Tree obstacle identification method and system, computer equipment and storage medium

Abstract

The invention relates to the technical field of tree obstacle recognition, and discloses a tree obstacle recognition method and system, computer equipment and a storage medium, and the method comprises the steps: obtaining tower coordinates and conductor parameters of a power transmission line, generating a conductor three-dimensional geometric model through employing an improved catenary model, and carrying out the interval sampling, and obtaining the point cloud data of a simulated conductor; obtaining original topographic data, and performing coordinate alignment, feature extraction and feature fusion on the original topographic data and the simulated traverse point cloud data to obtain multi-modal topographic features; inputting the to-be-predicted time point and the original tree data into the tree height prediction model to obtain a predicted tree height; and calculating the minimum distance between the conductor and the tree according to the simulated conductor point cloud data, the laser radar point cloud data, the multi-modal topographic features and the predicted tree height. According to the invention, through dynamic wire modeling, multi-source data fusion and intelligent risk prediction, risk pre-judgment of the whole life cycle of the power transmission line is realized, and the operation safety and stability of the power transmission line are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tree obstacle recognition, and in particular to a tree obstacle recognition method, system, computer equipment and storage medium. Background Art

[0002] In the power system, transmission lines, as a core component of the nation's backbone power grid, are critical infrastructure for ensuring stable power transmission and economic and social operations. In natural environments, overhead wires are susceptible to uncertainties such as abnormal vegetation growth and extreme weather disasters. This can lead to insufficient safe distances between wires and trees and buildings, thus compromising the safe operation of the power grid.

[0003] Traditional tree barrier identification methods generally rely on periodic manual inspections after line construction or on-site measurements using equipment such as drones and lidar. However, these methods have certain limitations. First, they suffer from a lag: traditional identification methods can only detect hidden dangers after the line is operational, failing to predict potential risks during the planning and construction phases, resulting in passive and costly operations and maintenance. Second, they are poorly adaptable to complex scenarios. In complex environments such as high-density urban blocks and densely vegetated mountainous areas, traditional identification methods struggle to accurately model the spatial relationship between the dynamic shape of conductors and obstacles, leading to prominent data occlusion and noise issues. Furthermore, there is a lack of dynamic prediction capabilities: traditional identification methods lack quantitative analysis of conductor sag changes caused by meteorological conditions and long-term tree growth trends, resulting in insufficient foresight in risk warnings. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a tree obstacle identification method, system, computer equipment and storage medium. Through dynamic conductor modeling, multi-source data fusion and intelligent risk prediction, it can realize risk prediction of the entire life cycle of transmission lines, thereby improving the operational safety and operation and maintenance efficiency of transmission lines.

[0005] In a first aspect, the present invention provides a tree obstacle identification method, the method comprising:

[0006] Obtaining the tower coordinates and conductor parameters of the transmission line, generating a three-dimensional geometric model of the conductor using an improved catenary model, and performing interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data;

[0007] Acquiring original terrain data of the transmission line, performing coordinate alignment, feature extraction, and feature fusion on the original terrain data and the simulated conductor point cloud data to obtain multimodal terrain features, wherein the original terrain data includes lidar point cloud data, image data, and geographic information system data;

[0008] Input the time point to be predicted and the original tree data into the preset tree height prediction model to obtain the predicted tree height;

[0009] The minimum distance between the wire and the tree is calculated based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height, and the potential tree obstacle risk is identified based on a comparison between the minimum distance and a preset safety distance.

[0010] Furthermore, the step of generating a three-dimensional geometric model of the conductor using the improved catenary model includes:

[0011] Performing data fitting on meteorological data to obtain deformation coefficients, wherein the meteorological data includes wind speed, temperature change, and ice thickness, and the deformation coefficients include vertical stretching coefficients, horizontal scaling coefficients, and vertical translation coefficients;

[0012] According to the deformation coefficient, the traditional catenary model is modified to obtain an improved catenary model, and a three-dimensional geometric model of the conductor is generated according to the improved catenary model.

[0013] Furthermore, the improved catenary model is expressed by the following formula:

[0014]

[0015] Where x and y represent coordinate values, cosh(*) represents the hyperbolic cosine function, a represents the vertical stretch coefficient, b represents the horizontal scaling coefficient, and c represents the vertical translation coefficient.

