A tree barrier hidden danger monitoring method and system of a power transmission channel

Abstract

The present application relates to power transmission channel inspection technical field, specifically to a kind of tree barrier hidden danger monitoring method and system of power transmission channel, solve the technical problem of low accuracy of existing technology tree barrier hidden danger monitoring.This method comprises: obtaining the vegetation 3D point cloud subset and 3D traverse center curve of the space where power transmission channel is located;According to vegetation 3D point cloud subset and 3D traverse center curve, generate 3D space voxel set, and medium topology construction is carried out to 3D space voxel set, obtain experience obstruction topology network;Based on experience obstruction topology network and 3D traverse center curve, tree barrier risk quantification analysis is carried out, and target intersection voxel and power transmission channel insulation strength index are determined;Target intersection voxel is used to represent the tree barrier anchor point position to be monitored in power transmission channel, and power transmission channel insulation strength index is used to represent the tree barrier insulation margin of power transmission channel;According to target intersection voxel and power transmission channel insulation strength index, tree barrier hidden danger monitoring is carried out.

Description

Technical Field

[0001] This invention relates to the field of power transmission channel inspection technology, specifically to a method and system for monitoring tree obstruction hazards in power transmission channels. Background Technology

[0002] Power grid maintenance departments typically use drones equipped with visible light cameras and lidar to perform transmission line inspections. During the inspection, the system extracts spatial coordinates from the 3D point cloud data, calculates the Euclidean distance between the vegetation coordinates and the conductor coordinates, and outputs an emergency defect alarm signal when this distance is less than a preset safety threshold.

[0003] However, in engineering practice, the above-mentioned scheme based on simple geometric distance measurement has several drawbacks. Euclidean linear measurement is difficult to distinguish the density of the spatial medium of tree obstructions and cannot differentiate between the sparse, soft leaves on the outer side of the canopy and the dense woody trunk inside. In windy conditions, the outer soft leaves sway with the wind and shorten the spatial distance between them and the conductor, resulting in the inspection system outputting a large number of invalid false alarm work orders based on single-point geometric displacement. In addition, the single shortest straight distance cannot reflect the overall thickness and density of the obstruction formed by multiple intersecting branches, making it difficult to quantitatively assess the actual comprehensive tree obstruction hazard level of the transmission channel. This leads to the low accuracy of existing tree obstruction hazard monitoring technology. Summary of the Invention

[0004] To address the low accuracy of existing tree obstruction hazard monitoring technologies, the present invention aims to provide a method and system for monitoring tree obstruction hazards in power transmission channels. The specific technical solution adopted is as follows: This invention provides a method for monitoring tree obstruction hazards in power transmission channels, comprising: Obtain a subset of the 3D point cloud of vegetation and the 3D center curve of the conductor in the space where the power transmission channel is located; A three-dimensional spatial voxel set is generated based on the vegetation three-dimensional point cloud subset and the three-dimensional traverse center curve, and a medium topology is constructed on the three-dimensional spatial voxel set to obtain an empirical barrier topology network; the edge weight of the connecting edge between adjacent voxels in the empirical barrier topology network corresponds to the medium type of the adjacent voxels. Based on the aforementioned empirical obstacle topology network and the aforementioned three-dimensional conductor center curve, a tree obstacle risk quantification analysis is performed to determine the target intersecting voxel and the transmission channel insulation strength index. The target intersecting voxel is used to characterize the location of the tree obstacle anchor point to be monitored within the transmission channel, and the transmission channel insulation strength index is used to characterize the tree obstacle insulation margin of the transmission channel. Tree obstruction hazard monitoring is conducted based on the target intersecting voxels and the insulation strength index of the power transmission channel.

[0005] Optionally, a three-dimensional spatial voxel set is generated based on the vegetation three-dimensional point cloud subset and the three-dimensional traverse center curve, and a medium topology is constructed on the three-dimensional spatial voxel set to obtain an empirical barrier topology network, including: Based on the three-dimensional point cloud subset of vegetation and the center curve of the three-dimensional conductor, a spatial bounding box containing the conductor and the ground in the power transmission channel is constructed, and the spatial bounding box is discretized to generate a three-dimensional spatial voxel set. Based on the echo characteristics of radar points inside each voxel in the three-dimensional spatial voxel set, the medium type corresponding to each voxel is determined; the medium type includes air medium voxels, soft leaf medium voxels, and woody branch voxels. The voxels in the three-dimensional space voxel set that contain the center curve of the three-dimensional conductor are taken as the high-voltage source node set, and the voxels in the three-dimensional space voxel set that intersect with the bottom surface of the spatial bounding box are taken as the grounding sink node set. Using voxels in the high-voltage source node set as source nodes and voxels in the ground sink node set as sink nodes, connect adjacent voxels to each other and assign edge weights based on the medium type of the voxels at both ends of each connect edge to obtain an empirical obstruction topology network.

[0006] Optionally, medium identification is performed based on the echo characteristics of radar points within each voxel in the three-dimensional spatial voxel set to determine the medium type corresponding to each voxel, including: The medium type corresponding to the voxel that does not contain radar points is determined to be the air medium voxel; The medium type corresponding to the voxel whose corresponding echo number is less than the total number of echoes for all internal radar points is determined as the soft leaf medium voxel. The medium type corresponding to the voxel containing at least one radar point whose corresponding echo number is equal to the total number of echoes is determined as the woody branch voxel.

