Tree information processing method and device, electronic equipment and computer readable medium

Abstract

The embodiment of the invention discloses a tree information processing method and device, electronic equipment and a computer readable medium. A specific embodiment of the method comprises the following steps: acquiring an aircraft patrol orthoimage set and aircraft patrol laser point cloud data; inputting the processed orthoimage set into a tree species image classification model to generate a tree species classification vector diagram; establishing a canopy height model and a vectorized lead three-dimensional model; according to the tree species classification vector diagram, the canopy height model and the tree species natural attribute knowledge base, generating a single tree information set; screening the single tree information set according to the vectorized traverse three-dimensional model and the vegetation growth prediction envelope network three-dimensional model; and sending the obstacle individual tree information set to the target terminal device. According to the implementation mode, the real three-dimensional environment around the power transmission line can be reconstructed, so that the real sag condition of a wire can be simulated, the growth condition of a tree crown can be pre-judged, a barrier tree is cut down in time to avoid frequent flashover, and the safety of the power transmission line is improved.

Description

Technical Field

[0001] Embodiments of this disclosure relate to the field of computer technology, and more specifically to tree information processing methods, apparatus, electronic devices, and computer-readable media. Background Technology

[0002] Tall trees along power transmission line corridors pose a significant safety hazard to these lines, and the continuous growth of obstructive trees can also cause flashover trips. Currently, the methods for felling obstructive trees typically involve identifying them through manual inspections, airborne radar scanning, and the installation of cameras on power poles, followed by timely felling.

[0003] However, when using the above methods to cut down obstructing trees, the following technical problems often arise: manual inspection requires a large number of inspectors, has a long inspection cycle, and is inefficient; airborne radar scanning is affected by flight path angle, line of sight obstruction, and instantaneous operating conditions, making it difficult to obtain a stable three-dimensional relationship between trees, terrain, and power lines; and video monitoring methods are difficult to capture line sag points and predict tree crown height growth, and can only identify obstructing trees in the current state, thus reducing the safety of power transmission lines.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure provide tree information processing methods, apparatuses, electronic devices, and computer-readable media to address one or more of the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide a tree information processing method, which includes: acquiring a set of orthophoto images taken by a drone during patrol of a target power transmission line and drone laser point cloud data, wherein each orthophoto image is an image of an orthophoto taken by a drone during patrol of a target power transmission line after manual annotation; inputting the processed orthophoto image set into a pre-trained tree species image classification model to generate a tree species classification vector map, wherein the processed orthophoto image set is generated after processing the drone orthophoto image set; and establishing a canopy height model and a vectorized conductor 3D model based on the drone laser point cloud data and the tree species classification vector map, wherein... The aforementioned canopy height model characterizes the ground elevation and distribution of vegetation along the target transmission line. Based on the aforementioned tree species classification vector map, the aforementioned canopy height model, and the pre-constructed tree species natural attribute knowledge base, a set of individual tree information is generated, wherein each set of individual tree information includes individual tree location information and individual tree segmentation point cloud data. Based on the aforementioned vectorized conductor 3D model and the pre-constructed vegetation growth prediction envelope network 3D model, the aforementioned set of individual tree information is filtered to determine the set of obstacle individual trees. The aforementioned set of obstacle individual tree information is sent to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation.

[0008] Secondly, some embodiments of this disclosure provide a tree information processing apparatus, comprising: an acquisition unit configured to acquire a set of orthophoto images taken by a drone during patrol of a target power transmission line and drone-taken laser point cloud data, wherein each orthophoto image is an image of an orthophoto taken by a drone during patrol of a target power transmission line and then manually annotated; an input unit configured to input the processed orthophoto image set into a pre-trained tree species image classification model to generate a tree species classification vector map, wherein the processed orthophoto image set is generated after processing the drone-taken orthophoto image set; and a modeling unit configured to establish a canopy height model and a vectorized conductor 3D model based on the drone-taken laser point cloud data and the tree species classification vector map, wherein... The aforementioned canopy height model characterizes the ground elevation and distribution of vegetation along the target transmission line. The generation unit is configured to generate a set of individual tree information based on the aforementioned tree species classification vector map, the aforementioned canopy height model, and a pre-constructed tree species natural attribute knowledge base. Each set of individual tree information includes individual tree location information and individual tree segmentation point cloud data. The filtering unit is configured to filter the set of individual tree information based on the aforementioned vectorized conductor 3D model and a pre-constructed vegetation growth prediction envelope 3D model to determine the set of obstacle individual tree information. The sending unit is configured to send the set of obstacle individual tree information to the target terminal device, allowing the target terminal device user to identify the trees to be felled and to perform the tree felling operation.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0011] The various embodiments of this disclosure have the following beneficial effects: the tree information processing method of some embodiments of this disclosure can construct a three-dimensional model of the transmission line, thereby reconstructing the real three-dimensional environment around the transmission line, simulating the real sag of the conductor, and predicting the growth of the tree crown, thus timely felling obstructing trees to avoid frequent flashovers and improve the safety of the transmission line. Specifically, the reasons for the low safety of the relevant transmission lines are: manual inspection requires a large number of inspectors, with long inspection cycles and low efficiency; airborne radar scanning is affected by flight path angle, line of sight obstruction, and instantaneous operating conditions, making it difficult to obtain a stable three-dimensional relationship between trees, terrain, and conductors; and video monitoring is difficult to capture the sag point of the line and predict the growth of the tree crown, and can only identify obstructing trees in the current state, thus reducing the safety of the transmission line. Based on this, the tree information processing method of some embodiments of this disclosure first acquires a set of orthophoto images taken by the UAV during patrol of the target transmission line and UAV laser point cloud data, wherein the orthophoto images in the above-mentioned set of orthophoto images are images after manual annotation of orthophoto images taken by the UAV during patrol of the target transmission line. Therefore, the true tree height can be obtained by laser penetration through vegetation, thus avoiding the limitation of cameras being unable to measure the true height of the canopy due to obstruction. Next, the processed orthophoto set is input into a pre-trained image classification network model to generate a tree species classification vector map. The processed orthophoto set is generated by processing the aforementioned machine-surveyed orthophoto set. Thus, the neural network model can automatically identify tree species types and generate vector data maps, replacing manual drawing of forest compartment maps, thereby facilitating the direct association of attribute fields (such as tree species type) contained in the vector data with a knowledge base. Next, based on the aforementioned machine-surveyed laser point cloud data and the aforementioned tree species classification vector map, a canopy height model and a vectorized conductor 3D model are established. The canopy height model represents the height and distribution of vegetation along the target transmission line. Thus, a highly accurate canopy height model and vectorized conductor 3D model can be constructed from point cloud data, thereby restoring the vertical distance between the tree canopy and the conductor. Finally, based on the aforementioned tree species classification vector map, the aforementioned canopy height model, and the pre-built tree species natural attribute knowledge base, a set of individual tree information is generated. The individual tree information set includes individual tree location information and individual tree segmentation point clouds. This enables individual tree-level management and dynamic updates of tree growth status, allowing for prediction of tree height within transmission lines. Then, based on the aforementioned vectorized conductor 3D model and the pre-constructed vegetation growth prediction envelope 3D model, the individual tree information set is filtered to obtain the obstacle individual tree information set. This allows for the identification of potential obstacle trees (e.g., trees that may exceed limits within the next year), and the filtering process can promptly remove isolated points, thereby reducing false alarms about obstacle trees.Finally, the aforementioned information on individual trees obstructing the power line is sent to the target terminal device, allowing the user to identify the target obstructing tree and perform tree felling operations. Thus, the point cloud data collected by airborne radar scanning can highly reconstruct the vegetation distribution and line sag along the transmission line, promptly capture sag points, and predict tree canopy height growth, thereby improving the safety of the transmission line. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of the tree information processing method according to the present disclosure; Figure 2 This is a schematic diagram of the structure of some embodiments of the tree information processing apparatus according to the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure; Figure 4 This is a schematic diagram showing the three-dimensional distribution of the target power transmission line and the vegetation along the route, as reconstructed by the tree information processing method disclosed herein. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0016] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0017] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0018] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0019] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Figure 1 A flow 100 of some embodiments of a tree information processing method according to the present disclosure is shown. The tree information processing method includes the following steps: Step 101: Obtain the set of machine-scanned orthophoto images and machine-scanned laser point cloud data.

