Safety inspection and identification evaluation method and system for overhead transmission line based on laser point cloud

Abstract

The application provides a kind of overhead transmission line safety inspection identification evaluation method and system based on laser point cloud, related to the technical field of vegetation growth evaluation, the application obtains vegetation point cloud distribution data by collecting power line corridor multi-period multi-view laser point cloud and meteorological data, point cloud splicing is segmented, three-dimensional reconstruction is generated in combination with multidimensional index to generate vegetation model, point cloud features are extracted therefrom and spatiotemporal coding fusion is carried out with meteorological characteristics, vegetation growth change data is calculated based on the fusion features through the growth model, and the growth evaluation result is generated after comparing with the safety threshold, accurate monitoring of vegetation growth state is realized, and automatic accurate monitoring and safety warning of vegetation growth state of power line corridor can be realized.

Description

Technical Field

[0001] This application relates to the technical field of vegetation growth assessment, and in particular to a safety inspection and identification assessment method and system for overhead power transmission lines based on laser point clouds. Background Technology

[0002] During the operation and maintenance of power transmission line corridors, continuous monitoring and accurate assessment of vegetation growth along the route is crucial. With the widespread adoption of lidar technology, a safety inspection and identification assessment method for overhead power transmission lines based on lidar point clouds provides a new technological approach to this need, enabling non-contact and precise measurement of the three-dimensional structure of vegetation.

[0003] Currently, existing technical solutions typically employ a combination of manual inspections and photogrammetry for vegetation monitoring. Operators periodically conduct on-site surveys of the corridor, acquiring basic vegetation parameters using handheld measuring devices, while simultaneously obtaining two-dimensional image data through aerial or ground photography. Image processing techniques are then used to analyze vegetation cover. Another common approach is to utilize single-period laser point cloud data, identifying vegetation areas through point cloud classification algorithms, and performing basic morphological parameter statistics based on this.

[0004] However, the most significant limitation of existing technologies lies in their inability to establish a precise correlation between multi-period observation data and dynamic changes in vegetation growth, and the lack of analytical methods to effectively link environmental meteorological factors with the vegetation growth process. This limitation leads to judgments on vegetation growth trends based on static observations and empirical inferences, which are insufficient to meet the needs for accurate prediction and early warning of vegetation growth status along transmission line corridors. Summary of the Invention

[0005] The purpose of this application is to provide a safety inspection, identification, and assessment method and system for overhead power transmission lines based on laser point clouds, in order to solve the problem that existing technologies cannot accurately predict and warn of the vegetation growth status of power transmission line corridors.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for safety inspection, identification, and assessment of overhead power transmission lines based on laser point clouds, comprising:

[0007] Collect multi-view laser point cloud data and corresponding meteorological environmental data of the same transmission line corridor area at different times, and extract meteorological features from the meteorological environmental data;

[0008] The multi-view laser point cloud data from various periods are stitched together to obtain complete point cloud data. The complete point cloud data is then segmented to separate the regions containing vegetation points, thus obtaining vegetation point cloud distribution data.

[0009] Based on the vegetation point cloud distribution data, combined with the multi-dimensional indicators of vegetation, a three-dimensional reconstruction process is performed to obtain a three-dimensional vegetation model.

[0010] Vegetation point cloud features are extracted from the three-dimensional vegetation model, and the vegetation point cloud features are fused with the meteorological features through spatiotemporal feature encoding to obtain spatiotemporal fused features.

[0011] Based on the spatiotemporal fusion features, vegetation growth change data is calculated using a pre-constructed vegetation growth model. The growth change data is then compared with a preset safety threshold, and a vegetation growth assessment result is generated based on the comparison result.

[0012] Optionally, the step of performing three-dimensional reconstruction processing based on the vegetation point cloud distribution data and combining it with multi-dimensional vegetation indicators to obtain a three-dimensional vegetation model includes:

[0013] Point cloud density analysis is performed on the vegetation point cloud distribution data to obtain the core growth area with a point cloud density greater than a first preset density threshold.

[0014] Based on spatial proximity, the point cloud in the core growth area is clustered to obtain multiple point groups, each point group corresponding to an independent vegetation.

[0015] For each point group, the outer contour shape of the point group is calculated to obtain a surface structure that reflects the external morphology of the vegetation. Combined with the multi-dimensional indicators of the vegetation, the detailed features of the surface structure are supplemented and optimized to obtain a three-dimensional model of the vegetation.

[0016] Optionally, the process of combining multi-dimensional vegetation indicators to supplement and optimize the detailed features of the surface structure to obtain a three-dimensional vegetation model includes:

[0017] The overall height of vegetation is determined based on the vertical distribution range of the point cloud, and the point cloud is divided into several layers in the vertical direction according to the overall height.

[0018] Calculate the point cloud density within each layer, and identify the layer with a point cloud density greater than the second preset density threshold and the largest horizontal coverage as the main area of ​​the vegetation canopy.

[0019] The point cloud of the main canopy region is divided into multiple sub-regions based on the local density gradient.

[0020] For each sub-region, the sub-volume corresponding to the sub-region is estimated based on the number of point clouds in the sub-region. Based on the point cloud density and sub-volume of the sub-region, the parameters of the surface reconstruction algorithm are adaptively adjusted. Based on the adjusted parameters, the sub-region is reconstructed in three dimensions to generate the sub-model corresponding to the sub-region. The sum of the sub-volumes corresponding to all sub-regions is the theoretical total volume, and the theoretical total volume, overall height, and point cloud density within the layer are used to constitute the multi-dimensional indicators of vegetation.

[0021] Merge the sub-models corresponding to all sub-regions and calculate the total volume of the merged model;

[0022] When the error between the total volume of the fused model and the theoretical total volume is greater than a preset error, the total volume of the fused model is corrected by scaling transformation until the error between the total volume of the fused model and the theoretical total volume is less than or equal to the preset error. During the correction process, the overall height remains unchanged, and finally a three-dimensional vegetation model is obtained.