[0007] Optionally, based on the empirical barrier topology network and the three-dimensional conductor center curve, a tree barrier risk quantification analysis is performed to determine the target intersecting voxels and the transmission channel insulation strength index, including: In the empirically hindered topology network, iteratively query the connected paths from the source node to the sink node, extract the minimum edge weight on the connected path, and subtract the edge weight of each connected edge on the connected path until there is no connected path from the source node to the sink node in the empirically hindered topology network, thus obtaining the residual network. Based on the residual network, reachability traversal is performed to extract the topological cut edge set, and the target intersecting voxel is determined according to the spatial intersection of the voxel where the endpoint of the connected edge in the topological cut edge set is located and the woody branch voxel in the three-dimensional spatial voxel set. The insulation strength index of the transmission channel is determined based on the edge weights of all connected edges in the set of topological cut edges and the spatial arc length of the center curve of the three-dimensional conductor.

[0008] Optionally, in the empirically hindering topology network, iteratively query the connected paths from the source node to the sink node, extract the minimum edge weight on the connected path, and subtract the edge weight of each connected edge on the connected path until there is no connected path from the source node to the sink node in the empirically hindering topology network, to obtain the residual network, including: In each iteration, the empirically hindering topology network of the current iteration is queried for connected paths from the source node to the sink node; the edge weights of each connected edge on the connected path are all positive numbers; Subtract the minimum edge weight from the edge weight of each connected edge on the connected path to generate the empirical obstacle topology network for the next iteration and proceed to the next iteration; If there is no connected path from the source node to the sink node in the current iteration's empirically hindering topology network, then the current iteration's empirically hindering topology network is considered the residual network.

[0009] Optionally, a subset of the 3D point cloud of vegetation and the 3D conductor center curve of the space where the transmission channel is located are obtained, including: Acquire visible light images and three-dimensional radar point clouds of the power transmission channel; The visible light image is segmented into instances to generate wire pixel masks and vegetation pixel masks; Based on the traverse pixel mask and the vegetation pixel mask, extract the 3D point cloud subsets of the traverse and the 3D point cloud subsets of the vegetation from the 3D radar point cloud, respectively. Curve fitting is performed on the subset of the three-dimensional point cloud of the conductor to generate the center curve of the three-dimensional conductor.

[0010] Optionally, tree obstruction hazard monitoring is performed based on the target intersecting voxel and the insulation strength index of the transmission channel, including: Obtain the empirical safety benchmark threshold corresponding to the voltage level of the power transmission channel; If the insulation strength index of the power transmission channel is less than the empirical safety benchmark threshold, a tree obstacle hazard alarm is triggered and a monitoring image calibration is performed based on the target intersecting voxels.

[0011] Optionally, monitoring image calibration is performed based on the target intersecting voxels, including: A two-dimensional mapping verification is performed on each target intersecting voxel to determine the set of trimming target point coordinates; the set of trimming target point coordinates includes the two-dimensional projection coordinates of the target intersecting voxels to be trimmed on the visible light image; Image rendering calibration is performed based on the set of pruning target coordinates and the insulation strength index of the power transmission channel to generate on-site maintenance images.

[0012] Optionally, a two-dimensional mapping verification is performed on each target intersecting voxel to determine the set of pruning target point coordinates, including: The three-dimensional coordinates of the geometric center of each intersecting target voxel are mapped to the pixel plane of the visible light image to generate the corresponding two-dimensional projection coordinates. Based on the vegetation pixel mask, the position of the two-dimensional projection coordinates is verified, and the two-dimensional projection coordinates that are not within the vegetation pixel mask are deleted to obtain the set of pruning target point coordinates.

[0013] This invention provides a tree obstruction hazard monitoring system for power transmission channels, comprising: a processor and a communication interface; the communication interface and the processor are coupled, and the processor is used to run computer programs or instructions to implement any of the methods described above.

[0014] The present invention has the following beneficial effects: This invention establishes basic spatial data for monitoring tree barrier hazards by acquiring a subset of three-dimensional point clouds of vegetation and the center curve of a three-dimensional conductor in the space where the power transmission channel is located. Based on the subset of three-dimensional point clouds and the center curve of the three-dimensional conductor, a set of three-dimensional spatial voxels is generated, and an empirical barrier topology network is constructed using media topology. This transforms the continuous three-dimensional space into discrete computational units and achieves quantitative characterization of the penetration resistance of different media at the data structure level. Subsequently, based on the empirical barrier topology network and the center curve of the three-dimensional conductor, quantitative analysis of insulation strength is performed to determine the target intersecting voxels and the insulation strength index of the power transmission channel. The tree barrier intrusion problem is transformed into a path blocking problem in graph theory networks, and the comprehensive tree barrier hazard level of the power transmission channel is evaluated. Tree barrier hazard monitoring is performed based on the target intersecting voxels and the insulation strength index of the power transmission channel, effectively improving the accuracy and reliability of tree barrier hazard monitoring. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a method for monitoring tree obstruction hazards in a power transmission channel, provided in one embodiment of the present invention. Figure 2 This is a system architecture diagram of a tree obstacle hazard monitoring system for a power transmission channel, provided as an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a tree obstruction hazard monitoring method and system for power transmission channels proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] The following description, in conjunction with the accompanying drawings, details the specific scheme of a tree obstruction hazard monitoring method and system for power transmission channels provided by the present invention.

[0019] Please see Figure 1 The diagram illustrates a flowchart of a method for monitoring tree obstruction hazards in a power transmission channel according to an embodiment of the present invention. The method includes the following steps: S100: Obtain a subset of the 3D point cloud of vegetation and the 3D conductor center curve of the space where the power transmission channel is located.