[0021] In some embodiments, the entity executing the tree information processing method (e.g., a computing device) can acquire a set of drone-based orthophoto images and drone-based laser point cloud data via a wired or wireless connection. The drone-based orthophoto images in the aforementioned set are images of orthophotos taken by a drone equipped with a camera and lidar during patrols of the target power transmission line, after manual annotation. These orthophoto images may include power transmission equipment and vegetation along the route. The power transmission equipment may include, but is not limited to, insulator strings, transmission towers, conductors, and utility poles. The manual annotation refers to manually selecting and marking areas with vegetation in each drone-based orthophoto image. The drone-based laser point cloud data can be point cloud data obtained after the drone's onboard radar performs laser scanning of the power transmission equipment and vegetation along the target power transmission line. The lidar may be a LiDAR lidar.

[0022] Specifically, drones equipped with cameras and LiDAR can generate flight paths along the centerline of power transmission lines, with the flight path height typically 35m-55m higher than the highest point of the transmission line. The onboard camera is set with a side-view angle of 60°–75° to ensure that tree canopies, power lines, and towers can all be captured simultaneously. The LiDAR scanner's scanning frequency and line density are also set to achieve a point cloud density of ≥20pts / m². Furthermore, the flight path overlap can be set according to the width of the transmission corridor (e.g., 120–150m) (e.g., forward overlap 75–80%, lateral overlap 65–70%) to ensure sufficient redundancy between imagery and point cloud data.

[0023] It should be noted that before the aforementioned UAV takes off, the onboard IMU is initialized and calibrated to ensure it can accurately record attitude changes during flight. During its flight path, the UAV simultaneously records imagery, GPS location information (i.e., longitude, latitude, and altitude), and IMU attitude information (e.g., pitch, roll, and yaw angles), and synchronizes all data using a unified timestamp. Furthermore, to ensure that the acquired orthophotos and laser point cloud data are in the same spatial coordinate reference system, several ground control points (GCPs) can be deployed at unobstructed ground locations within the power transmission line corridor. RTK (Real-Time Kinematic) carrier phase differential technology is used to measure and obtain their spatial coordinates with centimeter-level accuracy. During the structure from motion (S3D) image reconstruction process, the image point coordinates of the GCPs are jointly adjusted with their actual coordinates to ensure the geographic location accuracy of the orthophotos meets the requirements. Subsequently, through rigid translation, rotation, and scale correction, the entire airborne laser point cloud data was aligned to the coordinate frame of the GCP, achieving high-precision registration of the airborne orthophoto and the airborne laser point cloud data in the same coordinate system.

[0024] It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra wideband) connections, and other currently known or future wireless connection methods.

[0025] Step 102: Input the processed orthophoto set into the pre-trained tree species image classification network model to generate a tree species classification vector map.

[0026] In some embodiments, the execution entity can input each processed orthophoto from the processed orthophoto set into a pre-trained tree species image classification network model to generate a tree species classification vector map. The tree species classification vector map can be a vector map representing the types and locations of trees along the transmission line. The tree species image classification network model can be a neural network model that takes processed orthophotos as input and the tree species classification vector map as output to classify tree types. The tree species image classification network model can be a network model trained using an orthophoto set containing different types of trees. The tree species image classification network model can optionally use ResNet-50 or ResNet-101 as the feature extraction network and Deeplabv3 as the backbone structure. The input layer receives an RGB tensor corresponding to the processed orthophoto with a size of 512×512×3, and the output layer generates a 512×512 category matrix (each pixel corresponding to a tree species category number) with the same size as the input as a tree species classification mask. The loss function of the above tree image classification network model is the cross-entropy loss function, the optimizer is Adam or SGD, the initial learning rate is set to 0.001, and cosine decay or step decay is used to reduce the learning rate.

[0027] In practice, for each processed orthophoto in the aforementioned set of processed orthophotos, the executing entity can generate an image tensor corresponding to the processed orthophoto, and input the generated image tensor into the aforementioned tree species image classification network model to obtain the corresponding tree species classification mask. Then, the executing entity can stitch the obtained tree species classification masks according to coordinates to form a tree species classification map covering the entire target transmission line, display the tree species classification map in pseudo-color, and further vectorize it (that is, convert continuous areas of the same tree species into polygon vectors and record the corresponding tree species category in the attribute table) to obtain a tree species classification vector map.

[0028] Optionally, before inputting the aforementioned set of orthophoto images into a pre-trained tree species image classification model to generate a tree species classification vector map, the aforementioned execution entity may also perform the following steps: The first step is to perform the following preprocessing steps for each of the above-mentioned orthophoto images from the machine-guided orthophoto image set: The first sub-step involves labeling and cropping the aforementioned orthophoto images to obtain various cropped orthophoto images. In practice, the executing entity first crops out the marked regions within the aforementioned orthophoto images. Then, the executing entity crops the cropped marked region images into 512×512 pixel sub-images. The aforementioned tree species type labels can characterize the types of trees present in the corresponding marked regions.

[0029] The second sub-step involves image enhancement processing on each cropped orthophoto to obtain enhanced orthophotos. In practice, the aforementioned execution entity can first perform image enhancement processing on each cropped orthophoto using the Laplacian operator to obtain enhanced orthophotos.

[0030] The second step is to determine the obtained enhanced orthophotos as a set of processed orthophotos.

[0031] Step 103: Based on the laser point cloud data and tree species classification vector map, establish a canopy height model and a vectorized 3D model of the guide wire.

[0032] In some embodiments, the aforementioned executing entity can establish a canopy height model and a vectorized 3D model of the conductor based on the aforementioned machine-surveyed laser point cloud data and the aforementioned tree species classification vector map. The canopy height model characterizes the ground elevation and distribution of vegetation along the target transmission line.

[0033] In some optional implementations of certain embodiments, the aforementioned execution entity may establish a canopy height model and a vectorized traverse 3D model based on the aforementioned machine-surveyed laser point cloud data and the aforementioned tree species classification vector map through the following steps: The first step is to classify the above-mentioned machine-guided laser point cloud data according to the tree species classification vector map, resulting in vegetation point cloud data, ground point cloud data, conductor point cloud data, and pole / tower point cloud data. In practice, the execution entity first performs coarse classification processing on the above-mentioned machine-guided laser point cloud data according to the elevation threshold and the tree species classification vector map. Each machine-guided laser point cloud containing a Z-axis value less than 1m is classified as ground point cloud data, and each machine-guided laser point cloud containing a Z-axis value greater than or equal to 1m and less than or equal to 10m and whose X-axis and Y-axis coordinate values ​​are located within the vegetation area in the above-mentioned tree species classification vector map is classified as vegetation point cloud data. Then, the aforementioned executing entity can remove the vegetation point cloud data and ground point cloud data from the aforementioned machine-surveyed laser point cloud data, and use DBSCAN clustering and RANSAC algorithms to extract machine-surveyed laser point cloud data that presents an approximately rectangular or tower-shaped structure distribution and machine-surveyed laser point cloud data that presents a linear structure distribution, respectively, as pole-type point cloud data and conductor-type point cloud data.

[0034] The second step is to construct a digital surface model based on the aforementioned vegetation point cloud data. In practice, the aforementioned execution entity can use the Inverse Distance Weighting (IDW) algorithm to perform interpolation operations on all raster positions in the aforementioned vegetation point cloud data to construct the digital surface model.

[0035] The third step is to construct a digital elevation model (DEM) based on the aforementioned ground point cloud data. In practice, firstly, the implementing entity can fit the ground point cloud data to the surface using the iterative least squares interpolation method (ILSQ method). After the initial fitting, outliers with elevation residuals exceeding a threshold are removed, and the fitting is repeated until the constructed initial DEM converges. Then, the implementing entity can rasterize the fitting result (i.e., the initial DEM) to obtain the DEM (resolution such as 0.5m).

[0036] The fourth step is to construct a vectorized 3D model of the conductor based on the point cloud data of the conductor and the tower.