[0016] The vertical stretch coefficient is expressed by the following formula:

[0017] a=a0·(1+k1·V+k2·ΔT+k3·L)

[0018] The horizontal scaling factor is expressed as follows:

[0019] b=b0·(1+k4·V+k5·ΔT+k6·L)

[0020] The vertical translation coefficient is expressed by the following formula:

[0021] c=c0·(1+k7·V+k8·ΔT+k9·L)

[0022] Where a0 represents the initial value of the vertical stretch coefficient, b0 represents the initial value of the horizontal scaling coefficient, c0 represents the initial value of the vertical translation coefficient, V represents the wind speed, ΔT represents the temperature change, L represents the ice thickness, k1 represents the first wind speed coefficient, k4 represents the second wind speed coefficient, k7 represents the third wind speed coefficient, k2 represents the first temperature coefficient, k5 represents the second temperature coefficient, k8 represents the third temperature coefficient, k3 represents the first ice coefficient, k6 represents the second ice coefficient, and k9 represents the third ice coefficient.

[0023] Furthermore, the step of performing coordinate alignment, feature extraction and feature fusion on the original terrain data and the simulated wire point cloud data to obtain multimodal terrain features is as follows:

[0024] Coordinate alignment and spatial superposition are performed on the laser radar point cloud data and the simulated wire point cloud data to obtain fused point cloud data;

[0025] Preprocessing and coordinate aligning the fused point cloud data, image data, and geographic information system data;

[0026] Extracting geometric features of the fused point cloud data using an improved point cloud deep learning model;

[0027] Use a deep neural network model based on dilated convolution to segment image data and extract image features through texture analysis;

[0028] Perform context extraction on GIS data to obtain auxiliary features;

[0029] The geometric features, the image features and the auxiliary features are weightedly fused to obtain multimodal terrain features.

[0030] Furthermore, the steps of constructing the tree height prediction model include:

[0031] Establishing a tree growth prediction model based on historical tree growth data and historical meteorological environment data using a regression analysis algorithm, and obtaining a tree growth index based on the tree growth prediction model;

[0032] Correcting the tree growth prediction formula according to the tree growth index to obtain a tree height prediction model;

[0033] The tree height prediction model is expressed by the following formula:

[0034] H(t)=H0+r·t+ω·G(t)

[0035] Where t represents the time point, H(t) represents the predicted tree height at time point t, H0 represents the initial tree height, r represents the basic annual growth rate, ω represents the environmental correction coefficient, and G(t) represents the tree growth index at time point t.

[0036] Furthermore, the step of calculating the minimum distance between the wire and the tree based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height includes:

[0037] performing point cloud segmentation on the fused point cloud data according to the multimodal terrain features to obtain tree barrier point cloud data, and adjusting the tree barrier point cloud data according to the predicted tree heights to obtain tree barrier prediction point cloud data;

[0038] Establishing an original K-dimensional tree model and a predicted K-dimensional tree model according to the tree obstacle point cloud data and the tree obstacle predicted point cloud data;

[0039] Using a best-first search algorithm, querying the original K-dimensional tree model for each tree obstacle point closest to each conductor point in the simulated conductor point cloud data, and the corresponding first minimum distances;

[0040] Using a best-first search algorithm, querying the predicted tree obstacle prediction points closest to each guide wire point in the simulated guide wire point cloud data and the corresponding second minimum distances from the predicted K-dimensional tree model;

[0041] A minimum distance sequence is obtained according to the first minimum distance and the second minimum distance.

[0042] Furthermore, the step of identifying potential tree obstacle risks based on the comparison relationship between the minimum distance and the preset safety distance includes:

[0043] Compare each distance value in the minimum distance sequence with the preset safety distance, and generate a tree obstacle risk prediction result based on the comparison result;

[0044] A risk heat map is generated based on the tree barrier risk prediction results.

[0045] In a second aspect, the present invention provides a tree obstacle identification system, the system comprising:

[0046] A conductor point cloud generation module is used to obtain the tower coordinates and conductor parameters of the transmission line, generate a three-dimensional geometric model of the conductor using an improved catenary model, and perform interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data;

[0047] A terrain point cloud generation module is used to obtain original terrain data of the transmission line, perform coordinate alignment, feature extraction and feature fusion on the original terrain data to obtain multimodal terrain features. The original terrain data includes lidar point cloud data, image data and geographic information system data;

[0048] The tree height prediction module is used to input the time point to be predicted and the original tree data into a preset tree height prediction model to obtain the predicted tree height;

[0049] The risk identification module is used to calculate the minimum distance between the wire and the tree based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height, and identify potential tree obstacle risks based on a comparison between the minimum distance and a preset safety distance.