[0020] Among them, the vegetation 3D point cloud subset includes the spatial coordinates of the 3D point cloud of the vegetation surface in the space where the power transmission channel is located, and the 3D conductor center curve is used to characterize the continuous spatial trajectory of the conductor in the power transmission channel.

[0021] Specifically, S100 includes: S110: Acquire visible light images and three-dimensional radar point clouds of the power transmission channel.

[0022] For example, the system continuously monitors the unified timestamp of the UAV hardware, extracts the generation time of a visible light image containing pixel color information corresponding to a certain sampling moment, and the generation time of the corresponding batch of 3D radar point clouds containing spatial coordinates, echo sequence numbers, and total echo count parameters. The system calculates the time difference between the two generation times and compares this time difference with a preset synchronization tolerance threshold (e.g., set to 5 milliseconds). When the time difference is less than or equal to the synchronization tolerance threshold, the system determines that the data synchronization of the current frame is valid and writes the synchronized visible light image and 3D radar point cloud into memory. Through time alignment, the system ensures that the visible light image and 3D radar point cloud have a consistent spatiotemporal reference when spatially mapping, avoiding projection misalignment caused by data asynchrony. The visible light image is represented in the form of a two-dimensional visible light image matrix.

[0023] S120. Perform instance segmentation on the visible light image to generate wire pixel masks and vegetation pixel masks.

[0024] For example, the system inputs a visible light image into a pre-deployed instance segmentation network (which may employ the Mask R-CNN algorithm model). The instance segmentation network performs pixel-by-pixel category prediction on each pixel in the visible light image, outputting a wire pixel mask representing a two-dimensional connected region of a wire and a vegetation pixel mask representing a two-dimensional connected region of vegetation. Both the wire pixel mask and the vegetation pixel mask are two-dimensional matrices, with pixel values ​​of 1 for target regions and 0 for background regions. Furthermore, the system extracts the two-dimensional closed boundary coordinate sequence at the intersection of pixel values ​​of 1 and 0.

[0025] S130. Based on the traverse pixel mask and the vegetation pixel mask, extract the 3D point cloud subsets of the traverse and the 3D point cloud subsets of the vegetation from the 3D radar point cloud.

[0026] First, the system calls the pre-calibrated camera intrinsic and extrinsic transformation matrices. For each pixel coordinate on the two-dimensional closed boundary coordinate sequence of the guide vane pixel mask, it continuously left-multiplies the inverse of the camera intrinsic and extrinsic transformation matrices, transforming the two-dimensional pixel into a directional ray in a three-dimensional coordinate system. All rays corresponding to the guide vane boundary close together in three-dimensional space, generating a three-dimensional guide vane interception frustum. Using the same matrix inversion operation, the system generates a three-dimensional vegetation interception frustum for the two-dimensional closed boundary coordinate sequence of the vegetation pixel mask.

[0027] Subsequently, the system uses a 3D traverse to extract the view frustum and performs spatial Boolean cross-validation on the 3D radar point cloud, extracting radar points whose spatial coordinates fall within the 3D traverse traverse view frustum as the initial traverse point cloud. To filter out background terrain noise caused by the infinite divergence of the view frustum (such as mountains or ground behind the traverse), the system performs density-based spatial clustering on the initial traverse point cloud (using the DBSCAN algorithm, with a preset scanning neighborhood radius of 0.5 meters and a minimum number of core points of 10), dividing the initial traverse point cloud into multiple discrete point cloud clusters.

[0028] For each point cloud cluster, the system calculates the average elevation of all points within it and performs principal component analysis to calculate the length-to-width ratio of the cluster in 3D space. For example, first, the system constructs the covariance matrix of the 3D coordinates of the point cloud cluster, calculates its eigenvalues ​​and eigenvectors, and uses the direction of the eigenvector corresponding to the largest eigenvalue as the primary extension direction of the cluster. The spatial distance between all extreme points of projected coordinates along this primary extension direction is defined as the length. The direction of the eigenvector corresponding to the second largest eigenvalue is defined as the secondary extension direction, and the spatial distance between all extreme points of projected coordinates along this secondary extension direction is defined as the width. Then, the system calculates the length-to-width ratio of the cluster in 3D space and extracts the target point cloud cluster with the largest average elevation (highest position) and a length-to-width ratio greater than a preset ratio (exhibiting a linear extension shape) as the final denoised traverse 3D point cloud subset. Similarly, the system extracts radar points whose spatial coordinates fall within the 3D vegetation truncated frustum and outputs a vegetation 3D point cloud subset.

[0029] S140. Based on the 3D point cloud subset of the traverse, perform curve fitting to generate the 3D traverse center curve.

[0030] For example, the system reads the three-dimensional coordinates of all radar points in the three-dimensional point cloud subset of the traverse, runs the least squares fitting algorithm of the catenary equation on these three-dimensional coordinates, and generates a continuous smooth curve in the three-dimensional coordinate system based on the output coefficients of the fitting algorithm. This curve is defined as the center curve of the three-dimensional traverse.

[0031] S200. Generate a three-dimensional spatial voxel set based on the vegetation three-dimensional point cloud subset and the three-dimensional traverse center curve, and construct the medium topology of the three-dimensional spatial voxel set to obtain an empirical barrier topology network.

