensional model of the power line if both the smoothing verification result and the reliability verification result indicate that the verification has passed.

[0164] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: starting from the two endpoints of the catenary model, searching for candidate points within the preset neighborhood range of each endpoint; if the candidate point passes the curvature check, connecting the endpoints and the candidate point, and updating the catenary model; starting from the candidate point again, searching for candidate points until the search stopping condition is met, and using the current catenary model as the three-dimensional model of the electric power line.

[0165] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps: in response to a three-dimensional reconstruction instruction for power lines within a target power grid area, acquiring three-dimensional point cloud data of the target power grid area; classifying the terrain of the target power grid area based on the three-dimensional point cloud data to obtain a terrain classification result for the target power grid area; determining an elevation threshold that matches the terrain classification result; and, based on the elevation threshold, filtering candidate point cloud data that matches the power lines from the three-dimensional point cloud data; and performing three-dimensional reconstruction of the power lines based on the candidate point cloud data to obtain a three-dimensional model of the power lines, wherein the three-dimensional model is a catenary model obtained after smoothing and reliability verification.

[0166] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: segmenting the three-dimensional point cloud data into point clouds based on the terrain classification results to obtain at least one three-dimensional point cloud block; acquiring the elevation distribution characteristics of the three-dimensional point cloud set for each three-dimensional point cloud block; and determining an elevation threshold that matches the three-dimensional point cloud block based on the elevation distribution characteristics.

[0167] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: clustering the three-dimensional point cloud data based on an elevation threshold to obtain at least one three-dimensional point cloud class; analyzing the spatial distribution characteristics of each three-dimensional point cloud class; and, if the spatial distribution characteristics satisfy the power line distribution characteristics, using the three-dimensional point cloud data of the three-dimensional point cloud class as candidate point cloud data that matches the power line.

[0168] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: multi-scale feature extraction on the candidate point cloud data to obtain point cloud data features of the candidate point cloud data at different scales; feature fusion of each point cloud data feature based on an attention mechanism to obtain fused features; denoising the candidate point cloud data according to the fused features to obtain denoised target point cloud data; and three-dimensional reconstruction of the power line based on the target point cloud data to obtain a three-dimensional model of the power line.

[0169] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: constructing a catenary model of the power line based on candidate point cloud data; performing a smoothing verification on the catenary model to obtain a smoothing verification result of the catenary model; performing a reliability verification on the catenary model to obtain a reliability verification result of the catenary model; and using the catenary model as a three-dimensional model of the power line if both the smoothing verification result and the reliability verification result indicate that the verification has passed.

[0170] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: starting from the two endpoints of the catenary model, searching for candidate points within the preset neighborhood range of each endpoint; if the candidate point passes the curvature check, connecting the endpoints and the candidate point, and updating the catenary model; starting from the candidate point again, searching for candidate points until the search stopping condition is met, and using the current catenary model as the three-dimensional model of the electric power line.

[0171] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0172] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0173] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0174] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for three-dimensional reconstruction of power lines, characterized in that, The method includes: In response to a three-dimensional reconstruction command for power lines within a target power grid area, three-dimensional point cloud data of the target power grid area is acquired; Based on the three-dimensional point cloud data, the terrain of the target power grid area is classified to obtain the terrain classification result of the target power grid area; Based on the terrain classification results, the three-dimensional point cloud data is segmented to obtain at least one three-dimensional point cloud block. For each of the three-dimensional point cloud blocks, the elevation distribution characteristics of the three-dimensional point cloud set are obtained. Based on the elevation distribution characteristics, an elevation threshold that matches the three-dimensional point cloud block is dynamically generated. Based on the elevation threshold, candidate point cloud data that matches the power line are selected from the three-dimensional point cloud data; Based on the candidate point cloud data, the power line is reconstructed in three dimensions to obtain a catenary model after smoothing and reliability verification. Starting from each of the two endpoints of the catenary model, candidate points are searched within a preset neighborhood of each of the two endpoints. If the candidate point passes the curvature check, connect the endpoint and the candidate point, and update the catenary model; Starting from the candidate point, the candidate point search is performed again until the search stopping condition is met, and the current catenary model is used as the three-dimensional model of the electric power line.

2. The method according to claim 1, characterized in that, The step of filtering candidate point cloud data that matches the power line from the three-dimensional point cloud data based on the elevation threshold includes: Based on the elevation threshold, the three-dimensional point cloud data is clustered to obtain at least one three-dimensional point cloud class; For each of the three-dimensional point cloud classes, analyze the spatial distribution characteristics of the three-dimensional point cloud classes; If the spatial distribution features satisfy the power line distribution feature conditions, the three-dimensional point cloud data of the three-dimensional point cloud class is used as candidate point cloud data that matches the power line.

3. The method according to claim 1, characterized in that, The step of performing three-dimensional reconstruction of the power lines based on the candidate point cloud data includes: Multi-scale feature extraction is performed on the candidate point cloud data to obtain the point cloud data features of the candidate point cloud data at different scales; Based on the attention mechanism, feature fusion is performed on the features of each point cloud data to obtain fused features; Based on the fusion features, the candidate point cloud data is denoised to obtain the denoised target point cloud data. Based on the target point cloud data, the power line is reconstructed in three dimensions.

4. A three-dimensional reconstruction device for power lines, characterized in that, The device includes: The point cloud data acquisition module is used to acquire the three-dimensional point cloud data of the target power grid area in response to a three-dimensional reconstruction command for power lines within the target power grid area. The terrain classification module is used to classify the terrain of the target power grid area based on the three-dimensional point cloud data, and obtain the terrain classification result of the target power grid area. The point cloud data filtering module is used to segment the three-dimensional point cloud data according to the terrain classification results to obtain at least one three-dimensional point cloud block; for each three-dimensional point cloud block, the module obtains the elevation distribution characteristics of the three-dimensional point cloud set; dynamically generates an elevation threshold that matches the three-dimensional point cloud block based on the elevation distribution characteristics; and filters candidate point cloud data that matches the power line from the three-dimensional point cloud data based on the elevation threshold. The power line 3D reconstruction module is used to perform 3D reconstruction of the power line based on the candidate point cloud data to obtain a catenary model after smoothing and reliability verification; starting from the two endpoints of the catenary model, candidate points are searched within a preset neighborhood of each endpoint; if a candidate point passes curvature verification, the endpoints and the candidate point are connected, and the catenary model is updated; starting from the candidate point, the candidate point search is performed again until the search stopping condition is met, and the current catenary model is used as the 3D model of the power line.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

6. 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 3.

7. A computer program product, comprising a computer program, 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 3.

Images