[0050] In a third aspect, an embodiment of the present invention further provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0051] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0052] The present invention provides a tree barrier identification method, system, computer device, and storage medium. By incorporating meteorological data into conductor geometry correction through an improved catenary model, the present invention effectively addresses the sag calculation bias caused by existing methods ignoring meteorological conditions, enabling high-precision modeling of transmission line conductor morphology. Through deep fusion and precise registration of multimodal data, the accuracy of tree barrier identification in complex scenarios is enhanced. A tree growth prediction model based on meteorological environmental data accurately predicts tree height changes, and combined with tree barrier point clouds to generate tree barrier virtual point cloud data, this method enables risk prediction throughout the entire life cycle of a transmission line, thereby improving the operational efficiency of the transmission line and ensuring its safety and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 1 is a flow chart of a tree obstacle identification method according to an embodiment of the present invention;

[0054] Figure 2 2 is a schematic structural diagram of a tree obstacle identification system according to an embodiment of the present invention;

[0055] Figure 3 1 is a diagram showing the internal structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0057] See also Figure 1 A tree obstacle identification method proposed in the first embodiment of the present invention includes steps S10 to S40:

[0058] Step S10, obtaining the tower coordinates and conductor parameters of the transmission line, generating a three-dimensional geometric model of the conductor using an improved catenary model, and performing interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data;

[0059] Step S20, obtaining original terrain data of the transmission line, performing coordinate alignment, feature extraction, and feature fusion on the original terrain data to obtain multimodal terrain features, wherein the original terrain data includes lidar point cloud data, image data, and geographic information system data;

[0060] Step S30, inputting the time point to be predicted and the original tree data into a preset tree height prediction model to obtain a predicted tree height;

[0061] Step S40, calculating the minimum distance between the wire and the tree based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height, and identifying potential tree obstacle risks based on a comparison between the minimum distance and a preset safety distance.

[0062] The present invention provides a method for identifying tree barriers based on transmission lines. Tree barriers refer to trees in the canopy or corridor area below the line that threaten the safe operation of the transmission line. In order to accurately identify tree barriers, the conductor shape of the transmission line is first geometrically modeled. In order to improve the accuracy of the modeling, this embodiment uses the tower coordinates and conductor parameters of the transmission line and an improved catenary model to generate a three-dimensional geometric model of the conductor. The conductor parameters include material, cross-sectional dimensions, installation method, etc. Before the construction of the transmission line, the tower coordinates and conductor parameters can be obtained from the planning scheme. After the construction is completed, the tower coordinates and conductor parameters previously obtained can be corrected according to the actual construction situation to obtain more accurate data. After obtaining the tower coordinates and conductor parameters, this embodiment uses the improved catenary model to generate a three-dimensional geometric model of the conductor. The specific steps include:

[0063] Performing data fitting on meteorological data to obtain deformation coefficients, wherein the meteorological data includes wind speed, temperature change, and ice thickness, and the deformation coefficients include vertical stretching coefficients, horizontal scaling coefficients, and vertical translation coefficients;

[0064] According to the deformation coefficient, the traditional catenary model is modified to obtain an improved catenary model, and a three-dimensional geometric model of the conductor is generated according to the improved catenary model.

[0065] The improved catenary model in this embodiment introduces a deformation coefficient based on meteorological data into the traditional catenary model. Specifically, the traditional catenary model is a hyperbolic cosine function that only considers the conductor's own weight and initial tension. Its formula is expressed as:

[0066]

[0067] Where x and y represent the coordinates of the conductor, cosh(*) represents the hyperbolic cosine function, and d is a constant, which is the distance from the vertex of the curve to the abscissa axis. The d value can be calculated by dividing the horizontal tension of the conductor by the weight per unit length of the conductor.

[0068] The above equation shows that the traditional catenary model can only describe the conductor shape under ideal static conditions. In reality, conductors experience diverse meteorological conditions, and these conditions have varying impacts on conductor shape. For example, wind speed, temperature variation, and ice thickness can exert lateral loads, causing conductor sway. This sway alters the conductor's tension distribution, inhibiting its natural horizontal sag and reducing its sag distribution, thereby affecting its horizontal reach. Icing refers to the ice layer that forms on the conductor surface in low-temperature, high-humidity environments. This increases the conductor's vertical load. Under the combined effects of its own weight and ice load, the conductor experiences increased vertical tensile force, leading to increased sag (the amount of sag between its support points). Conductor material expands with heat and contracts with cold, and its linear expansion coefficient determines the extent to which conductor length changes with temperature. When the temperature rises, the conductor expands, increasing its length and increasing sag. When the temperature drops, the conductor contracts, decreasing sag and causing an overall height shift in the conductor.

[0069] In other words, different meteorological conditions have different effects on conductor deformation. For example, the effect of icing on vertical stretching is much greater than its effect on horizontal scaling. The correction of wind speed to horizontal scaling may be negative (suppressing the sag range), while the correction to vertical stretching may be positive (increasing the sag amplitude). Therefore, the impact of meteorological data on conductor deformation needs to be adjusted independently through multiple parameters.