[0023] Optionally, the step of extracting vegetation point cloud features from the 3D vegetation model and then performing spatiotemporal feature encoding and fusion of the vegetation point cloud features with the meteorological features to obtain spatiotemporal fused features includes:

[0024] Calculate the vertical dimensions of the vegetation 3D model to obtain the height feature; calculate the horizontal coverage of the vegetation 3D model to obtain the crown width feature; calculate the spatial size of the vegetation 3D model to obtain the volume feature.

[0025] Arrange the height feature, crown width feature, and volume feature in chronological order to obtain vegetation point cloud features;

[0026] By using a spatiotemporal feature encoding method, the vegetation point cloud features and corresponding meteorological features from the same period are processed to obtain spatiotemporal fusion features.

[0027] Optionally, the step of processing the vegetation point cloud features and corresponding meteorological features from the same period through a spatiotemporal feature encoding method to obtain spatiotemporal fusion features includes:

[0028] The vegetation point cloud features and meteorological features from the same period are combined to obtain multiple fused feature units;

[0029] All the fused feature units are arranged in chronological order to obtain time series features;

[0030] The fused feature units of adjacent periods in the time series features are analyzed to obtain feature change trend information;

[0031] By using a spatiotemporal feature encoding method, the time series features and the feature change trend information are integrated and processed to obtain spatiotemporal fusion features.

[0032] Optionally, the step of calculating vegetation growth change data based on the spatiotemporal fusion features using a pre-constructed vegetation growth model, comparing the growth change data with a preset safety threshold, and generating a vegetation growth assessment result based on the comparison result includes:

[0033] Based on the spatiotemporal fusion features, the height growth of vegetation in the vertical direction, the crown growth and density change in the horizontal direction are calculated using a pre-constructed vegetation growth model.

[0034] The height growth, crown width growth, and density change are compared with their respective preset safety thresholds to obtain vertical comparison results, horizontal comparison results, and density change comparison results.

[0035] The vertical comparison results, the horizontal comparison results, and the density change comparison results are comprehensively evaluated and processed to generate vegetation growth assessment results.

[0036] Optionally, the step of performing point cloud stitching processing on the multi-view laser point cloud data from different periods to obtain complete point cloud data, and then segmenting the complete point cloud data to separate the regions containing vegetation points, thereby obtaining vegetation point cloud distribution data, includes:

[0037] Identify the corresponding points in the overlapping areas between laser point cloud data from different viewpoints, and calculate the transformation parameters for converting the laser point cloud data from each viewpoint to a unified coordinate system based on the corresponding points.

[0038] Based on the conversion parameters, the multi-view laser point cloud data from each viewpoint is processed by point cloud stitching to obtain complete point cloud data.

[0039] By using a preset spatial height range and reflection intensity range, each point in the complete point cloud data is segmented to separate the regions containing vegetation points that simultaneously satisfy the spatial height range and reflection intensity range, thus obtaining vegetation point cloud distribution data.

[0040] Secondly, this application provides a safety inspection and identification assessment system for overhead power transmission lines based on laser point clouds, including:

[0041] The acquisition module is used to acquire multi-view laser point cloud data and corresponding meteorological environmental data of the same transmission line corridor area at different times, and extract meteorological features from the meteorological environmental data.

[0042] The separation module is used to perform point cloud stitching processing on the multi-view laser point cloud data of each period to obtain complete point cloud data, and to perform segmentation processing on the complete point cloud data to separate the area containing vegetation points to obtain vegetation point cloud distribution data.

[0043] The reconstruction module is used to perform three-dimensional reconstruction processing based on the vegetation point cloud distribution data and combined with the multi-dimensional indicators of vegetation to obtain a three-dimensional vegetation model.

[0044] The fusion module is used to extract vegetation point cloud features from the vegetation 3D model, and to perform spatiotemporal feature encoding and fusion of the vegetation point cloud features with the meteorological features to obtain spatiotemporal fusion features.

[0045] The comparison module is used to calculate vegetation growth change data based on the spatiotemporal fusion features and a pre-constructed vegetation growth model, compare the growth change data with a preset safety threshold, and generate vegetation growth assessment results based on the comparison results.

[0046] Thirdly, this application provides an electronic device, comprising:

[0047] Memory, used to store computer programs;

[0048] A processor is configured to execute the computer program to implement the steps of the safety inspection, identification, and evaluation method for overhead power transmission lines based on laser point clouds as described in the first aspect above.

[0049] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the safety inspection, identification, and evaluation method for overhead power transmission lines based on laser point clouds as described in the first aspect above.

[0050] The safety inspection and identification assessment method for overhead power transmission lines based on laser point clouds provided in this application involves collecting multi-view laser point cloud data and corresponding meteorological environmental data from different periods in the same power transmission line corridor area, extracting meteorological features from the meteorological environmental data, performing point cloud stitching processing on the multi-view laser point cloud data of each period to obtain complete point cloud data, segmenting the complete point cloud data to separate areas containing vegetation points, obtaining vegetation point cloud distribution data, performing three-dimensional reconstruction processing based on the vegetation point cloud distribution data and combining multi-dimensional vegetation indicators to obtain a three-dimensional vegetation model, extracting vegetation point cloud features from the three-dimensional vegetation model, and performing spatiotemporal feature encoding and fusion of the vegetation point cloud features and meteorological features to obtain spatiotemporal fusion features, calculating vegetation growth change data through a pre-constructed vegetation growth model based on the spatiotemporal fusion features, comparing the growth change data with a preset safety threshold, and generating a vegetation growth assessment result based on the comparison result.