[0032] In this system, each voxel in the three-dimensional spatial voxel set is a cubic unit containing spatial coordinate boundaries. The empirical barrier topology network is constructed with voxels as nodes and connections between adjacent voxels as edges. The edge weights of the edges between adjacent voxels in the empirical barrier topology network correspond to the media types of the adjacent voxels. Different media types reflect the differences in vegetation density within voxels. By assigning differentiated edge weights to different media types, the barrier characteristics of different media can be characterized.

[0033] Specifically, S200 includes: S210. Construct a spatial bounding box containing the conductor and the ground in the power transmission channel based on the vegetation 3D point cloud subset and the 3D conductor center curve, and discretize the spatial bounding box to generate a 3D spatial voxel set.

[0034] For example, the system extracts the maximum and minimum values ​​of all coordinate points in the three-dimensional traverse center curve and the vegetation three-dimensional point cloud subset on the horizontal two-dimensional plane (XY plane) to generate the initial horizontal boundary.

[0035] To prevent missing ground data due to vegetation being distributed only on one side of the guide, the system uses the projection trajectory of the 3D guide's center curve on the horizontal plane as a reference, and horizontally expands the width of the pre-defined passageway to both sides of the trajectory (e.g., expanding 15 meters to each side). If the expansion exceeds the initial horizontal boundary, the system updates the horizontal boundary to the expanded wider boundary.

[0036] Next, the system extracts the highest elevation point of all coordinate points in the 3D conductor center curve and the vegetation 3D point cloud subset. This highest elevation point is then extended upwards by a preset safety tolerance height (e.g., 5 meters) as the top boundary. Finally, the system searches the global 3D radar point cloud for all point clouds falling within this broad horizontal boundary and extracts the horizontal plane containing the lowest elevation point (representing the actual ground elevation) as the bottom boundary. Combining the updated broad horizontal boundary, top boundary, and bottom boundary, the system generates a vertically connected and completely enclosed spatial bounding box, ensuring that subsequent network flow algorithms have a complete spatial transition region from the high-voltage conductor to the ground.

[0037] Finally, the system discretizes the bounding box to generate a set of three-dimensional spatial voxels. The system loads a fixed spatial step size parameter (e.g., set to 0.2 meters) for mesh discretization into the configuration file. Following this step size, it performs uniform meshing along the length, width, and height of the bounding box, generating a set of discretized cubic elements, which constitutes the three-dimensional spatial voxel set. This set of voxels serves as the basic unit for assigning graph theory weights, with each voxel independently carrying its internal spatial coordinate boundaries and medium property information.

[0038] S220. Based on the echo characteristics of radar points inside each voxel in the three-dimensional space voxel set, perform medium identification and determine the medium type corresponding to each voxel.

[0039] The media types include air-based voxels, soft-leaf-based voxels, and woody branch / trunk voxels. When the radar laser penetrates the canopy, the outer soft leaves cannot completely block the laser beam, resulting in multiple penetrations and generating multiple sequentially ordered echo numbers. The dense woody trunk, however, completely intercepts the laser beam, producing the final echo. The system utilizes this physical property to distinguish geometrically similar grid cells.

[0040] Specifically, S220 includes: determining the medium type of voxels with no radar points inside as air medium voxels; determining the medium type of voxels with all radar points inside satisfying that the corresponding echo number is less than the total number of echoes as soft leaf medium voxels; and determining the medium type of voxels with at least one radar point inside satisfying that the corresponding echo number is equal to the total number of echoes as woody branch voxels (including single echo points with a total number of echoes equal to 1, and last echo points with a total number of echoes greater than or equal to 2).

[0041] For each voxel in the 3D spatial voxel set, the system extracts its geometric coordinate boundaries and retrieves all radar point cloud data whose spatial coordinates completely fall within the voxel from the vegetation 3D point cloud subset. If there is no point cloud data inside the voxel, it indicates that the area is an open background space, and the system marks it as an air medium voxel. If there is at least one point cloud data inside the voxel, the system extracts the current echo sequence number and total echo count for each laser reflection pulse inside the voxel. If all point cloud data inside the voxel satisfy that the current echo sequence number is strictly less than the total echo count, it indicates that all radar pulses passing through the voxel continue to propagate backward and generate subsequent echoes at a greater distance, and the voxel space does not completely block the beam, so the system marks it as a soft-leaved medium voxel. If the voxel contains at least one point cloud data that satisfies that the current echo sequence number is equal to the total echo count, it indicates that the voxel space completely intercepts the radar pulse spot, preventing it from penetrating further backward, and the system marks it as a woody branch voxel. The case where the echo number equals the total number of echoes includes single echo points with a total echo count of 1, and final echo points with a total echo count greater than or equal to 2. The system sequentially traverses all voxels in the 3D spatial voxel set, completes the medium definition, and outputs a 3D spatial voxel set with clear classification labels.

[0042] S230. The voxels in the three-dimensional space voxel set containing the center curve of the three-dimensional conductor are taken as the high voltage source node set, and the voxels in the three-dimensional space voxel set that intersect with the bottom surface of the space bounding box are taken as the grounding sink node set.

[0043] For example, after acquiring a set of three-dimensional spatial voxels with classification labels, the system begins to construct a network flow graph theory model. First, the system traverses the set of three-dimensional spatial voxels, extracting all voxels containing the coordinate trajectory of the three-dimensional conductor center curve, and packages these voxels into a high-voltage source node set, which represents the starting position of the electric field extending downwards. Subsequently, the system extracts all voxels that intersect or are adjacent to the bottom surface of the spatial bounding box, and packages these voxels into a ground sink node set, which represents the final discharge endpoint after the electric field penetrates the insulating medium.