[0037] The fifth step is to perform grid-by-grid subtraction between the above-mentioned digital elevation model and the above-mentioned digital surface model to generate the canopy height model.

[0038] In addressing the tree obstacle identification and safety distance assessment issues mentioned above using this technical solution, the following technical problems arise in the application scenario: within transmission line corridors with complex terrain such as mountains and hills, often consisting of multiple circuits and long spans. These problems are: the point cloud data collected by airborne lidar near the conductor is sparse, discrete, and highly susceptible to instantaneous attitude and operating conditions. Existing methods typically only allow for simple fitting of the conductor point cloud locally or approximation using empirical sag curves. This fails to obtain a continuous three-dimensional conductor model consistent with the tower structure, suspension point location, and conductor mechanical parameters across the entire span. This can lead to significant geometric deviations in the conductor position when calculating the minimum spatial distance between trees and the conductor, making it difficult to obtain a true, continuous, and usable three-dimensional spatial position for safety distance assessment. Consequently, the reliability of identifying obstructing trees based on this type of conductor model is low, and the safety of the transmission line decreases. To address the following requirements for this application scenario: In complex corridor environments, the 3D model of the conductor must not only reflect the actual suspension shape of the conductor, but also be consistent with the tower apex position, mounting point distribution, and mechanical parameters such as conductor type and weight. Furthermore, it must be repeatable and comparable across different inspection batches. Therefore, we have decided to adopt the following solution: In some optional implementations of certain embodiments, the aforementioned execution entity may construct a vectorized 3D model of the conductor based on the aforementioned conductor-type point cloud data and tower-type point cloud data through the following steps: The first step is to obtain the equipment's mechanical parameter information. This includes the conductor type identifier, conductor length, conductor weight, and the weight of the load. The load weight can be the weight of the insulator string. The conductor length can be the designed span or the actually measured span length. The conductor weight can be the weight per unit length of the conductor. In practice, the implementing entity can retrieve the corresponding target transmission line's equipment mechanical parameter information from transmission line design documents, equipment ledgers, or operation and maintenance management systems.

[0039] The second step is to establish a two-dimensional data index in the aforementioned traverse point cloud data to update the traverse point cloud data. In practice, the aforementioned execution entity can first obtain the three-dimensional coordinates of each point in the traverse point cloud (…). , , Projecting the data onto a horizontal plane, retaining only the (x, y) coordinates, and using these plane coordinates as input, a two-dimensional spatial index structure is constructed using the KD-tree algorithm. The constructed two-dimensional KD-tree allows for two main functions: firstly, neighborhood search and local fitting of the conductor point cloud, filtering out isolated noise points and retaining points continuously distributed along the line direction as conductor skeleton points; secondly, it enables efficient querying of conductor points within a specified area based on spatial location, thus obtaining updated conductor-type point cloud data for subsequent segmented fitting of the conductor according to span and search for mounting points.

[0040] The third step is to establish a 3D data index in the aforementioned pole and tower point cloud data to update the data. In practice, the execution entity can directly use the 3D coordinates (x, y, z) of each point in the pole and tower point cloud data as the key value of the KD tree to construct a KD tree index for the pole and tower point cloud in 3D space. This 3D KD tree allows for the rapid retrieval of point cloud clusters in high-elevation spatial regions to identify the apex clusters of each pole. Furthermore, spatial neighborhood queries can be used to cluster the pole and tower point clouds, filtering out scattered points unrelated to the pole and tower structure, thus obtaining updated pole and tower point cloud data. Considering that poles and towers are significant vertical structures, using a 3D KD tree is beneficial for distinguishing the top of the pole from other surrounding structures in the elevation direction.

[0041] The fourth step involves locating the vertices of the updated pole-type point cloud data to determine the position information of each vertices. In practice, the aforementioned execution entity can first stratify the updated pole-type point cloud data according to height values, designating point clouds with elevations within a preset percentile (e.g., the highest 5% or the highest 2m range) as "candidate tower vertex clouds." Subsequently, using these candidate tower vertex clouds as input, a clustering algorithm based on Euclidean distance (e.g., DBSCAN or KD-tree-based neighborhood clustering) can be used to segment the point cloud into multiple spatially separated point clusters, each corresponding to the top structure of a pole. Within each tower vertex cluster, the position of the vertex is determined by calculating the geometric center or taking the point with the maximum elevation, thereby determining the three-dimensional coordinate information (x, y, y) of each tower vertex. t y t , z t This will be used as a spatial reference for subsequent conductor mounting point searches.

[0042] The fifth step involves extracting mounting points from the updated conductor point cloud data based on the determined tower apex positions to determine the location information of each mounting point. In practice, the execution entity can use the 2D KD tree index of the conductor point cloud constructed in the second step to retrieve all conductor skeleton points within a spatial neighborhood of a predetermined radius (e.g., 2m to 5m) for each determined tower apex position. Then, the spatial distance between these conductor skeleton points and the tower apex is determined, and the points with the closest distance and reasonable height (e.g., height difference within 1m) are selected as the conductor mounting points on that tower. In multi-circuit or multi-split conductor scenarios, multiple mounting points with different heights or directions can be identified near the same tower apex, and their three-dimensional coordinates are recorded respectively. Through the above processing, mounting point location information corresponding to each tower is obtained, providing geometric constraints for endpoints and intermediate support points for subsequent conductor grading and catenary equation solving.

[0043] The sixth step involves extracting load points from the updated conductor point cloud data based on the determined tower apex positions to determine the end positions of each conductor. In practice, the aforementioned execution entity can sort each conductor path along the route using the filtered and smoothed conductor skeleton point cloud, and search for the "end positions" of the conductor point cloud in the region near the tower. Specifically, starting from the mounting point, the point cloud sequence can be traced outward along the conductor skeleton direction. When the point cloud density drops below a preset threshold or the spatial connectivity of the conductor skeleton is significantly interrupted, the geometric center of the point cloud near that location is taken as the end position information of that conductor segment. In other words, the conductor end position information corresponds to the start and end positions of the conductor on each tower side, used to delineate the spatial range of conductors with different spans.

[0044] The seventh step involves sorting the position information of each conductor end to obtain a sequence of conductor end position information. In practice, the executing entity can sort the position information of each conductor end according to the overall direction of the transmission line (which can be represented by the order of data collection; typically, a drone starts from the starting point of the target transmission line, patrols along the line, and collects data to the end point). For example, the information can be sorted first by the coordinate values ​​of the plane coordinates of the end points projected onto the line direction (such as the projection distance along the line centerline), or by ascending order using the x-coordinate or chain length parameter along the line direction as the key value. After sorting, all end points are sequentially arranged into a sequence of conductor end position information, such that two adjacent end points in the sequence correspond to the two ends of the same span in space, thus providing ordered geometric endpoint data for the subsequent "construction of the catenary equation according to the span".

[0045] Step 8: Sort the location information of each mounting point to obtain a sequence of mounting point location information. In practice, the executing entity can sort all mounting point location information according to the line direction in a similar manner to the conductor end points. For example, sorting by the projected coordinates of the mounting points in the direction of the line centerline in ascending order yields the mounting point location information sequence. The sorted mounting point location information sequence can spatially correspond to the conductor end point location information sequence, making it easy to select one or more mounting points located within the planar projection interval of any two adjacent end points from the mounting point sequence to serve as intermediate support points for the conductor of that span, thereby achieving automatic matching of mounting points according to the span.

[0046] Step 9: For every two conductor end position information in the above conductor end position information sequence, perform the following steps: The first sub-step involves selecting, based on the two conductor end position information provided above, each mounting point position information sub-sequence that satisfies the interval condition from the aforementioned mounting point position information sequence. In practice, the executing entity can take two adjacent end points P from the conductor end position information sequence. i and P i+1 The projected coordinates of the conductor along the line direction are used as the endpoints of the interval. All mounting points whose projected coordinates fall within this interval are searched in the mounting point location information sequence, forming a subsequence of mounting point location information corresponding to that span. Specifically, for each span of conductor, both ends are determined by adjacent conductor end points, and the subsequence of mounting points located between these two end points serves as the set of intermediate support points for the conductor within that span, thus forming the geometric constraint point set used for solving the catenary equation.