[0051] The technical solution of this application has the following beneficial effects:

[0052] This application provides a complete multi-source data foundation for vegetation growth analysis by acquiring multi-period, multi-view laser point cloud and meteorological environmental data, ensuring the spatiotemporal continuity of observations. Through point cloud stitching and segmentation, it transforms raw discrete point clouds into structured vegetation distribution data, providing clean input data for subsequent 3D reconstruction. Based on the vegetation point cloud distribution data and multi-dimensional indicators, it performs 3D reconstruction to construct a geometric model that accurately reflects the actual spatial structure of vegetation. By spatiotemporally encoding and fusing vegetation point cloud features with meteorological features, it establishes a correlation model between vegetation growth status and environmental factors, enhancing the comprehensiveness of growth assessment. Using the pre-constructed growth model, it calculates growth change data and, through safety threshold comparison, achieves quantitative assessment and risk warning of vegetation growth status.

[0053] Furthermore, this application determines the core growth area by performing point cloud density analysis on vegetation point cloud distribution data, clusters the point clouds in the area into point groups corresponding to independent vegetation based on spatial proximity, calculates the outer contour shape of each point group to form a surface structure, and supplements and optimizes the surface structure details by combining multi-dimensional indicators, and finally generates a high-precision three-dimensional vegetation model.

[0054] This application realizes the automated reconstruction process from discrete point clouds to refined 3D models. It accurately identifies individual vegetation plants through density analysis and spatial clustering, and optimizes model details by combining multi-dimensional indicators, thereby improving the realism and accuracy of the 3D vegetation model and providing a reliable geometric basis for subsequent growth feature extraction.

[0055] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 A flowchart illustrating a safety inspection, identification, and evaluation method for overhead power lines based on laser point clouds, provided for an embodiment of this application;

[0058] Figure 2 A schematic diagram illustrating a specific implementation of a safety inspection, identification, and evaluation method for overhead power transmission lines based on laser point clouds, provided in this application embodiment;

[0059] Figure 3 A schematic diagram of the structure of a safety inspection, identification, and evaluation system for overhead power transmission lines based on laser point clouds, provided for an embodiment of this application;

[0060] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0061] The core problem facing current vegetation monitoring technology along power transmission line corridors is that existing methods struggle to establish precise correlations between multi-period observation data and dynamic changes in vegetation growth. Furthermore, there is a lack of analytical tools to effectively link environmental meteorological factors with the vegetation growth process. This technological limitation leads to judgments about vegetation growth trends based on static observations and empirical inferences, failing to meet the practical needs for accurate prediction and early warning of vegetation growth status.

[0062] To overcome the aforementioned technical limitations, this application proposes a safety inspection and identification assessment method for overhead power transmission lines based on laser point clouds. This method employs the following technical concept: by collecting multi-period, multi-view laser point cloud and meteorological data, vegetation distribution data is obtained through point cloud stitching and segmentation. A three-dimensional vegetation model is generated by combining multi-dimensional indicators, and then point cloud features are extracted and spatiotemporally encoded and fused with meteorological features. Finally, vegetation growth changes are calculated based on the fused features using a growth model and compared with a safety threshold. This scheme, through multi-source data fusion and spatiotemporal feature analysis, achieves accurate monitoring and scientific assessment of vegetation growth dynamics, effectively solving the problems of insufficient dynamic correlation and lack of environmental factor consideration in existing technologies.

[0063] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0064] The core of this application is to provide a safety inspection, identification, and assessment method for overhead power transmission lines based on laser point clouds. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0065] S101. Collect multi-view laser point cloud data and corresponding meteorological environment data of the same transmission line corridor area at different times, and extract meteorological features from the meteorological environment data.

[0066] In the above scheme, multi-view laser point cloud data refers to a point cloud dataset containing three-dimensional coordinates and reflection intensity, obtained by scanning the same power transmission line corridor area from different angles using lidar equipment installed on platforms such as mobile survey vehicles and drones; meteorological environmental data refers to environmental parameters affecting vegetation growth, including temperature, humidity, precipitation, and wind speed; the corresponding meteorological environmental data refers to meteorological data that is spatially consistent with the laser scanning area and temporally synchronized with each point cloud data acquisition period. Meteorological features refer to quantitative characteristic indicators that characterize the environmental conditions for vegetation growth, obtained from the original meteorological data through feature extraction methods.

[0067] In this application example, firstly, a lidar system deployed on a drone and a mobile survey vehicle is used to scan the target power transmission line corridor area from multiple angles, acquiring raw point cloud data containing spatial location information and reflection intensity values. Simultaneously, environmental data such as temperature, humidity, and wind speed are collected using meteorological sensors deployed on-site. Secondly, the collected multi-view lidar point cloud data is preprocessed, establishing a correspondence between data from different perspectives by associating timestamps and spatial coordinates, forming a point cloud dataset that fully covers the target area. Then, time-series analysis is performed on the meteorological environmental data, extracting statistical features such as the mean, extreme values, and variance of each meteorological element using a sliding window statistical method, forming meteorological feature vectors corresponding to the point cloud acquisition period. Finally, the processed point cloud data and meteorological features are associated with time labels to establish a multi-source dataset with spatiotemporal consistency.

[0068] In practical applications, in a power transmission line corridor monitoring project, a UAV lidar system was first used to collect data on a 50-kilometer corridor area in spring and autumn. The 120 million point cloud data collected in spring were obtained by setting the flight altitude to 80 meters and the scanning frequency to 100 Hz, and scanning continuously for 4 hours. The specific calculation is 100 Hz × 3600 seconds × 4 hours × 10 scan lines = 144 million points. After noise reduction, 120 million valid points were retained. The 150 million point cloud data collected in autumn were obtained by scanning for 5 hours under the same parameter settings. The calculation process is 100 Hz × 3600 seconds × 5 hours × 10 scan lines = 180 million points. After noise reduction, 150 million valid points were retained.