[0044] S240. Establish connecting edges between adjacent voxels using voxels in the high-voltage source node set as source nodes and voxels in the ground sink node set as sink nodes, and assign edge weights according to the medium type of the voxels at both ends of each connecting edge to obtain an empirical obstruction topology network.

[0045] For example, the system establishes a pair of directed connected edges with opposite directions for any two adjacent voxels in three-dimensional space, connecting discrete voxels into a global three-dimensional directed connected graph. Specifically, for any two adjacent voxels, the system calculates the Euclidean distance from the geometric center of each voxel to the nearest voxel in the high-pressure source node set. The voxel with the smaller distance is determined to be closer to the high-pressure source node, and the other voxel is determined to be farther away from the high-pressure source node. The direction from the voxel closer to the high-pressure source node to the voxel farther away from the high-pressure source node is the positive direction of the connected edge, with the corresponding endpoints being the start and end endpoints, respectively. After establishing the connected edges, the system pre-loads three edge weight constants, corresponding to air medium, soft-leaf medium, and woody branch medium, respectively, and they satisfy a monotonically decreasing relationship, i.e., the edge weight constant corresponding to air medium is greater than the edge weight constant corresponding to soft-leaf medium, and the edge weight constant corresponding to soft-leaf medium is greater than the edge weight constant corresponding to woody branch medium. For example, the edge weight constant corresponding to air medium is 1000, the edge weight constant corresponding to soft-leaf medium is 500, and the edge weight constant corresponding to woody branch medium is 10.

[0046] For each connected edge in the directed connected graph, the system assigns edge weights based on the medium type of the voxels connected to its two ends. If both ends of the connected edge are air medium voxels, its edge weight is assigned the edge weight constant corresponding to air medium. If the voxels connected to both ends of the connected edge include at least one soft-leaf medium voxel and no woody branch voxels, its edge weight is assigned the edge weight constant corresponding to soft-leaf medium. If the voxels connected to both ends of the connected edge include at least one woody branch voxel, regardless of whether the other end is air, soft-leaf, or wood, its edge weight is assigned the edge weight constant corresponding to woody branch medium. After completing the numerical assignment of all connected edges, the system outputs a global empirical obstacle topology network with the high-voltage source node set as the starting point and the ground sink node set as the ending point, and all paths carrying quantified resistance values. By setting the edge weight constants in this way, the resistance to traversing soft leaves is greater than that to traversing wood in the pathfinding algorithm. When the algorithm attempts to find the path of least resistance, it will avoid the grid filled with soft leaves, thus achieving automatic filtering of wind deflection noise at the underlying data structure level.

[0047] S300. Based on empirical obstacle topology network and three-dimensional conductor center curve, tree obstacle risk quantification analysis is performed to determine target intersecting voxels and transmission channel insulation strength index.

[0048] Among them, the target intersecting voxel is used to characterize the location of the tree barrier anchor point to be monitored within the transmission channel, that is, the dense lignin voxel on the critical section where the tree barrier intrusion resistance is minimized, and its spatial coordinates indicate the location of the potential tree barrier entity. The transmission channel insulation strength index is used to characterize the tree barrier insulation margin of the transmission channel.

[0049] Specifically, the S300 includes: S310. In the empirically hindered topology network, iteratively query the connected paths from the source node to the sink node, extract the minimum edge weight on the connected path, and subtract the weight of each connected edge on the connected path until there is no connected path from the source node to the sink node in the empirically hindered topology network, and obtain the residual network.

[0050] Specifically, S310 includes: In each iteration, the system queries the empirically hindered topology network for the current iteration to find a connected path from the source node to the sink node. The edge weights of all connected edges on this path are positive. For example, a breadth-first search (BFS) strategy is used to find a connected path in the empirically hindered topology network that starts from the set of high-voltage source nodes and ends at the set of grounded sink nodes, with the edge weight of each edge on this path being strictly greater than 0 in the current iteration. The system reads the current edge weights of all connected edges on this path and extracts the minimum value as the minimum available weight of the path. This minimum available weight represents the maximum capacity bottleneck that the current connected path can withstand.

[0051] Next, the minimum edge weight is subtracted from the edge weight of each connected edge in the connected path to generate the empirical barrier topology network for the next iteration and proceed to the next iteration. The system synchronously performs a numerical deduction operation on each connected edge in the connected path, subtracting the minimum available weight of the path from the current edge weight of these edges. After the deduction operation is completed, at least one connected edge in the connected path will have a remaining edge weight of zero, indicating that the connected path has been completely consumed and blocked. The system generates an updated empirical barrier topology network and proceeds to the next iteration, continuing to search for the next connected path with a weight greater than zero in the empirical barrier topology network. As the number of iterations increases, the number of connected edges with remaining edge weights greater than zero in the empirical barrier topology network gradually decreases.

[0052] If there is no connected path from the source node to the sink node in the current iteration's empirically hindering topology network, then the current iteration's empirically hindering topology network is used as the residual network. When no connected path from the source node to the sink node with all edge weights greater than zero can be found in a certain iteration, the system determines that the current network has reached a saturated blocking state, terminates the iteration operation, and uses the current iteration's empirically hindering topology network as the residual network.

[0053] S320. Based on the residual network, perform reachability traversal, extract the topological cut edge set, and determine the target intersecting voxel based on the spatial intersection of the voxels where the endpoints of the connected edges in the topological cut edge set are located and the woody branch voxels in the three-dimensional space voxel set.