[0047] The second sub-step involves determining the equations for each target curve based on the aforementioned information on the positions of the two conductor ends, the sequence of mounting point positions, and the mechanical parameters of the equipment. These target curve equations can be catenary equations. In practice, the executing entity can use the position information of the two conductor ends within the span as boundary conditions, the sequence of mounting points within the span as additional geometric constraints, and combine this with the mechanical parameters obtained in the first step, such as conductor type, conductor length, conductor unit weight, and load weight, to construct a set of equations for solving the catenary parameters, using the catenary equation as the mathematical model for the conductor shape. Preferably, the catenary equation can be in the form of: z(x) = a × cosh((x - x0) / a) + z0. Wherein, the parameters... These are the parameters of the catenary, and the horizontal tension of the conductor. and the unit weight of the above-mentioned conductors satisfy = / x0 and z0 are translation parameters that allow the curve to pass through spatial constraints such as end points and mounting points. , , respectively, are the coordinates of the conductor in the horizontal projection direction and the height direction. cosh() is the hyperbolic cosine function. The above-mentioned execution entity can use numerical iterative algorithms (such as Newton's iteration, gradient descent, or other nonlinear optimization methods) to solve the parameters of the catenary equation, so that the model curve fits the spatial position of the conductor end and the mounting point as closely as possible, and conforms to the actual distribution position of the conductor-like point cloud, in the sense of minimum mean square error.

[0048] The third sub-step involves establishing a 3D model of the traverse based on the determined equations of each target curve and the point cloud data of the traverse. In practice, the execution entity can discretize the catenary curve within the horizontal projection range of each span according to a preset spatial step size (e.g., 0.5m to 2m). The planar coordinates of the sampling positions are then substituted into the target curve equation (i.e., the corresponding catenary equation) to calculate the corresponding 3D coordinate points, thereby generating a traverse spatial trajectory composed of multiple 3D points. Simultaneously, this curve can be compared with the traverse point cloud, and areas with significant deviations can be locally adjusted or the sampling density increased to improve the fitting accuracy of the 3D traverse model to the actual point cloud. By performing the above operations on all spans, a continuous 3D traverse model can be established across the entire route.

[0049] Step 10: Vectorize the aforementioned 3D traverse model to obtain a vectorized 3D traverse model. In practice, the executing entity can connect the discrete 3D sampling points of each span of the traverse in sequence to generate the corresponding 3D polyline object, or directly store the parameters of the catenary equation as attributes of vector data. For each traverse polyline, information such as its line number, span number, starting and ending tower numbers, and catenary parameters can be added, thus forming a vectorized 3D traverse model that can be directly called in a geographic information system or 3D visualization platform.

[0050] The above-described technical solution and its related content, as an inventive point of this disclosure, combined with step 106, solve the technical problem of "difficulty in obtaining a true, continuous, and usable three-dimensional spatial position of the conductor for safety distance assessment, leading to a reduction in the safety of transmission lines." The factors causing the difficulty in accurately obtaining the three-dimensional position of the conductor in the prior art are often as follows: the point cloud data collected by airborne lidar near the conductor is sparsely distributed in a striped pattern, with insufficient points and significant influence from instantaneous flight attitude, failing to directly reflect the suspension shape of the conductor within the entire span; simultaneously, due to the high overlap between the conductor and the background environment (such as ground and vegetation) in the point cloud, the conductor point cloud is often interfered with by noise points, making it difficult for existing methods to accurately identify key structural information such as conductor suspension points and end points, and failing to provide a continuous and resolvable conductor geometric model for subsequent tree-to-line distance calculations. Solving these factors can improve the safety of transmission lines. To achieve the technical effects of improving the spatial positioning accuracy of conductors, enhancing the accuracy of tree obstacle identification, and improving the reliability of transmission line safety analysis, this disclosure firstly establishes a two-dimensional data index for the conductor point cloud and performs local curve fitting, which effectively suppresses conductor point cloud noise and enables continuous extraction of the conductor skeleton, thereby forming a stable conductor point sequence that can be used for further modeling. Secondly, by establishing a three-dimensional index for the tower point cloud and performing tower top positioning processing, the true spatial position of each tower can be obtained, providing an accurate geometric benchmark for mounting point search and span division. Subsequently, this disclosure utilizes the spatial correspondence between the conductor end point sequence and the mounting point sequence, combined with mechanical parameters such as conductor type, conductor weight, and load weight, to construct the catenary equation, so that the generated three-dimensional conductor model not only geometrically conforms to the distribution trend of the conductor point cloud, but also physically conforms to the force balance and natural suspension law of the conductor. Finally, by spatially discretizing the catenary model over the entire span, a continuous, smooth, and physically meaningful three-dimensional vector model of the conductor is established. Therefore, this disclosure can obtain a three-dimensional spatial position model consistent with the actual conductor state, which can significantly reduce spatial errors caused by sparse point clouds, noise interference, or local missing data, and improve the accuracy of calculating the minimum distance between trees and conductors. Furthermore, because this vectorized three-dimensional conductor model can be used to perform a unified safety distance analysis of the entire line, it can not only reliably and repeatedly identify potentially dangerous trees, but also avoid misjudgments or omissions caused by conductor position estimation deviations. Combined with step 106, this can improve the accuracy of tree obstacle management and enhance the operational safety of transmission lines.

[0051] Step 104: Generate a set of individual tree information based on the tree species classification vector map, canopy height model, and pre-built tree species natural attribute knowledge base.

[0052] In some embodiments, the executing entity can generate a set of individual tree information based on the tree species classification vector map, the canopy height model, and a pre-built tree species natural attribute knowledge base. The individual tree information in the set includes individual tree location information and individual tree segmentation point clouds. The natural attribute knowledge base can be a database storing natural attribute information of different trees.

[0053] In some optional implementations of certain embodiments, the aforementioned execution entity can generate a set of individual tree information based on the aforementioned tree species classification vector map, the aforementioned canopy height model, and a pre-built knowledge base of tree species natural attributes through the following steps: The first step is to smooth the canopy height model to obtain the processed canopy height model. In practice, the execution entity can use a moving window (such as 3×3 or 5×5 pixels) to create a local neighborhood window on the surface of the canopy height model, and fit the quadratic surface equation z(x,y)=ax within the local neighborhood window using the least squares method. 2 +by 2 The coefficients of +cxy+dx+ey+f are used to minimize the error (i.e., model pit) between the fitted surface and the neighborhood height. The residual of each height point is calculated (i.e., the difference between the actual elevation value of the corresponding height point and the corresponding quadratic surface equation value). The weight is determined by the ratio of the residual threshold (e.g., 0.2 to 0.3 m) to each residual. Finally, weighted least squares fitting is performed using the determined weights to obtain the smoothed local height, thereby smoothing the above canopy height model.

[0054] The second step involves processing the processed canopy height model to mark the canopy top, resulting in a set of canopy top location information. In practice, firstly, the execution entity can use a sliding window to mark the maximum pixel value of the processed canopy height model within the window as a candidate pixel. For example, the sliding window size can be a preset maximum canopy diameter. For instance, the window size could be 5m × 5m. Then, the execution entity can remove pixels with a corresponding grid height less than a preset minimum tree height threshold (e.g., 3 to 4m) to prevent shrubs or noise from being misidentified as canopy tops. Next, the execution entity can use a non-maximum suppression algorithm to remove non-maximum pixel values ​​from each candidate pixel at a preset pixel distance, ensuring that each canopy region corresponds to only one canopy top. For example, the preset pixel distance could be 3m. Finally, the execution entity can determine the pixel positions of the removed candidate pixels as the canopy top location information set.

[0055] The third step involves performing background labeling on the processed canopy height model to obtain a background region mask image. In practice, the execution entity can label grids in the processed canopy height model with heights below a preset threshold (e.g., 0.5m) as background categories, and further use morphological opening operations to remove small-area noise, thereby generating the background region mask image. This mask image can serve as a boundary condition for subsequent single-tree segmentation, preventing the segmentation algorithm from misidentifying the ground or gaps as canopy areas.