[0069] Subsequently, raw meteorological data were collected from four meteorological stations along the route. The temperature data was calculated by taking the arithmetic mean of 288 sampled values, using the following formula: ,in Let represent the i-th temperature sample value, and N represent the total number of sampling points, 288. Substituting the 288 temperature values ​​collected in spring into the formula yields 18.6℃, and substituting the 288 temperature values ​​collected in autumn yields 25.3℃. Humidity data is calculated using the same method, substituting the 288 humidity sample values ​​into the formula... The calculated average humidity for spring is 65.2% and for autumn is 58.7%. This represents the i-th humidity sample value, and N represents the total number of sampling points, 288. The cumulative precipitation is calculated by substituting the 24 hourly precipitation samples into the formula. ,in, This indicates the cumulative precipitation. Let M represent the precipitation in hour j, and M represent the hour number 24. The cumulative precipitation in spring is calculated to be 45.6 mm and the cumulative precipitation in autumn is 32.1 mm.

[0070] Finally, the meteorological feature data obtained through the complete calculation process are associated with the point cloud data of the corresponding period through precise timestamps to form a dataset with strict spatiotemporal consistency, which is then directly sent to the subsequent point cloud stitching and segmentation processing flow.

[0071] The above-mentioned S101 overall solution, through a systematic data acquisition and processing process, establishes a spatiotemporally aligned multi-source dataset, providing a complete data foundation for vegetation growth assessment, ensuring the spatiotemporal consistency and data reliability required for subsequent analysis, and solving the limitation of a single data source in vegetation growth assessment.

[0072] S102. Perform point cloud stitching processing on the multi-view laser point cloud data of each period to obtain complete point cloud data. Perform segmentation processing on the complete point cloud data to separate the areas containing vegetation points to obtain vegetation point cloud distribution data.

[0073] Optionally, step S102 involves performing point cloud stitching processing on the multi-view laser point cloud data from each period to obtain complete point cloud data, and then segmenting the complete point cloud data to separate areas containing vegetation points, thereby obtaining vegetation point cloud distribution data, including:

[0074] Step 1021: Identify the corresponding points in the overlapping areas between laser point cloud data from different viewpoints, and calculate the transformation parameters to convert the laser point cloud data from each viewpoint to a unified coordinate system based on the corresponding points.

[0075] Step 1022: According to the conversion parameters, perform point cloud stitching processing on the multi-view laser point cloud data from each viewpoint to obtain complete point cloud data.

[0076] Step 1023: Segment each point in the complete point cloud data using a preset spatial height range and reflection intensity range to separate the regions containing vegetation points that simultaneously satisfy the spatial height range and reflection intensity range, thereby obtaining vegetation point cloud distribution data.

[0077] In the above scheme, corresponding points refer to feature points with the same spatial location in the overlapping area of ​​point cloud data from different viewpoints; transformation parameters refer to rotation matrices and translation vectors used to transform point cloud data from different coordinate systems to a unified coordinate system; spatial height range refers to the effective range in the vertical direction set according to the safety distance requirements of power transmission lines; and reflection intensity range refers to the range of laser reflection intensity values ​​unique to vegetation point clouds.

[0078] In this application example, firstly, common feature points in the overlapping areas of point cloud data acquired from different scanning perspectives are identified using feature extraction technology. These feature points have the same spatial characteristics in different coordinate systems. Based on the spatial correspondence of these feature points, a transformation model between different coordinate systems is established, and complete transformation parameters, including rotation angle and translation distance, are calculated. Secondly, the calculated transformation parameters are used to perform coordinate system unification processing on the point cloud datasets from each perspective, transforming the point cloud data originally scattered in different coordinate systems into a unified global coordinate system. During this process, overlapping area optimization processing is performed on the transformed point cloud data to eliminate duplicate points and balance the point cloud density, forming a point cloud dataset that completely covers the monitoring area. Finally, a reasonable spatial height range is set according to the safety operation specifications of the transmission line corridor, and the reflection intensity range is determined by combining the reflection characteristics of vegetation to laser light. Based on these two conditions, the complete point cloud dataset is screened point by point to extract the point set that simultaneously meets the height and reflection intensity conditions, forming accurate and reliable vegetation point cloud distribution data.

[0079] In a practical application, specifically in a power transmission line corridor monitoring project, the point cloud data acquired from three different scanning perspectives were first processed. A feature extraction algorithm identified 156 sets of corresponding feature points from the overlapping area, and coordinate transformation parameters were calculated based on these points. The rotation matrix R was calculated using singular value decomposition, with the specific formula as follows: Where U and V are orthogonal matrices obtained by decomposing the covariance matrix of the corresponding point sets. The formula for calculating the translation vector T is: ,in and These represent the centroid coordinates of the two point clouds, obtained by calculating the average coordinates of all points in each point cloud.

[0080] Subsequently, the point cloud data from the three perspectives were unified using these transformation parameters to form a complete point cloud dataset containing 28.5 million points. Then, the spatial height range was set to 0.5 meters to 30 meters above the ground, and the reflection intensity range was set to 0.15 to 0.85. Point-by-point conditional checks were performed on the complete point cloud dataset, selecting point sets that simultaneously met both conditions, ultimately yielding vegetation point cloud distribution data containing 6.3 million points. This vegetation point cloud distribution data was directly fed into the subsequent 3D reconstruction processing flow to construct a 3D vegetation model.

[0081] The S102 overall solution described above establishes a transformation process from multi-view discrete point clouds to complete vegetation distribution data through systematic point cloud stitching and segmentation. This effectively solves the inconsistency problem in spatial coordinate systems and density distribution of multi-source point cloud data. By accurately calculating coordinate transformation parameters and implementing strict point cloud selection criteria, the accuracy of vegetation point cloud data in terms of spatial location and feature attributes is ensured. This provides a high-quality data foundation for subsequent 3D vegetation reconstruction and growth feature extraction, improving the reliability and accuracy of vegetation growth assessment.