[0070] In this embodiment, according to the influence of the above-mentioned meteorological conditions on the wire deformation, three different coefficients are set, namely the vertical stretching coefficient, the horizontal scaling coefficient and the vertical translation coefficient. Each coefficient is represented by a linear function:

[0071] The vertical stretch coefficient is expressed by the following formula:

[0072] a=a0·(1+k1·V+k2·ΔT+k3·L)

[0073] The horizontal scaling factor is expressed as follows:

[0074] b=b0·(1+k4·V+k5·ΔT+k6·L)

[0075] The vertical translation coefficient is expressed by the following formula:

[0076] c=c0·(1+k7·V+k8·ΔT+k9·L)

[0077] Where a0 represents the initial value of the vertical stretch coefficient, b0 represents the initial value of the horizontal scaling coefficient, c0 represents the initial value of the vertical translation coefficient, V represents the wind speed, ΔT represents the temperature change, L represents the ice thickness, k1 represents the first wind speed coefficient, k4 represents the second wind speed coefficient, k7 represents the third wind speed coefficient, k2 represents the first temperature coefficient, k5 represents the second temperature coefficient, k8 represents the third temperature coefficient, k3 represents the first ice coefficient, k6 represents the second ice coefficient, and k9 represents the third ice coefficient.

[0078] Since different meteorological data have different effects on the geometric characteristics of the conductor, for each deformation coefficient, actual historical data or experimental data are used to perform data fitting to obtain the independent coefficient k in the linear function. i , (i = 1, …, 9). The unified function form simplifies the parameter calibration process while allowing each parameter to respond independently to meteorological changes, avoiding model rigidity caused by shared coefficients. This allows for accurate quantification of the weights of different meteorological data on different geometric characteristics.

[0079] According to the deformation coefficient obtained by data fitting, the traditional catenary model is modified to obtain the improved catenary model:

[0080]

[0081] Where x and y represent the coordinate values of the wire, cosh(*) represents the hyperbolic cosine function, a represents the vertical stretch coefficient, b represents the horizontal scaling coefficient, and c represents the vertical translation coefficient.

[0082] Finally, based on the generated 3D conductor geometry, sampling is performed at preset intervals to generate a simulated conductor point cloud. This embodiment uses deformation coefficients to separate the multidimensional effects of different meteorological conditions on conductor shape. By using independent coefficients to respond to changes in complex meteorological conditions, the conductor shape can be modified in real time, achieving high-precision conductor simulation modeling.

[0083] This embodiment uses a multi-source data fusion approach to analyze the actual terrain to identify obstacles such as trees or buildings. The multi-source data includes lidar point cloud data, image data, and geographic information system data. These data are used as the original terrain data and are then aligned with the simulated wire point cloud data for coordinate alignment, feature extraction, and feature fusion. The specific steps include:

[0084] Coordinate alignment and spatial superposition are performed on the laser radar point cloud data and the simulated wire point cloud data to obtain fused point cloud data;

[0085] Preprocessing and coordinate aligning the fused point cloud data, image data, and geographic information system data;

[0086] Extracting geometric features of the fused point cloud data using an improved point cloud deep learning model;

[0087] Use a deep neural network model based on dilated convolution to segment image data and extract image features through texture analysis;

[0088] Perform context extraction on GIS data to obtain auxiliary features;

[0089] The geometric features, the image features and the auxiliary features are weightedly fused to obtain multimodal terrain features.

[0090] In this embodiment, the lidar point cloud data mainly includes obstacle point cloud data such as trees, vegetation and buildings. Therefore, it is necessary to align and fuse the simulated wire point cloud data with the lidar point cloud data. Specifically, the simulated wire point cloud and the obstacle point cloud (lidar scan) are aligned to the same coordinate system using the nearest point iteration algorithm, and the simulated wire point cloud and the obstacle point cloud are superimposed to form fused point cloud data.

[0091] The fused point cloud data, image data, and GIS data are then preprocessed. Preprocessing the fused point cloud data includes denoising and downsampling. Specifically, statistical filtering (such as radius filtering) is used to remove outlier noise points. Downsampling (such as voxel gridding) is performed on non-critical areas away from the wires, while retaining the high-density point cloud in critical areas near the wires. Furthermore, a random sampling consensus algorithm is used to segment the ground point cloud, retaining point clouds of ground features such as vegetation and buildings.

[0092] In a preferred embodiment, the weight of each point cloud is calculated using the following weighted denoising formula:

[0093]

[0094] Where d(p) represents the distance from point cloud p to the target wire, ε represents a constant used to prevent division by zero, and λ represents the attenuation exponent.