[0054] To accurately identify the weakest cross section causing network blockage, the system performs a reachability traversal in the residual network after iteration termination to extract the set of topological cut edges. Starting from the set of high-voltage source nodes, the system uses a breadth-first traversal strategy to extend outwards along connected edges with strictly greater than zero weights on the remaining edges. All voxels that can be reached are marked as reachable connected sets of the source nodes, while those that cannot be reached are marked as unreachable connected sets of the source nodes.

[0055] Subsequently, the system traverses all directed connected edges established in the initial state of the global network, extracting connected edges that satisfy the cross-boundary condition, i.e., connected edges whose starting endpoint is located within the reachable connected set of the source node and whose ending endpoint is located within the unreachable connected set of the source node. The system combines all extracted connected edges that satisfy the cross-boundary condition and outputs a set of topological cut edges. This set of topological cut edges is physically equivalent to the weakest continuous section where the barrier defense is completely penetrated. Because the edge weights of soft leaves are set relatively high in the previous steps, it is extremely difficult for soft leaf connected edges to be reduced to zero. Therefore, the extracted weakest section automatically avoids the soft leaf region, achieving immunity to wind deflection noise.

[0056] After acquiring the topological cut edge set, the system extracts the endpoints of all connected edges constituting this set and locates the voxels containing these endpoints. These voxels represent the locations where the defense line is most easily breached. The system performs a spatial intersection extraction operation on these voxels and the woody branch voxels in the 3D spatial voxel set, retaining voxels that are both located at the endpoints of the topological cut edge set and belong to the woody branch voxel category. The system defines this extracted intersection set of voxels as the target intersection voxels, whose 3D coordinates represent the location of the most critical dense tree barrier anchor point within the current transmission span. The system caches the topological cut edge set and the target intersection voxels and passes them to subsequent stages for final risk quantification and image calibration.

[0057] S330. Determine the insulation strength index of the transmission channel based on the edge weights of all connected edges in the topological cut edge set and the spatial arc length of the three-dimensional conductor center curve.

[0058] For example, the system extracts the edge weights assigned to all connected edges in the topology cut edge set during the initial network construction phase, performs a scalar addition operation on these edge weights, and calculates the real-number sum reflecting the total obstruction cost of the current defense line. To eliminate the scale effect of conductor cut length on the total resistance assessment, the system extracts the three-dimensional conductor center curve and calculates its actual spatial arc length in three-dimensional space, which is a real number greater than zero. Subsequently, the system calculates the transmission channel insulation strength index using the following formula: in, Indicates the insulation strength index of the power transmission channel. This represents the sum of the edge weights of all connected edges in the set of topological cut edges. This represents the spatial arc length of the center curve of the three-dimensional conductor. This calculation method distributes the total resistive force of the intercepting electric field in three-dimensional space across each meter of conductor length, ensuring that an increase in conductor length does not lead to a false expansion of the insulation strength value. The smaller the value, the higher the proportion of dense lignin with minimal resistance within the space of a unit length of conductor, or the fewer the safe air gaps that hinder discharge, meaning a higher degree of risk of tree obstruction.

[0059] S400, monitor tree obstruction hazards based on target intersecting voxels and transmission channel insulation strength index.

[0060] When the insulation strength index of the transmission channel indicates the presence of tree obstruction hazards, the system triggers a corresponding monitoring response and outputs a quantitative indicator characterizing the degree of tree obstruction danger. The spatial coordinates of the target intersecting voxels are used to locate specific tree obstruction anchor points, providing precise targeting guidance for on-site maintenance personnel.

[0061] Specifically, S400 includes: acquiring an empirical safety reference threshold corresponding to the voltage level of the transmission channel; if the insulation strength index of the transmission channel is less than the empirical safety reference threshold, triggering a tree obstacle hazard alarm and performing monitoring image calibration based on the target intersecting voxels.

[0062] Among them, the empirical safety benchmark threshold is a pre-calibrated constant, with different thresholds corresponding to different voltage levels, representing the minimum insulation strength that needs to be maintained per unit length of conductor space at that voltage level. For example, for a 220kV line, the minimum insulation strength that needs to be maintained per meter is set at 400.

[0063] If the insulation strength index of the transmission channel is greater than or equal to the empirical safety benchmark threshold, the system determines that the barrier within a unit length of the current channel is sufficient to maintain a safe state and does not output an alarm command. If the insulation strength index of the transmission channel is strictly less than the empirical safety benchmark threshold, the system determines that the dielectric distribution within the transmission span is too dense and has exceeded the minimum safety limit, immediately triggers a tree obstacle hazard alarm command, and enters the monitoring image calibration process.

[0064] Specifically, image calibration based on target intersecting voxels includes: performing two-dimensional mapping verification on each target intersecting voxel to determine the set of pruning target point coordinates.

[0065] The set of target point coordinates for trimming includes the two-dimensional projection coordinates of the intersecting voxels of the target to be trimmed on the visible light image.

[0066] Specifically, the system maps the three-dimensional coordinates of the geometric center of each intersecting target voxel to the pixel plane of the visible light image, generates the corresponding two-dimensional projection coordinates, performs position verification on the two-dimensional projection coordinates based on the vegetation pixel mask, deletes the two-dimensional projection coordinates that are not in the vegetation pixel mask, and obtains the set of pruning target coordinates.