[0082] S103. Based on the vegetation point cloud distribution data and combined with the multi-dimensional indicators of vegetation, perform three-dimensional reconstruction processing to obtain a three-dimensional vegetation model.

[0083] Optionally, step S103 involves performing three-dimensional reconstruction processing based on the vegetation point cloud distribution data and combining it with multi-dimensional vegetation indicators to obtain a three-dimensional vegetation model, including:

[0084] Step 1031: Perform point cloud density analysis on the vegetation point cloud distribution data to obtain the core growth area with a point cloud density greater than the first preset density threshold.

[0085] Step 1032: Based on spatial proximity, the point cloud in the core growth area is clustered to obtain multiple point groups, each point group corresponding to an independent vegetation.

[0086] Step 1033: For each point group, calculate the outer contour shape of the point group to obtain the surface structure that reflects the external morphology of the vegetation. Combine the multi-dimensional indicators of the vegetation to supplement and optimize the detailed features of the surface structure to obtain a three-dimensional model of the vegetation.

[0087] Step 1033 may specifically include the following steps: determining the overall height of vegetation based on the vertical distribution range of the point cloud, and dividing the point cloud into several layers in the vertical direction according to the overall height; calculating the point cloud density within each layer, identifying the layer with a point cloud density greater than a second preset density threshold and the largest horizontal coverage as the main canopy area of ​​the vegetation; dividing the point cloud of the main canopy area into multiple sub-regions according to the local density gradient; for each sub-region, estimating the sub-volume corresponding to the sub-region based on the number of point clouds in the sub-region, adaptively adjusting the parameters of the surface reconstruction algorithm based on the point cloud density and sub-volume of the sub-region, and performing the sub-region reconstruction algorithm based on the adjusted parameters. The domain is reconstructed in 3D to generate sub-models corresponding to sub-regions. The sum of the sub-volumes corresponding to all sub-regions is the theoretical total volume. The theoretical total volume, overall height, and point cloud density within the layer are used to constitute the multi-dimensional indicators of vegetation. The sub-models corresponding to all sub-regions are fused, and the total volume of the fused model is calculated. When the error between the total volume of the fused model and the theoretical total volume is greater than a preset error, the total volume of the fused model is corrected by scaling transformation until the error between the total volume of the fused model and the theoretical total volume is less than or equal to the preset error. The overall height remains unchanged during the correction process, and finally, a 3D vegetation model is obtained.

[0088] In the above scheme, the core growth area refers to the concentrated distribution area where the point cloud density in the vegetation point cloud distribution data is higher than that of the surrounding area; the point cluster refers to the set of independent point clouds with continuous spatial distribution characteristics obtained by spatial clustering analysis; the surface structure refers to the three-dimensional structure reflecting the external geometric morphology of vegetation obtained by calculating the outer contour of the point cluster; and the multi-dimensional index refers to multi-angle quantitative parameters reflecting the vegetation growth status, including vegetation height, canopy volume, and point cloud density distribution.

[0089] In this application example, firstly, a comprehensive density statistical analysis is performed on the vegetation point cloud distribution data. By calculating the number of point clouds contained in each unit volume space, concentrated distribution areas with point cloud density higher than the surrounding areas are identified. These areas are defined as core growth areas, and their density values ​​must exceed a preset first density threshold.

[0090] Secondly, based on the three-dimensional spatial distance relationship between point clouds, cluster analysis is used to group the point clouds in the core growth area, dividing the spatially adjacent and density-distributed point cloud sets into independent point clusters, with each point cluster corresponding to a complete vegetation individual.

[0091] Then, for each independent point group, a preliminary surface structure model is generated by calculating its spatial outer contour shape, reflecting the basic geometric morphology of the vegetation. Based on this, multi-dimensional vegetation indicators, including height distribution, volume characteristics, and density variations, are used to optimize the surface structure and supplement its features. Furthermore, the overall height of the vegetation is determined based on the vertical distribution range of the point cloud, and the point cloud is divided into multiple layers vertically using either equal spacing or an adaptive method. The point cloud density distribution characteristics of each layer are analyzed, and the layer with the highest density and largest horizontal coverage is identified as the main canopy region. The point cloud of the main canopy region is divided into multiple sub-regions based on local density gradient changes, and its sub-volume is estimated based on the number of points and related spatial characteristics of each sub-region. Based on the point cloud density and estimated volume of the sub-regions, the parameter settings of the surface reconstruction algorithm are dynamically adjusted to perform refined 3D modeling for each sub-region. Subsequently, the 3D models of all sub-regions are fused, the total volume of the fused model is calculated, and compared with the theoretical total volume. When the error between the two exceeds the preset range, volume correction is performed through scaling transformation while keeping the overall height of the vegetation unchanged, thus obtaining an accurate 3D model of the vegetation.

[0092] In a practical application, a vegetation monitoring project along a power transmission line corridor involved 3D reconstruction of the vegetation point cloud distribution data obtained from preliminary processing. This data contained 6.3 million points. First, point cloud density analysis was performed, setting a first density threshold of 50 points per cubic meter. By calculating the number of points within each 1-cubic-meter grid, the core growth region was identified as containing 2.8 million points, all exceeding the threshold density. Next, a clustering algorithm based on Euclidean distance was applied, with a cluster radius of 0.5 meters, dividing the point cloud within the core growth region into 85 independent point clusters.