[0095] For each point cloud p, its weight is calculated and compared with the preset weight threshold. Point clouds with weights lower than the weight threshold are removed as noise points to obtain denoised point cloud data.

[0096] Image data preprocessing includes geometric correction and multispectral fusion. Specifically, image distortion is corrected through camera calibration parameters (intrinsic and extrinsic parameters) and distortion models (such as the Brown-Conrady model). For multispectral cameras, RGB and near-infrared bands can also be fused to enhance vegetation recognition capabilities.

[0097] For geographic information system data, or GIS data, preprocessing includes geographic coordinate conversion and vector rasterization. Specifically, GIS data is uniformly converted to the same coordinate system as the lidar point cloud, and vector data (such as transmission line corridor boundaries) is converted to raster layers to facilitate overlay analysis with point clouds and images.

[0098] The preprocessed data also requires coordinate alignment. First, the image data is aligned with the fused point cloud data. Specifically, feature points are extracted from the image data and matched with the 3D feature points of the fused point cloud data. Iterative optimization is performed using the nearest point iteration algorithm, and the image is projected into the point cloud coordinate system. If the drone that captured the image data is equipped with a high-precision global positioning system / inertial measurement unit (GPS / IMU), coarse alignment can be achieved directly using the pose data, reducing the number of iterations. Finally, the geographic coordinates of the image data are overlaid with the GIS raster data to ensure spatial consistency.

[0099] After coordinate alignment, different methods are used to extract features from the fused point cloud data, image data, and geographic information system data. For the fused point cloud data, a buffer zone is first defined around the simulated wire (e.g., 10 meters on both sides), and only the obstacle point cloud within this area is extracted to reduce computational effort and focus on high-risk areas. An improved point cloud deep learning model (PointNet++ model) is then used to perform local geometric feature analysis on the obstacle point cloud, extracting geometric features such as curvature, normal vectors, and height distribution. For the image data, a deep neural network model based on dilated convolution (DeepLabv3+ model) or a deep learning model based on convolutional neural networks (U-Net model) is first used to segment the tree, building, and wire areas in the image data. The vegetation index of the tree area is then extracted to distinguish healthy trees from dead vegetation. For the GIS data, contextual information is extracted to extract terrain slope, aspect, land use type, and other features from the GIS data as auxiliary features. Finally, the geometric features, image features, and auxiliary features are weighted fused to obtain the multimodal terrain feature F:

[0100] F=α·F p +β·F i +γ·F G

[0101] Where, F p represents the geometric feature, α represents the geometric feature weight, F i represents image features, β represents image feature weights, F G represents the auxiliary feature, and γ represents the auxiliary feature weight.

[0102] This embodiment achieves deep collaboration among point cloud data, image data, and GIS data through multi-source data fusion, taking into account geometric accuracy, semantic understanding, and geographic context, thereby providing accurate data support for subsequent tree barrier identification of transmission lines.

[0103] This embodiment establishes a tree height prediction model by analyzing the impact of meteorological environment data on tree growth, thereby predicting tree growth. The specific steps of constructing the model include:

[0104] Based on historical tree growth data and historical meteorological environment data, a tree growth prediction model was established using regression analysis algorithms;

[0105] The tree growth prediction formula is modified according to the tree growth index to obtain a tree height prediction model.

[0106] In this embodiment, the impact of weather and environment on tree growth is first analyzed to establish a tree growth prediction model. The weather and environment data include average temperature, precipitation, wind speed, and soil moisture. Based on historical weather and environment data and historical tree growth data, a multiple regression algorithm or a long-short-term memory neural network model is used to construct a tree growth prediction model. Based on the tree growth prediction model, a tree growth index is determined:

[0107] G=f(T,P,V,D)

[0108] Where G represents the tree growth index, f(*) represents the tree growth prediction model, T represents the average temperature, P represents precipitation, V represents wind speed, and D represents soil moisture.

[0109] Then, based on the tree growth index, the traditional tree growth prediction formula is modified to obtain the tree height prediction model:

[0110] H(t)=H0+r·t+ω·G(t)

[0111] Where t represents the time point, H(t) represents the predicted tree height at time point t, H0 represents the initial tree height, r represents the basic annual growth rate, ω represents the environmental correction coefficient, and G(t) represents the tree growth index at time point t.

[0112] After obtaining the tree height prediction model, the current tree height and the time point to be predicted are input into the model to obtain the predicted tree height.