[0067] For example, in response to a tree obstacle hazard alarm command, the system calculates the three-dimensional coordinates of the geometric center of each intersecting voxel, calls the pre-calibrated camera intrinsic parameter matrix and camera extrinsic parameter transformation matrix, and continuously left-multiplies the three-dimensional coordinates of each geometric center by the camera extrinsic parameter transformation matrix and camera intrinsic parameter matrix, so as to positively map the position of each voxel center in the three-dimensional coordinate system back to the two-dimensional pixel plane and generate the corresponding two-dimensional projected coordinates.

[0068] To eliminate false multiple echo noise generated by radar when scanning surface metal or reflective rocks, the system calls a vegetation pixel mask to determine whether the row and column pixel positions of each two-dimensional projection coordinate fall within the closed boundary of the vegetation pixel mask (pixel value 1). If the two-dimensional projection coordinate is outside the closed boundary, the system determines it to be a false ground noise point outside vegetation and performs numerical removal on its coordinates. If the two-dimensional projection coordinate is inside the closed boundary, the system retains the coordinate point, confirms it as a real high-risk tree anchor point, and packages it into a set of pruning target coordinates. The set of pruning target coordinates includes the two-dimensional projection coordinates of the intersecting voxels of the target to be pruned on the visible light image. Through secondary verification of the vegetation mask, the system effectively filters out false ground noise in non-vegetated areas, ensuring that the output target coordinates are all located within real tree areas.

[0069] Finally, image rendering calibration was performed based on the set of pruning target coordinates and the insulation strength index of the transmission channel to generate on-site maintenance images.

[0070] The system retrieves a visible light image from memory as the rendering base map, and uses each row and column coordinate point in the set of trimmed target point coordinates as the center anchor point, extending a fixed pixel distance (e.g., radiating outwards) around that anchor point. The system generates a square pixel array to create a target rendering area. It then modifies the values ​​of all pixels within this target area to extreme values ​​in specific color channels (e.g., forcibly overwriting the red channel value to 255), and overlays these values ​​onto the original base image to create a highlighted selection area. Simultaneously, the system writes the power transmission channel insulation strength index as a text label into the blank areas at the image edges. Finally, it packages and saves the image matrix carrying this quantized score and the highlighted selection area, outputting the on-site inspection image.

[0071] This invention also provides a hardware structure diagram of a tree obstruction hazard monitoring system for power transmission channels (referred to as tree obstruction hazard monitoring system 20 for power transmission channels), see [link to diagram]. Figure 2 The tree obstruction hazard monitoring system 20 of the power transmission channel includes a processor 21, and optionally, a memory 22 connected to the processor 21.

[0072] Optional, see Figure 2 The tree obstruction hazard monitoring system 20 for power transmission channels also includes a communication interface 23. The processor 21, memory 22, and communication interface 23 are connected via a bus. The communication interface 23 is used to communicate with other devices or communication networks. Optionally, the communication interface 23 may include a transmitter and a receiver. The device in the communication interface 23 used to implement the receiving function can be considered as a receiver, which is used to perform the receiving steps in this embodiment. The device in the communication interface 23 used to implement the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in this embodiment.

[0073] Optional, Figure 2 The structural diagram shown can be used to illustrate the structure of the tree obstacle hazard monitoring system for power transmission channels involved in the above embodiments.

[0074] in, Figure 2 The diagram can also illustrate the system chip in the tree obstruction hazard monitoring system of the power transmission channel. In this case, the actions performed by the aforementioned tree obstruction hazard monitoring system of the power transmission channel can be implemented by this system chip. The specific actions performed can be found above and will not be repeated here.

[0075] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in this embodiment can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0076] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for monitoring tree obstruction hazards in power transmission channels, characterized in that, include: Obtain a subset of the 3D point cloud of vegetation and the 3D center curve of the conductor in the space where the power transmission channel is located; A three-dimensional spatial voxel set is generated based on the vegetation three-dimensional point cloud subset and the three-dimensional traverse center curve, and a medium topology is constructed on the three-dimensional spatial voxel set to obtain an empirical barrier topology network; the edge weight of the connecting edge between adjacent voxels in the empirical barrier topology network corresponds to the medium type of the adjacent voxels. Based on the empirical obstacle topology network and the three-dimensional conductor center curve, a tree obstacle risk quantification analysis is performed to determine the target intersecting voxels and the insulation strength index of the transmission channel. The target intersecting voxel is used to characterize the location of the tree barrier anchor point to be monitored within the transmission channel, and the transmission channel insulation strength index is used to characterize the tree barrier insulation margin of the transmission channel. Tree obstruction hazard monitoring is conducted based on the target intersecting voxels and the insulation strength index of the power transmission channel.

2. The method for monitoring tree obstruction hazards in power transmission channels according to claim 1, characterized in that, A three-dimensional spatial voxel set is generated based on the vegetation three-dimensional point cloud subset and the three-dimensional traverse center curve, and a medium topology is constructed on the three-dimensional spatial voxel set to obtain an empirical barrier topology network, including: Based on the three-dimensional point cloud subset of vegetation and the center curve of the three-dimensional conductor, a spatial bounding box containing the conductor and the ground in the power transmission channel is constructed, and the spatial bounding box is discretized to generate a three-dimensional spatial voxel set. Based on the echo characteristics of radar points inside each voxel in the three-dimensional spatial voxel set, the medium type corresponding to each voxel is determined; the medium type includes air medium voxels, soft leaf medium voxels, and woody branch voxels. The voxels in the three-dimensional space voxel set that contain the center curve of the three-dimensional conductor are taken as the high-voltage source node set, and the voxels in the three-dimensional space voxel set that intersect with the bottom surface of the spatial bounding box are taken as the grounding sink node set. Using voxels in the high-voltage source node set as source nodes and voxels in the ground sink node set as sink nodes, connect adjacent voxels to each other and assign edge weights based on the medium type of the voxels at both ends of each connect edge to obtain an empirical obstruction topology network.