[0093] Subsequently, taking a typical point cluster as an example, which contains 32,500 points, its overall height was calculated. By finding the maximum and minimum values ​​of the Z-coordinate in the point cloud, the height was determined to be 8.6 meters. The point cloud was then vertically divided into 9 levels, with each level representing one meter, and the point cloud density of each level was calculated using the following formula: ,in, The density is represented by n, the number of points in the layer is represented by V, and the layer volume is represented by V. It was found that the density of the 4th layer reached 120 points per cubic meter and had the largest horizontal coverage, so it was identified as the main area of ​​the canopy.

[0094] Subsequently, the point cloud of the main canopy region was divided into six sub-regions based on local density gradients, with each sub-region containing 4800, 5200, 4500, 5100, 4900, and 4600 points, respectively. The volume of each sub-region was estimated based on the number of point clouds, using the following formula: ,in, This represents the estimated volume, where n represents the number of point clouds. The average point cloud density is 100 points per cubic meter. The estimated volumes of each sub-region are 2.4 cubic meters, 2.6 cubic meters, 2.25 cubic meters, 2.55 cubic meters, 2.45 cubic meters and 2.3 cubic meters, respectively, and the theoretical total volume is the sum of the volumes of all sub-regions, which is 15.15 cubic meters.

[0095] Finally, the parameters of the surface reconstruction algorithm were adjusted based on the point cloud density and estimated volume of each sub-region, primarily adjusting the surface smoothness and detail retention parameters. Three-dimensional reconstruction was performed on each sub-region to generate a sub-model. After merging all sub-models, a preliminary three-dimensional model was obtained, with a calculated total volume of 15.6 cubic meters. This volume was compared with the theoretical total volume of 15.15 cubic meters, showing an error of 2.97%, exceeding the preset error range of 2%. Therefore, volume correction was performed, adjusting the model volume to 15.15 cubic meters through scaling transformation, while maintaining the overall height of 8.6 meters, ultimately obtaining an accurate three-dimensional vegetation model. This model will be directly used in subsequent vegetation point cloud feature extraction steps, providing fundamental data for growth change analysis.

[0096] The aforementioned S103 overall solution establishes a complete transformation chain from discrete point cloud data to accurate 3D vegetation models through a systematic 3D reconstruction processing workflow. Multi-level density analysis and spatial clustering ensure accurate identification of individual vegetation plants. The model reconstruction process is optimized using multi-dimensional indicators, and regional modeling and volume correction mechanisms improve the accuracy of the model in terms of geometry and spatial structure. This processing provides a reliable 3D data foundation for subsequent vegetation growth feature extraction and spatiotemporal change analysis, enhancing the scientific rigor and accuracy of vegetation growth status assessment, and providing technical support for the safety management of vegetation along transmission line corridors.

[0097] S104. Extract vegetation point cloud features from the vegetation 3D model, and perform spatiotemporal feature encoding and fusion of the vegetation point cloud features and the meteorological features to obtain spatiotemporal fusion features.

[0098] Optionally, step S104, extracting vegetation point cloud features from the 3D vegetation model and fusing the vegetation point cloud features with the meteorological features through spatiotemporal feature encoding to obtain spatiotemporal fused features, includes:

[0099] Step 1041: Calculate the vertical dimension of the vegetation 3D model to obtain the height feature; calculate the horizontal coverage of the vegetation 3D model to obtain the crown width feature; and calculate the spatial size of the vegetation 3D model to obtain the volume feature.

[0100] Both the height feature and the overall height originate from the vertical dimension measurement of the same 3D vegetation model and share the same numerical basis. The difference lies in their application scenarios and purposes: the overall height is an intermediate geometric parameter in the 3D reconstruction process, primarily used for structural division and volume correction reference during model building. The height feature, on the other hand, is a quantitative indicator extracted from the final 3D model for growth assessment, serving as direct input data for spatiotemporal feature fusion and reflecting the transformation relationship from model building parameters to growth assessment features. Both the volume feature and the total volume of the fused model characterize the spatial occupancy of the 3D vegetation model and are numerically consistent. The difference lies in their technical roles and output stages: the total volume of the fused model is the final output of the 3D reconstruction process, used to verify reconstruction accuracy and complete model optimization; it is an internal quality indicator of the modeling process. The volume feature, however, is a feature parameter extracted from the validated model for growth analysis, serving as the foundational data for subsequent spatiotemporal fusion and growth assessment, reflecting the process connection from modeling output to feature application.

[0101] Step 1042: Arrange the height feature, the crown width feature, and the volume feature in chronological order to obtain vegetation point cloud features.

[0102] Step 1043: Using a spatiotemporal feature encoding method, the vegetation point cloud features and corresponding meteorological features of the same period are processed to obtain spatiotemporal fusion features.

[0103] Step 1043 may specifically include the following steps: combining the vegetation point cloud features and meteorological features from the same period to obtain multiple fused feature units; arranging all the fused feature units in chronological order to obtain time series features; analyzing the fused feature units from adjacent periods in the time series features to obtain feature change trend information; and integrating the time series features and the feature change trend information through a spatiotemporal feature encoding method to obtain spatiotemporal fused features. The specific implementation process of the spatiotemporal feature encoding method is as follows: first, concatenating the feature vector of each period in the time series features with the corresponding feature change trend information vector to form an enhanced feature representation; then, capturing the evolutionary patterns and dependencies of features in the time dimension through a time series analysis module, focusing on extracting the feature contributions of key time nodes; simultaneously, combining spatial correlation analysis to evaluate the spatial interaction of different feature dimensions; finally, using feature compression technology to reduce and optimize the dimensionality of the concatenated high-dimensional features, eliminating redundant information and retaining the most discriminative feature combinations, thereby generating a unified and compact spatiotemporal fused feature vector that comprehensively reflects the historical state, dynamic change trends, and environmental factors of vegetation growth.