[0056] The fourth step involves performing individual tree segmentation on the crown height model based on the aforementioned crown position information set and background mask image, resulting in an individual tree grid information set. In practice, the execution entity can use the crown position information in the crown position information set as "seed points" and the background mask as "forbidden regions," then execute a gradient-based or inverse elevation map-based watershed algorithm on the processed CHM. This causes each crown vertex to "expand" into an independent segmented region, consistent with the actual crown extent, ultimately yielding the individual tree grid information set, i.e., the pixel set corresponding to each tree.

[0057] The fifth step is to generate a single-tree information set based on the above single-tree grid information set.

[0058] In some optional implementations of certain embodiments, the aforementioned execution entity may generate a single-tree information set based on the aforementioned single-tree grid information set through the following steps: The first step is to perform the following steps for each individual raster cell in the above set of individual raster cell information: The first sub-step involves segmenting the vegetation point cloud data to obtain individual tree point cloud data. This individual tree point cloud data can be the point cloud data corresponding to a single tree. In practice, the executing entity can retrieve all points falling within the geographic coordinate range represented by each individual tree raster cell in the vegetation point cloud data, and treat these points as a subset of the individual tree point cloud. Subsequently, the executing entity can perform spatial clustering (e.g., DBSCAN) on the individual tree point cloud subset to remove isolated noise, ultimately obtaining the individual tree 3D point cloud data.

[0059] The second sub-step involves determining the tree species and individual tree boundary vector map based on the aforementioned single-tree point cloud data and tree species classification vector map. In practice, the executing entity can map the boundary range represented by the single-tree raster information onto the tree species classification vector map, and crop out the raster region corresponding to the single-tree location. Furthermore, for the tree species category within the raster region, a voting method can be used to select the category with the largest area to determine the category with the highest probability of tree species, thereby determining the tree species category for each tree. Additionally, the executing entity can vectorize the outer envelope (Convex Hull or Alpha Shape) according to the segmentation boundary to form the single-tree boundary vector map.

[0060] The third sub-step involves selecting tree species natural attribute information corresponding to the determined tree species from the aforementioned tree species natural attribute knowledge base as individual tree attribute information. This individual tree attribute information can be information characterizing the natural attributes of the corresponding tree. In practice, the executing entity can retrieve tree species natural attribute information corresponding to the determined tree species from the tree species knowledge base, including but not limited to: typical tree height range, crown width coefficient, average annual growth, crown expansion model parameters, volume and biomass coefficients, etc. Among these, the crown width coefficient is used to describe the empirical proportional relationship between crown width and tree height, generally expressed as C=k. H. C is the crown diameter, k is the crown spread coefficient, and H is the tree height. Crown expansion model parameters are a set of mathematical model parameters used to describe how the crown grows and expands in space. Common mathematical models and corresponding parameters include linear expansion models (common in fast-growing tree species): C t+1 =C t +r, where t is the year and r is the annual crown spread growth (m / year); Logistic growth model (applicable to natural forests): C(t) = Cmax / (1+e -k(t-t 0 ) Where Cmax is the maximum crown size, k is the growth rate, and t0 is the inflection point time (i.e., the time from accelerated to decelerated growth); Allometric biomass ratio model, etc. The above volume and biomass coefficients are empirical formula parameters used to estimate tree volume or tree biomass (including dry matter weight). Timber Volume Equation: V=a DBH b H c Where V is the trunk volume (m³), DBH is the diameter at breast height (cm), H is the tree height, and a, b, and c are the model parameters corresponding to the tree species.

[0061] The fourth sub-step involves determining the aforementioned single-tree point cloud data, single-tree attribute information, and single-tree boundary vector map as single-tree information.

[0062] The second step is to define the individual tree information as a set of individual tree information.

[0063] Step 105: Based on the vectorized 3D model of the conductor and the pre-constructed 3D model of the vegetation growth prediction envelope network, the information set of individual trees is filtered to obtain the information set of obstacle individual trees.

[0064] In some embodiments, the aforementioned execution entity can filter the aforementioned individual tree information set based on the aforementioned vectorized traverse 3D model and the pre-constructed vegetation growth prediction envelope 3D model to obtain the obstacle individual tree information set. The aforementioned vegetation growth prediction envelope 3D model can be a 3D canopy envelope model constructed using triangular mesh (TIN) or voxel form, based on data such as individual tree point clouds, tree species natural attributes, and canopy boundary information, to predict the spatial occupancy range that trees may reach at a certain time in the future (e.g., 1 year, 3 years, 5 years).

[0065] In addressing the technical problems mentioned above, the application scenario—a continuously growing vegetated area within a transmission line corridor—often presents the following challenges: the actual growth rate of individual trees exhibits significant uncertainty due to differences in tree species, terrain conditions, and climate, making it difficult to predict the future height of trees accurately and timely using traditional methods relying on manual inspections or two-dimensional imagery. Simultaneously, the sag position of transmission lines exhibits spatial nonlinearity, making risk analysis based solely on the current tree-to-line distance ineffective in identifying potential obstructive trees that are currently safe but will likely touch the line in the future, thus reducing transmission line safety. Considering the following requirements for this application scenario: the transmission line corridor contains a large number of trees, widely distributed, and of complex species; the future growth trend of these trees impacts the operational safety of the line; to avoid accidents such as line tripping and breakdown, the system needs the ability to predict the potential height and spatial range of different tree species over the next few years, and also the ability to identify potential obstructive trees in advance and perform regional clustering analysis to reduce obstructive vegetation along the transmission line and improve its safety. Therefore, we have decided to adopt the following solution: In some optional implementations of certain embodiments, the execution entity may, through the following steps, filter the individual tree information set based on the vectorized traverse 3D model and the pre-constructed vegetation growth prediction envelope 3D model to obtain the obstacle individual tree information set: The first step is to perform the following processing steps for each piece of information in the above set of individual tree information: The first sub-step involves determining the growth prediction time corresponding to the aforementioned individual tree information. This prediction time corresponds to the target transmission line. In practice, the implementing entity can determine the growth prediction time for the target transmission line based on its operation and maintenance cycle, inspection cycle, or tree obstruction management planning cycle. For example, if the transmission line stipulates "tree obstruction clearing every 3 years," then the growth prediction time can be determined as 3 years. The determined growth prediction time is used to determine the timescale of future tree growth, providing time input parameters for subsequent growth prediction model calculations.

[0066] The second sub-step involves determining the individual tree growth model based on the individual tree attribute information included in the aforementioned individual tree information. This growth model corresponds to model parameters. In practice, the executing entity can retrieve the corresponding tree species' growth model parameters, including average annual height growth, typical maximum height, and crown expansion model parameters, from a pre-built tree species natural attribute knowledge base, according to the tree species category identifier corresponding to the individual tree information. Subsequently, the executing entity can select a suitable growth model for the current individual tree based on the tree species' growth pattern. For example, a linear expansion model can be used for some tree species with obvious linear growth patterns, while a logistic growth model can be used for tree species with a growth slowdown phase. The selected growth model will be used to calculate the height change of the individual tree within the predicted future timeframe.

[0067] The third sub-step involves determining the current tree height based on the individual tree point cloud data included in the aforementioned individual tree information. In practice, the executing entity can extract the three-dimensional coordinates of the highest point of the tree crown from the individual tree point cloud data and determine the corresponding ground elevation value from the ground point cloud or digital elevation model (DEM). The executing entity can then determine the current tree height as the difference between the height of the highest point of the tree crown and the ground elevation. For example, if the height of the highest point of the tree crown is 8.5m and the corresponding ground elevation is 1.0m, then the current tree height can be determined as 7.5m.

[0068] The fourth sub-step generates the height increment based on the determined tree height, the determined individual tree growth model, and the growth prediction time. In practice, the executing entity can use the determined current tree height, the model parameters corresponding to the determined individual tree growth model, and the growth prediction time as input. It then calculates the predicted future height using the individual tree growth model and calculates the height increment based on the difference between the predicted future height and the current height. For example, if the current tree height is 7.5m and the individual tree growth model predicts the tree height will be 8.7m in 3 years, the height increment will be 1.2m.