3. The method for monitoring tree obstruction hazards in power transmission channels according to claim 2, characterized in that, Based on the echo characteristics of radar points within each voxel in the three-dimensional spatial voxel set, medium identification is performed to determine the medium type corresponding to each voxel, including: The medium type corresponding to the voxel that does not contain radar points is determined to be the air medium voxel; The medium type corresponding to the voxel whose corresponding echo number is less than the total number of echoes for all internal radar points is determined as the soft leaf medium voxel. The medium type corresponding to the voxel containing at least one radar point whose corresponding echo number is equal to the total number of echoes is determined as the woody branch voxel.

4. The method for monitoring tree obstruction hazards in power transmission channels according to claim 1, characterized in that, Based on the aforementioned empirical obstacle topology network and the aforementioned three-dimensional conductor center curve, a tree obstacle risk quantification analysis is performed to determine the target intersecting voxels and the insulation strength index of the transmission channel, including: In the empirically hindered topology network, iteratively query the connected paths from the source node to the sink node, extract the minimum edge weight on the connected path, and subtract the edge weight of each connected edge on the connected path until there is no connected path from the source node to the sink node in the empirically hindered topology network, thus obtaining the residual network. Based on the residual network, reachability traversal is performed to extract the topological cut edge set, and the target intersecting voxel is determined according to the spatial intersection of the voxel where the endpoint of the connected edge in the topological cut edge set is located and the woody branch voxel in the three-dimensional spatial voxel set. The insulation strength index of the transmission channel is determined based on the edge weights of all connected edges in the set of topological cut edges and the spatial arc length of the center curve of the three-dimensional conductor.

5. The method for monitoring tree obstruction hazards in power transmission channels according to claim 4, characterized in that, In the empirically hindered topology network, iteratively query the connected paths from the source node to the sink node, extract the minimum edge weight on the connected path, and subtract the edge weight of each connected edge on the connected path until there is no connected path from the source node to the sink node in the empirically hindered topology network, obtaining the residual network, including: In each iteration, the empirically hindering topology network of the current iteration is queried for connected paths from the source node to the sink node; the edge weights of each connected edge on the connected path are all positive numbers; Subtract the minimum edge weight from the edge weight of each connected edge on the connected path to generate the empirical obstacle topology network for the next iteration and proceed to the next iteration; If there is no connected path from the source node to the sink node in the current iteration's empirically hindering topology network, then the current iteration's empirically hindering topology network is considered the residual network.

6. The method for monitoring tree obstruction hazards in power transmission channels according to claim 1, characterized in that, Obtain a subset of the 3D point cloud of vegetation and the 3D conductor center curve of the space where the transmission channel is located, including: Acquire visible light images and three-dimensional radar point clouds of the power transmission channel; The visible light image is segmented into instances to generate wire pixel masks and vegetation pixel masks; Based on the traverse pixel mask and the vegetation pixel mask, extract the 3D point cloud subsets of the traverse and the 3D point cloud subsets of the vegetation from the 3D radar point cloud, respectively. Curve fitting is performed on the subset of the three-dimensional point cloud of the conductor to generate the center curve of the three-dimensional conductor.

7. The method for monitoring tree obstruction hazards in power transmission channels according to claim 6, characterized in that, Tree obstruction hazard monitoring is performed based on the target intersecting voxels and the insulation strength index of the transmission channel, including: Obtain the empirical safety benchmark threshold corresponding to the voltage level of the power transmission channel; If the insulation strength index of the power transmission channel is less than the empirical safety benchmark threshold, a tree obstacle hazard alarm is triggered and a monitoring image calibration is performed based on the target intersecting voxels.

8. The method for monitoring tree obstruction hazards in power transmission channels according to claim 7, characterized in that, Image calibration based on the target intersecting voxels includes: A two-dimensional mapping verification is performed on each target intersecting voxel to determine the set of trimming target point coordinates; the set of trimming target point coordinates includes the two-dimensional projection coordinates of the target intersecting voxels to be trimmed on the visible light image; Image rendering calibration is performed based on the set of pruning target coordinates and the insulation strength index of the power transmission channel to generate on-site maintenance images.

9. The method for monitoring tree obstruction hazards in power transmission channels according to claim 8, characterized in that, Perform two-dimensional mapping verification on each intersecting target voxel to determine the set of pruning target point coordinates, including: The three-dimensional coordinates of the geometric center of each intersecting target voxel are mapped to the pixel plane of the visible light image to generate the corresponding two-dimensional projection coordinates. Based on the vegetation pixel mask, the position of the two-dimensional projection coordinates is verified, and the two-dimensional projection coordinates that are not within the vegetation pixel mask are deleted to obtain the set of pruning target point coordinates.

10. A tree obstruction hazard monitoring system for power transmission channels, characterized in that, include: A processor and a communication interface; the communication interface is coupled to the processor, the processor being used to run computer programs or instructions to implement the tree obstruction hazard monitoring method for power transmission channels as described in any one of claims 1-9.

Images