[0113] Based on the predicted tree heights, combined with simulated traverse point cloud data, LiDAR point cloud data, and multimodal terrain features, the minimum distance between the traverse and the trees can be calculated. The specific steps include:

[0114] performing point cloud segmentation on the fused radar point cloud data according to the multimodal terrain features to obtain tree barrier point cloud data, and adjusting the tree barrier point cloud data according to the predicted tree heights to obtain tree barrier prediction point cloud data;

[0115] Establishing an original K-dimensional tree model and a predicted K-dimensional tree model according to the tree obstacle point cloud data and the tree obstacle predicted point cloud data;

[0116] Using a best-first search algorithm, querying the original K-dimensional tree model for each tree obstacle point closest to each conductor point in the simulated conductor point cloud data, and the corresponding first minimum distances;

[0117] Using a best-first search algorithm, querying the predicted tree obstacle prediction points closest to each guide wire point in the simulated guide wire point cloud data and the corresponding second minimum distances from the predicted K-dimensional tree model;

[0118] A minimum distance sequence is obtained according to the first minimum distance and the second minimum distance.

[0119] In this example, the simulated traverse point cloud data and the LiDAR point cloud data are spatially superimposed and then segmented based on the semantic labels in the multimodal terrain features to generate tree barrier point cloud data. The height coordinates of the tree point clouds in the tree barrier point cloud data are then adjusted based on the predicted tree heights to generate predicted tree barrier point cloud data for a future time point.

[0120] A primitive K-dimensional tree (kd-tree) model of the tree barrier point cloud data is then established. A kd-tree is a spatial partitioning data structure that organizes the point cloud into a tree structure by hierarchically segmenting the tree barrier point cloud. For each conductor point in the simulated conductor point cloud data, a kd-tree nearest neighbor search is performed using a best-first search algorithm to find the nearest neighbor of that conductor point. Specifically, the nodes of the kd-tree are added to a priority queue, which is sorted by the minimum possible distance between the node and the conductor point. The node with the closest distance is extracted from the queue (leaf nodes are prioritized). If it is a leaf node, the Euclidean distance between all points within the node and the conductor point is calculated, and the current nearest point is recorded. If it is a non-leaf node, the spatial relationship between its left and right child nodes and the conductor point is examined, and the closest child node is added to the queue. If the minimum possible distance between the remaining nodes in the queue is greater than the currently recorded nearest distance, the search is terminated. This process is repeated iteratively to determine the nearest tree barrier point for each conductor point and the minimum distance between the conductor point and the nearest tree barrier point.

[0121] Similarly, for the tree barrier prediction point cloud data, the same steps are used to build the original K-dimensional tree model for prediction. The closest tree barrier prediction point to each traverse point is then iteratively calculated, as well as the minimum distance between the traverse point and the nearest tree barrier prediction point. Finally, these minimum distances are sorted and combined to obtain the minimum distance sequence corresponding to each traverse point.

[0122] In a preferred embodiment, to improve query efficiency, a spatial search tag is added when constructing the K-dimensional tree model to quickly filter out candidate point clouds within the buffer zone. This reduces the amount of computation by querying only the tree barrier point clouds within a certain range on both sides of the wire. In addition, when the tree barrier point cloud is updated, an incremental update method is used to reconstruct the K-dimensional tree model only for the new point cloud portion, avoiding global reconstruction and further improving computational efficiency.

[0123] After obtaining the minimum distance sequence for each traverse point, potential tree obstacle risks can be identified based on the minimum distance sequence and the preset safety distance. The specific steps include:

[0124] Compare each distance value in the minimum distance sequence with the preset safety distance, and generate a tree obstacle risk prediction result based on the comparison result;

[0125] A risk heat map is generated based on the tree barrier risk prediction results.

[0126] In this embodiment, a safety distance is first set, and then each distance value in the minimum distance sequence is compared with the safety distance. If the distance value is less than the safety distance, a tree barrier risk is considered to have occurred, and a risk warning reminder is issued. In addition, red warning areas can be marked in the three-dimensional GIS, and based on the dynamic comparison results, risk heat maps for the current and future time periods can be generated to provide accurate data support for subsequent line maintenance.

[0127] This embodiment provides a tree barrier identification method that uses an improved catenary model to introduce meteorological data into conductor geometric parameter correction. This method can effectively solve the sag calculation deviation problem caused by existing methods ignoring meteorological conditions, achieve high-precision modeling of the transmission line conductor shape, and improve the accuracy of tree barrier identification in complex scenarios through deep fusion and precise alignment of multimodal data. Through a tree growth prediction model based on meteorological environmental data, the height changes of trees are accurately predicted, and tree barrier virtual point cloud data is generated in combination with the tree barrier point cloud. This achieves risk avoidance in the transmission line planning stage and risk prediction for the entire life cycle of the transmission line, effectively reducing the subsequent rectification costs, improving the operation and maintenance efficiency of the transmission line, and further ensuring the safety and stability of the transmission line operation.