[0069] The fifth sub-step involves determining the sum of the current tree height and the height increment as the predicted height of a single tree. In practice, the executing entity can sum the height increment with the current tree height to obtain the predicted height of the single tree at the predicted time scale. For example, if the current height is 7.5m and the height increment is 1.2m, the predicted height is 8.7m.

[0070] The sixth sub-step involves aligning the individual tree point cloud data (including the tree information) with the vectorized traverse 3D model. In practice, the execution entity can transform the individual tree point cloud data from its local coordinate system to a global coordinate system consistent with the vectorized traverse 3D model based on sensor extrinsic parameters, attitude calculation parameters, or registration algorithms (such as ICP registration algorithm, NDT registration algorithm, etc.). The transformation process can employ a homogeneous transformation matrix to achieve rigid body transformation, enabling both types of data to be expressed in the same coordinate space, thereby ensuring a consistent reference benchmark for subsequent calculations of tree-line spatial distance.

[0071] The seventh sub-step involves generating tree line prediction distance information based on the aforementioned single tree point cloud data, the three-dimensional model of the vegetation growth prediction envelope network, and the aforementioned three-dimensional model of the vectorized traverse.

[0072] The second step involves clustering the individual tree information set based on the determined interval distances and predicted heights of each tree, resulting in a set of individual tree information groups. In practice, the executing entity can use the predicted treeline distance, predicted height, and planar position of each tree as feature vectors to input into a density-based clustering algorithm (such as DBSCAN or HDBSCAN). By setting neighborhood radii and minimum point count conditions, spatially adjacent trees with small intervals and the potential for future growth can be grouped into individual tree information groups, thereby identifying tree-blocking areas with significant risk. The final set of individual tree information groups represents multiple sub-regions with potential tree-blocking risks.

[0073] The third step involves selecting tree information groups that meet preset quantity conditions from the aforementioned set of individual tree information groups as the obstacle tree information set. In practice, the implementing entity can analyze indicators such as the number of information items included in each tree information group, the average predicted height (average spatial distance of the target tree line, average spatial predicted distance of the target tree line), and the minimum interval distance within the group (specifically, a weighted method can be used to determine the comprehensive scoring method), and filter out eligible tree information groups according to preset quantity conditions (e.g., the number of individual trees in the group ≥ 3). The aforementioned eligible tree information groups can be identified as the obstacle tree information set, representing tree barrier areas with high governance priority.

[0074] It should be noted that the safe tree height for eucalyptus, Christmas trees, water melon trees, rubber trees, and birch trees is generally 25m; for pine, fir, and bamboo forests, it is generally 20m; for fruit trees such as olive and chestnut, it is generally 15m; for fruit trees such as longan, lychee, and orange, it is generally 12m; and for shrubs and miscellaneous trees, it is generally 5-18m.

[0075] The above-mentioned technical solution and its related content, as an inventive point of this disclosure, combined with step 106, solve the technical problem of "difficulty in identifying potential obstructing trees along the target transmission line in advance, resulting in a large amount of obstructing vegetation along the transmission line and reduced transmission line safety." Factors leading to reduced transmission line safety often include: the growth rate and height changes of trees have significant uncertainties, and the growth patterns of different tree species differ significantly, making it difficult to accurately predict future tree heights using simple extrapolation based on current tree height; furthermore, traditional methods typically cannot incorporate the true three-dimensional spatial structure of the conductor, resulting in tree-to-line distance assessments only remaining at the current moment, failing to identify trees that will touch the line in a future year; furthermore, the distribution of trees over a large area exhibits spatial clustering characteristics, while traditional assessment methods often rely on repeated calculations of individual trees, lacking regionalized identification of tree obstacles. To achieve the technical effect of identifying potential obstructing trees in advance and improving transmission line safety, this disclosure, firstly, introduces a growth time scale that matches the line operation and maintenance cycle by determining a growth prediction time for each tree, thereby enabling the assessment of the future growth range of different trees within a unified prediction framework. Secondly, by selecting the corresponding growth model based on the tree species attributes of each tree and combining the current tree height with the predicted future height, the system can more accurately predict the potential height of trees in the coming years. Then, by aligning the coordinate systems of the individual tree point cloud and the conductor 3D model, a unified spatial benchmark is provided for the accurate calculation of the tree-line spatial relationship. Subsequently, by generating tree-line prediction distance information based on the vegetation growth prediction envelope 3D model and conductor discrete points, the system can obtain not only the current minimum tree-line distance but also the minimum tree-line distance for the next few years, thereby identifying trees that may potentially cross the line in advance. Finally, by inputting the tree-line prediction distance, predicted future height, and spatial location features into a density clustering algorithm, potentially hazardous tree areas are identified, and high-risk tree groups meeting preset conditions are selected as the obstacle tree information set. This allows for the early identification of high-risk areas, reducing the frequency of power line trips and flashovers caused by vegetation growth, and improving the safety of power line operation.

[0076] In addressing the technical problems mentioned above, and considering the application scenario—trees in power transmission line corridors continuously grow over time while conductor sag changes with environmental conditions—the following technical issues arise: traditional tree obstruction detection methods only obtain the current tree-to-line spatial distance, failing to assess the future spatial occupancy of trees and accurately predict whether a potential line contact hazard will occur before the next inspection cycle. Furthermore, due to the catenary structure of the conductors, relying solely on current point cloud data for direct minimum distance calculation easily misses crucial risk points such as the lowest sag point, making the tree obstruction risk assessment unreliable. Given the following requirements for this application scenario: power transmission line tree obstruction management not only needs accurate detection of the current tree-to-line distance but also the ability to predict the minimum tree-to-line spatial distance after tree growth over a future period to determine whether a line contact hazard is likely to occur before the next inspection cycle, we have decided to adopt the following solution: In some optional implementations of certain embodiments, the execution entity can generate treeline prediction distance information based on the single tree point cloud data, the vegetation growth prediction envelope 3D model, and the vectorized traverse 3D model through the following steps: The first step is to discretize the aforementioned vectorized 3D traverse model to generate a set of discrete points for the traverse. In practice, the execution entity can spatially sample the traverse according to the catenary curve equation or the spatial shape represented by the vectorized traverse model, at a preset step size (e.g., 0.2–1m), and determine the 3D coordinates of each sampling location as a discrete point of the traverse. After the above sampling process, a complete set of discrete points for the traverse can be obtained, which is used to calculate the minimum spatial distance between the traverse and the trees.

[0077] The second step involves generating the spatial distances of each treeline based on the aforementioned 3D model of the vegetation growth prediction envelope. In practice, the executing entity can represent the canopy space as a triangular network (TIN) or voxel structure according to the 3D model of the vegetation growth prediction envelope. Subsequently, the executing entity can calculate the minimum 3D distance from each discrete point of the traverse to the surface of the predicted canopy envelope, thereby obtaining the spatial distances of each treeline.

[0078] The third step is to select the minimum value from the generated treeline spatial distances as the first treeline spatial distance. In practice, the aforementioned execution entity can select the minimum value from the generated treeline spatial distances and determine this minimum value as the first treeline spatial distance. This distance represents the position of the traverse closest to the predicted tree crown in the entire set of discrete points of the traverse.

[0079] The fourth step involves determining the treeline spatial distance corresponding to the target curve equation based on the individual tree information, using this distance as the second treeline spatial distance. In practice, the executing entity can obtain the three-dimensional coordinates of the lowest sag point by using the target curve equation (i.e., the corresponding catenary equation) corresponding to the individual tree information. Subsequently, the executing entity can calculate the minimum distance from the three-dimensional coordinates of the lowest sag point to the predicted canopy envelope network, and determine this distance as the second treeline spatial distance to supplement any extreme points that may have been missed by discrete sampling.

[0080] The fifth step is to determine the target tree-line spatial distance based on the first and second tree-line spatial distances mentioned above. In practice, the executing entity can compare the first and second tree-line spatial distances and determine the smaller one as the target tree-line spatial distance, which characterizes the minimum spatial distance between the tree and the guide wire at the current prediction time.

[0081] Step 6: Based on the generated predicted height of each individual tree, generate the spatial predicted distances for each treeline. In practice, the above-mentioned execution entity can expand the crown height or crown extent according to the predicted height of each individual tree (for example, through the crown width coefficient and the corresponding crown width equation, C=k). (H, C are the canopy diameter, k is the canopy width coefficient, and H is the tree height) to construct a future prediction canopy envelope network. Subsequently, the aforementioned execution entity can recalculate the minimum distance from each traverse discrete point to the future prediction canopy surface, forming a future prediction distance sequence.

[0082] Step 7: Based on the target curve equation corresponding to the individual tree information, determine the treeline spatial prediction distance corresponding to the target curve point. In practice, the executing entity can determine the minimum distance from the three-dimensional coordinates of the lowest point of the sag obtained by the target curve equation (i.e., the corresponding catenary equation) corresponding to the individual tree information to the predicted canopy envelope network, and determine this distance as the treeline spatial prediction distance.

[0083] Step 8: Determine the target treeline spatial prediction distance based on the determined spatial prediction distances of each treeline. In practice, the executing entity can select the minimum value from the various treeline spatial prediction distances as the target treeline spatial prediction distance.

[0084] The ninth step involves defining the determined target treeline spatial distance and the determined target treeline spatial predicted distance as the treeline predicted distance information. In practice, the executing entity can combine the aforementioned target treeline spatial distance (current moment) with the target treeline spatial predicted distance (future moment) to generate the treeline predicted distance information. This treeline predicted distance information can represent: the current minimum treeline distance and the future minimum predicted treeline distance, thereby determining whether the current distance between treelines constitutes an obstacle to the target transmission line, and whether the distance between treelines will constitute an obstacle to the target transmission line before the next inspection cycle.

[0085] The above-mentioned technical solution and its related content, combined with the processing steps of the conductor model and vegetation growth prediction envelope network by the implementing entity, serve as an inventive point of this disclosure embodiment. Combined with step 106, it solves the problem of "the inability to assess the spatial occupancy range of trees after future growth, and the inability to accurately determine whether there will be a potential line contact hazard before the next inspection cycle, thus leading to a reduction in the safety of transmission lines." Factors leading to a reduction in transmission line safety mainly include: the inability to quantify the future growth trend of trees, the inability of the current tree-to-line distance to reflect future potential hazards, and the tendency for key risk points under the catenary structure of the conductor (such as the lowest point of sag) to be easily ignored in routine sampling, thus reducing the safety of the transmission line. To solve the above problems, this disclosure first performs discrete sampling of the conductor based on the catenary model to obtain a set of discrete points that can completely characterize the spatial morphology of the conductor. Subsequently, based on the vegetation growth prediction envelope network model, it calculates the minimum distance from each discrete point of the conductor to the canopy surface, obtaining the spatial distance of each tree-line, and selecting the minimum value among them as the first tree-line spatial distance. Next, the location of the lowest point of the sag is obtained through the target curve equation of the conductor, and the distance from it to the canopy envelope is calculated to obtain the second treeline spatial distance. The minimum value is taken as the target treeline spatial distance at the current prediction time. Simultaneously, this disclosure also expands the canopy envelope based on the predicted height of a single tree to recalculate the future treeline spatial distance, thus obtaining the predicted target treeline spatial distance at the future prediction time. Finally, by combining the current treeline distance and the future treeline distance to form treeline predicted distance information, it is possible to determine whether a tree currently constitutes an obstacle and whether it will form an obstacle before the next inspection cycle. This allows for accurate identification of current obstructing trees and early identification of potential obstructing trees that may touch the line in the future, reducing the risk of tripping due to tree obstruction growth and improving the safety of transmission line operation.

[0086] Step 106: Send the set of obstacle tree information to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation.

[0087] In some embodiments, the executing entity may send the set of individual obstacle tree information to a target terminal device, so that the user of the target terminal device can identify the trees to be felled and carry out the tree felling operation. The target terminal device may be a mobile terminal device. For example, the target terminal may be a mobile phone or tablet device used by a power transmission line manager. In practice, the executing entity may send the set of individual obstacle tree information to the target terminal device, so that the user of the target terminal device can identify the target obstacle trees and use tree felling equipment to carry out the tree felling operation.

[0088] In some optional implementations of certain embodiments, the aforementioned executing entity may send the aforementioned set of obstacle tree information to the target terminal device through the following steps, so that the target terminal device user can identify the trees to be felled and perform the tree felling operation: The first step is to encrypt the aforementioned set of individual obstacle tree information to obtain an encrypted set of individual obstacle tree information. In practice, the executing entity can use an encryption algorithm (such as the RSA algorithm) to encrypt the aforementioned set of individual obstacle tree information to obtain an encrypted set of individual obstacle tree information.

[0089] The second step is to send the encrypted set of individual obstacle tree information to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation.

[0090] The various embodiments of this disclosure have the following beneficial effects: the tree information processing method of some embodiments of this disclosure can construct a three-dimensional model of the transmission line, thereby reconstructing the real three-dimensional environment around the transmission line, simulating the real sag of the conductor, and predicting the growth of the tree crown, thus timely felling obstructing trees to avoid frequent flashovers and improve the safety of the transmission line. Specifically, the reasons for the low safety of the relevant transmission lines are: manual inspection requires a large number of inspectors, with long inspection cycles and low efficiency; airborne radar scanning is affected by flight path angle, line of sight obstruction, and instantaneous operating conditions, making it difficult to obtain a stable three-dimensional relationship between trees, terrain, and conductors; and video monitoring is difficult to capture the sag point of the line and predict the growth of the tree crown, and can only identify obstructing trees in the current state, thus reducing the safety of the transmission line. Based on this, the tree information processing method of some embodiments of this disclosure first acquires a set of orthophoto images taken by the UAV during patrol of the target transmission line and UAV laser point cloud data, wherein the orthophoto images in the above-mentioned set of orthophoto images are images after manual annotation of orthophoto images taken by the UAV during patrol of the target transmission line. Therefore, the true tree height can be obtained by laser penetration through vegetation, thus avoiding the limitation of cameras being unable to measure the true height of the canopy due to obstruction. Next, the processed orthophoto set is input into a pre-trained image classification network model to generate a tree species classification vector map. The processed orthophoto set is generated by processing the aforementioned machine-surveyed orthophoto set. Thus, the neural network model can automatically identify tree species types and generate vector data maps, replacing manual drawing of forest compartment maps, thereby facilitating the direct association of attribute fields (such as tree species type) contained in the vector data with a knowledge base. Next, based on the aforementioned machine-surveyed laser point cloud data and the aforementioned tree species classification vector map, a canopy height model and a vectorized conductor 3D model are established. The canopy height model represents the height and distribution of vegetation along the target transmission line. Thus, a highly accurate canopy height model and vectorized conductor 3D model can be constructed from point cloud data, thereby restoring the vertical distance between the tree canopy and the conductor. Finally, based on the aforementioned tree species classification vector map, the aforementioned canopy height model, and the pre-built tree species natural attribute knowledge base, a set of individual tree information is generated. The individual tree information set includes individual tree location information and individual tree segmentation point clouds. This enables individual tree-level management and dynamic updates of tree growth status, allowing for prediction of tree height within transmission lines. Then, based on the aforementioned vectorized conductor 3D model and the pre-constructed vegetation growth prediction envelope 3D model, the individual tree information set is filtered to obtain the obstacle individual tree information set. This allows for the identification of potential obstacle trees (e.g., trees that may exceed limits within the next year), and the filtering process can promptly remove isolated points, thereby reducing false alarms about obstacle trees.Finally, the aforementioned information on individual trees obstructing the power line is sent to the target terminal device, allowing the user to identify the target obstructing tree and perform tree felling operations. Thus, the point cloud data collected by airborne radar scanning can highly reconstruct the vegetation distribution and line sag along the transmission line, promptly capture sag points, and predict tree canopy height growth, thereby improving the safety of the transmission line.

[0091] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a tree information processing device, which are similar to... Figure 1 Corresponding to the method embodiments shown, this tree information processing device can be specifically applied to various electronic devices.