[0128] See also Figure 2 Based on the same inventive concept, a tree obstacle identification system proposed in a second embodiment of the present invention includes:

[0129] The conductor point cloud generation module 10 is used to obtain the tower coordinates and conductor parameters of the transmission line, generate a three-dimensional geometric model of the conductor using an improved catenary model, and perform interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data;

[0130] A terrain point cloud generation module 20 is used to obtain original terrain data of the transmission line, perform coordinate alignment, feature extraction and feature fusion on the original terrain data to obtain multimodal terrain features, wherein the original terrain data includes lidar point cloud data, image data and geographic information system data;

[0131] The tree height prediction module 30 is used to input the time point to be predicted and the original tree data into a preset tree height prediction model to obtain the predicted tree height;

[0132] The risk identification module 40 is used to calculate the minimum distance between the wire and the tree based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height, and identify potential tree obstacle risks based on a comparison between the minimum distance and a preset safety distance.

[0133] The technical features and effects of the tree obstacle identification system proposed in the embodiments of the present invention are the same as those of the method proposed in the embodiments of the present invention and are not further elaborated here. Each module in the tree obstacle identification system described above can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, allowing the processor to call and execute the corresponding operations of each module.

[0134] In addition, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0135] See also Figure 3 , an internal structure diagram of a computer device in one embodiment, the computer device can specifically be a terminal or a server. The computer device includes a processor, a memory, a network interface, a display and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a tree obstacle identification method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0136] It can be understood by those skilled in the art that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have the same component arrangement.

[0137] In addition, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when the computer program is executed by a processor.

[0138] In summary, embodiments of the present invention propose a tree obstacle identification method, system, computer device, and storage medium. The method obtains the tower coordinates and conductor parameters of a transmission line, uses an improved catenary model to generate a three-dimensional geometric model of the conductor, and performs interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data. The method also obtains original terrain data of the transmission line, and performs coordinate alignment, feature extraction, and feature fusion on the original terrain data and the simulated conductor point cloud data to obtain multimodal terrain features. The original terrain data includes lidar point cloud data, image data, and geographic information system data. The predicted time point and original tree data are input into a preset tree height prediction model to obtain a predicted tree height. The method calculates the minimum distance between the conductor and the tree based on the simulated conductor point cloud data, the lidar point cloud data, the multimodal terrain features, and the predicted tree height, and identifies potential tree obstacle risks based on a comparison between the minimum distance and a preset safety distance. The present invention overcomes the physical limitations of traditional static models through an improved catenary model, achieving high-precision modeling of the transmission line conductor morphology. Through deep fusion and precise alignment of multimodal data, the accuracy of tree obstacle identification in complex scenarios is improved. In combination with the tree growth prediction model, risk avoidance in the transmission line planning stage and risk prediction for the entire life cycle of the transmission line are achieved. The present invention effectively improves the safety and stability of transmission line operation through dynamic conductor modeling, multi-source data fusion and intelligent risk prediction.

[0139] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the various technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0140] The above-described embodiments merely represent several preferred implementations of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art could make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the scope of protection of the claims.

Claims

1. A tree obstacle identification method, characterized in that: include: Obtaining the tower coordinates and conductor parameters of the transmission line, generating a three-dimensional geometric model of the conductor using an improved catenary model, and performing interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data; Acquiring original terrain data of the transmission line, performing coordinate alignment, feature extraction, and feature fusion on the original terrain data and the simulated conductor point cloud data to obtain multimodal terrain features, wherein the original terrain data includes lidar point cloud data, image data, and geographic information system data; Input the time point to be predicted and the original tree data into the preset tree height prediction model to obtain the predicted tree height; The minimum distance between the wire and the tree is calculated based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height, and the potential tree obstacle risk is identified based on a comparison between the minimum distance and a preset safety distance.

2. The tree obstacle identification method according to claim 1, characterized in that: The step of using the improved catenary model to generate a three-dimensional geometric model of the conductor includes: Performing data fitting on meteorological data to obtain deformation coefficients, wherein the meteorological data includes wind speed, temperature change, and ice thickness, and the deformation coefficients include vertical stretching coefficients, horizontal scaling coefficients, and vertical translation coefficients; According to the deformation coefficient, the traditional catenary model is modified to obtain an improved catenary model, and a three-dimensional geometric model of the conductor is generated according to the improved catenary model.