[0092] like Figure 2 As shown, a tree information processing device 200 in some embodiments includes: an acquisition unit 201, an input unit 202, a modeling unit 203, a generation unit 204, a filtering unit 205, and a sending unit 206. The acquisition unit 201 is configured to acquire a set of orthophoto images taken by a drone during patrol of a target power transmission line and drone-taken laser point cloud data. Each orthophoto image is an image of an orthophoto taken by a drone during patrol of a target power transmission line, after manual annotation. The input unit 202 is configured to input the processed orthophoto image set into a pre-trained tree species image classification model to generate a tree species classification vector map. The processed orthophoto image set is generated after processing the drone-taken orthophoto image set. The modeling unit 203 is configured to establish a canopy height model and a vectorized conductor 3D model based on the drone-taken laser point cloud data and the tree species classification vector map. The canopy height model represents the target power transmission line. The system includes the height and distribution of vegetation along the power line; the generation unit 204 is configured to generate a set of individual tree information based on the tree species classification vector map, the canopy height model, and a pre-built tree species natural attribute knowledge base, wherein each set of individual tree information includes individual tree location information and individual tree segmentation point cloud data; the filtering unit 205 is configured to filter the set of individual tree information based on the vectorized conductor 3D model and a pre-built vegetation growth prediction envelope 3D model to determine the set of obstacle individual tree information; and the sending unit 206 is configured to send the set of obstacle individual tree information to the target terminal device so that the target terminal device user can identify the trees to be felled and perform the tree felling operation.

[0093] It is understandable that the units recorded in the tree information processing device 200 are related to the reference. Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the tree information processing device 200 and the units contained therein, and will not be repeated here.

[0094] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0095] like Figure 3 As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory 302 or a program loaded from a storage device 308 into a random access memory 303. The random access memory 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, the read-only memory 302, and the random access memory 303 are interconnected via a bus 304. An input / output interface 305 is also connected to the bus 304.

[0096] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0097] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 309, or installed from a storage device 308, or installed from a read-only memory 302. When the computer program is executed by the processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0098] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0099] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0100] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire a set of drone-guided orthophoto images and drone-guided laser point cloud data, wherein each drone-guided orthophoto image is an image of an orthophoto taken by a drone during patrol of a target power transmission line and then manually annotated; input the processed orthophoto image set into a pre-trained tree species image classification model to generate a tree species classification vector map, wherein the processed orthophoto image set is generated after processing the aforementioned drone-guided orthophoto image set; and establish a canopy height model and a vectorization guide based on the aforementioned drone-guided laser point cloud data and the aforementioned tree species classification vector map. A three-dimensional model of the line is generated, wherein the canopy height model represents the ground elevation and distribution of vegetation along the target transmission line; based on the tree species classification vector map, the canopy height model, and a pre-constructed tree species natural attribute knowledge base, a set of individual tree information is generated, wherein each set of individual tree information includes individual tree location information and individual tree segmentation point cloud data; based on the vectorized conductor three-dimensional model and the pre-constructed vegetation growth prediction envelope three-dimensional model, the set of individual tree information is filtered to determine the set of obstacle individual trees; the set of obstacle individual tree information is sent to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation.

[0101] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0102] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0103] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, an image preprocessing unit 202, an input unit, a modeling unit, a generation unit, a filtering unit, and a sending unit. The names of these units do not necessarily limit the specific unit; for example, the acquisition unit may also be described as "a unit that acquires a set of machine-scanned orthophoto images and machine-scanned laser point cloud data."

[0104] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0105] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A tree information processing method, comprising: Acquire a collection of orthophoto images and laser point cloud data from drone patrols. Each orthophoto image is an image with manual annotations of an orthophoto taken by a drone while patrolling a target power transmission line. The processed orthophoto set is input into a pre-trained tree species image classification model to generate a tree species classification vector map, wherein the processed orthophoto set is generated by processing the machine-surveyed orthophoto set; Based on the laser point cloud data and the tree species classification vector map, a canopy height model and a vectorized 3D model of the conductor are established. The canopy height model represents the height and distribution of vegetation along the target transmission line. Based on the tree species classification vector map, the canopy height model, and the pre-built tree species natural attribute knowledge base, a set of individual tree information is generated, wherein each individual tree information includes individual tree location information and individual tree segmentation point cloud data; Based on the vectorized traverse 3D model and the pre-constructed vegetation growth prediction envelope 3D model, the individual tree information set is filtered to determine the obstacle individual tree information set; The information set of individual trees in the obstacle is sent to the target terminal device so that the user of the target terminal device can identify the trees to be felled and carry out the tree felling operation.

2. The method according to claim 1, wherein, Before inputting the machine-surveyed orthophoto set into a pre-trained tree species image classification model to generate a tree species classification vector map, the method further includes: For each machine-scan orthophoto, perform the following image processing steps: The orthophoto images are labeled and cropped to obtain various cropped orthophoto images; Image enhancement processing is performed on each cropped orthophoto to obtain each enhanced orthophoto; The obtained enhanced orthophotos are defined as a set of processed orthophotos.

3. The method according to claim 1, wherein, The step of establishing a canopy height model and a vectorized 3D traverse model based on the machine-guided laser point cloud data and the tree species classification vector map includes: Based on the tree species classification vector map, the machine-guided laser point cloud data is classified to obtain vegetation point cloud data, ground point cloud data, conductor point cloud data, and tower point cloud data. Based on the vegetation point cloud data, a digital surface model is constructed; Based on the aforementioned ground point cloud data, a digital elevation model is constructed; Based on the point cloud data of conductors and towers, a vectorized 3D model of the conductor is constructed. The digital elevation model and the digital surface model are subtracted grid by grid to generate the canopy height model.

4. The method according to claim 3, wherein, The step of generating a set of individual tree information based on the tree species classification vector map, the canopy height model, and a pre-built knowledge base of tree species natural attributes includes: The canopy height model is smoothed to obtain the processed canopy height model; The processed canopy height model is then subjected to canopy top marking to obtain a set of canopy top position information. The processed canopy height model is subjected to background marking processing to obtain a background region mask image; Based on the crown position information set and the background region mask image, the crown height model is segmented into individual trees to obtain a set of individual tree grid information. Based on the set of individual tree grid information, generate a set of individual tree information.

5. The method according to claim 4, wherein, The step of generating a single-tree information set based on the single-tree grid information set includes: For each individual grid cell in the set of individual grid cell information, perform the following steps: The vegetation point cloud data is segmented to obtain individual tree point cloud data; Based on the single-tree point cloud data and the tree species classification vector map, the tree species and single-tree boundary vector map are determined; Select the tree species natural attribute information corresponding to the determined tree species from the tree species natural attribute knowledge base as the individual tree attribute information; The single-tree point cloud data, single-tree attribute information, and single-tree boundary vector map are defined as single-tree information; The determined information of each individual tree is defined as a set of individual tree information.

6. The method according to claim 5, wherein, The step of sending the set of obstacle tree information to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation includes: The set of obstacle tree information is encrypted to obtain an encrypted set of obstacle tree information. The encrypted set of individual obstacle tree information is sent to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation.

7. A tree information processing device, comprising: The acquisition unit is configured to acquire a set of orthophoto images and laser point cloud data from the drone patrol. Each orthophoto image is an image that has been manually annotated from an orthophoto image taken by the drone while patrolling the target power transmission line. The input unit is configured to input the processed orthophoto set into a pre-trained tree species image classification model to generate a tree species classification vector map, wherein the processed orthophoto set is generated by processing the machine-surveyed orthophoto set. The modeling unit is configured to establish a canopy height model and a vectorized conductor 3D model based on the machine-scanned laser point cloud data and the tree species classification vector map, wherein the canopy height model represents the ground height and distribution of vegetation along the target transmission line; The generation unit is configured to generate a set of individual tree information based on the tree species classification vector map, the canopy height model, and a pre-built tree species natural attribute knowledge base, wherein each set of individual tree information includes individual tree location information and individual tree segmentation point cloud data; The filtering unit is configured to filter the set of individual tree information based on the vectorized traverse 3D model and the pre-constructed 3D model of vegetation growth prediction envelope network, so as to determine the set of obstacle individual tree information. The sending unit is configured to send the set of obstacle tree information to the target terminal device so that the target terminal device user can identify the trees to be felled and carry out the tree felling operation.

8. An electronic device, comprising: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 6.

9. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images