3. The tree obstacle identification method according to claim 1, characterized in that: The improved catenary model is expressed by the following formula: Where x and y represent coordinate values, cosh(*) represents the hyperbolic cosine function, a represents the vertical stretch coefficient, b represents the horizontal scaling coefficient, and c represents the vertical translation coefficient. The vertical stretch coefficient is expressed by the following formula: a=a0·(1+k1·V+k2·ΔT+k3·L) The horizontal scaling factor is expressed as follows: b=b0·(1+k4·V+k5·ΔT+k6·L) The vertical translation coefficient is expressed by the following formula: c=c0·(1+k7·V+k8·ΔT+k9·L) Where a0 represents the initial value of the vertical stretch coefficient, b0 represents the initial value of the horizontal scaling coefficient, c0 represents the initial value of the vertical translation coefficient, V represents the wind speed, △T represents the temperature change, L represents the ice thickness, k1 represents the first wind speed coefficient, k4 represents the second wind speed coefficient, k7 represents the third wind speed coefficient, k2 represents the first temperature coefficient, k5 represents the second temperature coefficient, k8 represents the third temperature coefficient, k3 represents the first ice coefficient, k6 represents the second ice coefficient, and k9 represents the third ice coefficient.

4. The tree obstacle identification method according to claim 1, characterized in that: The step of performing coordinate alignment, feature extraction and feature fusion on the original terrain data and the simulated wire point cloud data to obtain multimodal terrain features: Coordinate alignment and spatial superposition are performed on the laser radar point cloud data and the simulated wire point cloud data to obtain fused point cloud data; Preprocessing and coordinate aligning the fused point cloud data, image data, and geographic information system data; Extracting geometric features of the fused point cloud data using an improved point cloud deep learning model; Use a deep neural network model based on dilated convolution to segment image data and extract image features through texture analysis; Perform context extraction on GIS data to obtain auxiliary features; The geometric features, the image features and the auxiliary features are weightedly fused to obtain multimodal terrain features.

5. The tree obstacle identification method according to claim 1, characterized in that: The steps of constructing the tree height prediction model include: Establishing a tree growth prediction model based on historical tree growth data and historical meteorological environment data using a regression analysis algorithm, and obtaining a tree growth index based on the tree growth prediction model; Correcting the tree growth prediction formula according to the tree growth index to obtain a tree height prediction model; The tree height prediction model is expressed by the following formula: H(t)=H0+r·t+ω·G(t) Where t represents the time point, H(t) represents the predicted tree height at time point t, H0 represents the initial tree height, r represents the basic annual growth rate, ω represents the environmental correction coefficient, and G(t) represents the tree growth index at time point t.

6. The tree obstacle identification method according to claim 4, characterized in that: The step of calculating the minimum distance between the wire and the tree based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features, and the predicted tree height includes: performing point cloud segmentation on the fused point cloud data according to the multimodal terrain features to obtain tree barrier point cloud data, and adjusting the tree barrier point cloud data according to the predicted tree heights to obtain tree barrier prediction point cloud data; Establishing an original K-dimensional tree model and a predicted K-dimensional tree model according to the tree obstacle point cloud data and the tree obstacle predicted point cloud data; Using a best-first search algorithm, querying the original K-dimensional tree model for each tree obstacle point closest to each conductor point in the simulated conductor point cloud data, and the corresponding first minimum distances; Using a best-first search algorithm, querying the predicted tree obstacle prediction points closest to each guide wire point in the simulated guide wire point cloud data and the corresponding second minimum distances from the predicted K-dimensional tree model; A minimum distance sequence is obtained according to the first minimum distance and the second minimum distance.

7. The tree obstacle identification method according to claim 6, characterized in that: The step of identifying potential tree obstacle risks based on a comparison between the minimum distance and a preset safety distance includes: Compare each distance value in the minimum distance sequence with the preset safety distance, and generate a tree obstacle risk prediction result based on the comparison result; A risk heat map is generated based on the tree barrier risk prediction results.

8. A tree obstacle recognition system, characterized in that: include: A conductor point cloud generation module is used to obtain the tower coordinates and conductor parameters of the transmission line, generate a three-dimensional geometric model of the conductor using an improved catenary model, and perform interval sampling on the three-dimensional geometric model of the conductor to obtain simulated conductor point cloud data; A terrain point cloud generation module is used to obtain original terrain data of the transmission line, perform coordinate alignment, feature extraction and feature fusion on the original terrain data to obtain multimodal terrain features. The original terrain data includes lidar point cloud data, image data and geographic information system data; The tree height prediction module is used to input the time point to be predicted and the original tree data into a preset tree height prediction model to obtain the predicted tree height; The risk identification module is used to calculate the minimum distance between the wire and the tree based on the simulated wire point cloud data, the lidar point cloud data, the multimodal terrain features and the predicted tree height, and identify potential tree obstacle risks based on a comparison between the minimum distance and a preset safety distance.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images