[0104] In the above scheme, height feature refers to the maximum size value of the vegetation 3D model in the vertical direction; crown width feature refers to the maximum projection coverage of the vegetation 3D model on the horizontal plane; volume feature refers to the size of the space occupied by the vegetation 3D model; vegetation point cloud feature refers to the feature sequence formed by arranging height feature, crown width feature and volume feature in chronological order; spatiotemporal feature encoding refers to the technical method of fusing vegetation point cloud feature and meteorological feature in time and space dimensions.

[0105] In this application example, firstly, key geometric features are extracted from the three-dimensional vegetation model of each period, including height features obtained by measuring the maximum distance of the model in the vertical direction, crown features obtained by calculating the projected area of ​​the model on the horizontal plane, and volume features obtained by evaluating the spatial volume occupied by the model. These features together describe the static spatial properties of the vegetation.

[0106] Secondly, the height, crown width, and volume features extracted at different times are arranged and organized according to the collection time sequence to form a temporal feature sequence reflecting the vegetation growth process, namely vegetation point cloud features. This sequence can intuitively show the state evolution of vegetation at multiple time points.

[0107] Then, the vegetation point cloud features from the same period are combined with the corresponding meteorological features to construct fused feature units, ensuring that each unit contains both the geometric state of the vegetation and environmental impact information, providing a multi-source data foundation for subsequent analysis.

[0108] Next, the fusion feature units of all periods are arranged in chronological order to construct a complete time series feature. By analyzing the numerical differences between fusion feature units of adjacent periods, the feature change trend information of vegetation growth process is extracted, thereby capturing the dynamic growth law.

[0109] Finally, a spatiotemporal feature encoding method is used to integrate time series features with feature change trend information to generate a spatiotemporal fusion feature that contains both static features and dynamic changes. This feature comprehensively reflects the growth status of vegetation in the spatiotemporal dimension.

[0110] In practical applications, such as a vegetation monitoring project along a power transmission line corridor, feature extraction was first performed based on three-dimensional vegetation models from three different periods. For height features, the maximum vertical distance of the model in each period was calculated using the formula... Where h represents the height feature and z represents the vertical coordinate value of the model point, the maximum and minimum values ​​of the vertical coordinates of all points are found and the difference is calculated, resulting in height feature values ​​of 5.2 meters, 5.8 meters, and 6.3 meters for the three periods. For the crown feature, the model is projected onto a horizontal plane and the area of ​​the projected region is calculated using the formula... Where 'a' represents the crown width feature and 'S' represents the projected area, the crown width feature values ​​for the three periods are obtained as 12.6 square meters, 14.2 square meters, and 15.8 square meters by meshing the projected surface and calculating the effective mesh area. For the volume feature, the space occupied by the model is calculated using a voxelization method, employing the formula... Where v represents the volume feature and n represents the number of voxels. The volume of a single voxel is represented by 0.001 cubic meters. By counting the number of voxels within the model and multiplying them, the volume characteristic values ​​for the three periods are 28.5 cubic meters, 33.7 cubic meters, and 38.9 cubic meters, respectively.

[0111] Subsequently, these features were arranged chronologically to form a vegetation point cloud feature sequence, while meteorological features for the corresponding periods, including average temperature, precipitation, and humidity, were acquired. Vegetation point cloud features and meteorological features from the same period were combined to form fused feature units; three fused feature units were obtained for the three periods. These fused feature units were arranged chronologically to construct a time-series feature, and the changes in feature values ​​between adjacent periods were analyzed. The height increase was calculated using a formula. ,in, Indicates a high growth rate. The height characteristics of period t are represented by the height increase of 0.6 meters and 0.5 meters respectively. Similarly, the crown width increase and volume increase are calculated using the same method, and the crown width increase is 1.6 square meters and 1.6 square meters respectively, and the volume increase is 5.2 cubic meters and 5.2 cubic meters respectively. These increases together constitute the characteristic change trend information.

[0112] Finally, the time series features and feature change trend information are integrated through spatiotemporal feature encoding to generate spatiotemporal fusion features. These features will be directly used in subsequent vegetation growth model calculations as input data for growth change analysis.

[0113] The aforementioned S104 overall scheme, through systematic feature extraction and fusion processing, achieves a complete transformation from a three-dimensional vegetation model to multi-dimensional features, establishes a deep spatiotemporal correlation between vegetation geometric features and meteorological environmental features, and comprehensively reveals the key laws in the vegetation growth process through refined time series analysis and dynamic change trend capture. The generated spatiotemporal fusion features provide a high-precision, multi-dimensional data foundation for vegetation growth assessment, improve the analytical accuracy and predictive reliability of subsequent growth models, and provide solid technical support for the vegetation safety management and scientific decision-making of transmission line corridors.

[0114] S105. Based on the spatiotemporal fusion features, calculate the vegetation growth change data through a pre-constructed vegetation growth model, compare the growth change data with a preset safety threshold, and generate a vegetation growth assessment result based on the comparison result.

[0115] Optionally, step S105 involves calculating vegetation growth change data based on the spatiotemporal fusion features using a pre-constructed vegetation growth model, comparing the growth change data with a preset safety threshold, and generating a vegetation growth assessment result based on the comparison result, including:

[0116] Step 1051: Using a pre-constructed vegetation growth model, calculate the vertical height growth, horizontal crown growth, and density change of vegetation based on the spatiotemporal fusion features. The vegetation growth model employs a segmented ensemble model structure, including a feature preprocessing module, a spatiotemporal feature parsing module, and a multi-output prediction module. The feature preprocessing module is responsible for standardizing and dimensionality-aligning the input spatiotemporal fusion features. The spatiotemporal feature parsing module uses a multilayer perceptron structure, including an input layer, three hidden layers, and an output layer, extracting deep features through a fully connected network. The multi-output prediction module uses a parallel computing structure, corresponding to three output branches for height growth, crown growth, and density change. Each branch contains a dedicated fully connected layer and activation function, ultimately outputting three predicted values ​​simultaneously. The model calculates the three growth values ​​through a forward propagation mechanism: first, the spatiotemporal fusion features are input into a shared feature extraction layer to obtain latent feature representations; then, they are calculated separately through three independent branches, with the height growth branch performing a linear transformation. ,in, This is the weight matrix. As a bias term, the crown growth branch is calculated using the ReLU activation function. The density change branch constrains the output range using the Sigmoid function. The three branches share a feature extraction layer but optimize independently, ultimately outputting a quantitative indicator of vegetation growth status synchronously. The model construction adopts a modular design process: first, the shared feature extraction layer is initialized, setting the number of hidden layer neurons in the sequence 256-128-64, and weights are allocated using the He initialization method; then, three parallel output layers are constructed, configuring dedicated activation functions and regularization coefficients according to the characteristics of each growth level; residual structures and batch normalization layers are introduced into the network connections to enhance gradient flow and training stability; finally, all modules are integrated to form a multi-task learning framework, using the Adam optimizer and a multi-objective loss function to achieve end-to-end parameter optimization; the model training adopts a multi-stage optimization strategy: first, a training set containing historical vegetation features and meteorological data is constructed, dividing the training / validation / test sets in a 7:2:1 ratio; during the training phase, a composite loss function is used. ,in , , These represent the mean squared error losses for the three growth rates. , , To balance the weights, parameters are updated iteratively through backpropagation. The initial learning rate is set to 0.001 and dynamically adjusted using cosine annealing. An early stopping mechanism is used to prevent overfitting. Finally, the model with the best performance on the validation set is selected as the deployment version.

[0117] Step 1052: Compare the height growth, crown width growth, and density change with the corresponding preset safety thresholds to obtain the vertical comparison result, the horizontal comparison result, and the density change comparison result.

[0118] Step 1053: Perform comprehensive judgment and processing on the vertical comparison results, the horizontal comparison results, and the density change comparison results to generate vegetation growth assessment results. When comprehensively judging the vertical, horizontal, and density change comparison results, a decision-making rule based on a three-level early warning mechanism is first established: when all three comparison results show no exceedances, a "safe" assessment result is directly generated; when any comparison result exceeds its safety threshold but does not reach the emergency threshold, a "caution" level assessment result is initiated, and the specific growth parameter exceeding the limit is marked; when two or more comparison results exceed the limit, or any direction exceeds the emergency threshold, a "danger" assessment result is generated, and the degree of numerical deviation of the exceeding parameters and the combined risk pattern are recorded simultaneously. In this process, a multi-dimensional feature coupling analysis method is used to focus on analyzing the correlation between different growth parameters. For example, when height growth and crown growth both show an accelerating trend, even if each parameter individually does not reach the danger threshold, the "caution" level will be triggered and an abnormal growth trend will be indicated. Finally, through this multi-level, multi-condition comprehensive judgment mechanism, a vegetation growth assessment result with a clear risk level and specific early warning direction is generated.

[0119] In the above scheme, the vegetation growth model refers to a computational model trained by machine learning methods that can predict vegetation growth changes based on spatiotemporal fusion features; height growth refers to the degree of vegetation growth change in the vertical direction; crown growth refers to the degree of vegetation expansion change in the horizontal direction; density change refers to the change in vegetation point cloud density per unit volume; and safety threshold refers to the allowable limit values ​​of various growth parameters set according to the safety operation specifications of transmission lines.

[0120] In this application example, the spatiotemporal fusion features obtained after preprocessing are first input into a pre-trained vegetation growth model. The model generates the vertical height growth, horizontal crown growth, and spatial density change of vegetation through deep analysis and calculation of the input features. These three key indicators together constitute a dataset reflecting the changes in vegetation growth status.

[0121] Secondly, the height growth calculated by the model is compared in detail with the preset vertical safety threshold, and the safety status assessment result in the vertical direction is obtained based on the comparison. At the same time, the crown growth is compared with the horizontal safety threshold to obtain the safety status assessment result in the horizontal direction. And the density change is compared with the density safety threshold to obtain the safety status assessment result in terms of density change.

[0122] Finally, based on the comparison results of the three dimensions, a multi-factor comprehensive analysis method was adopted to comprehensively consider the degree of exceeding the limit and the combined effect of various indicators, and a final assessment result that can accurately reflect the impact of vegetation growth status on the safe operation of transmission lines was generated, providing a scientific basis for subsequent vegetation management decisions.

[0123] In practical applications, a vegetation monitoring project along a power transmission line corridor used spatiotemporal fusion features obtained from preprocessing as input data, and performed calculations and analyses using a pre-constructed vegetation growth model. This model was trained on a large amount of historical data, and a multivariate regression analysis method was used to establish the relationship between features and growth.

[0124] First, the model calculates the height growth using the formula. ,in, Let β represent height growth, β represent model coefficients, and F represent the various feature components in the spatiotemporal fusion feature. By substituting the specific feature values ​​into the formula, the height growth for the current monitoring period is calculated to be 0.45 meters; then, the crown growth is calculated using the formula... ,in, This represents the crown width growth, and γ represents a specific coefficient of the crown width growth model. The calculated crown width growth for the current cycle is 1.2 square meters. Finally, the density change is calculated using the formula... ΔD represents the density change, and δ represents a specific coefficient of the density change model. The calculation results show that the density change in the current cycle is an increase of 15 points per cubic meter.

[0125] Next, the calculated growth change data was systematically compared with preset safety thresholds. According to the transmission line operation safety regulations, the vertical safety threshold was set at 0.5 meters. The calculated height growth of 0.45 meters was compared with this threshold; since 0.45 is less than 0.5, the vertical comparison result was deemed safe. The horizontal safety threshold was set at 1.5 square meters. The crown growth of 1.2 square meters was compared with this threshold, and the horizontal comparison result was deemed safe. The density safety threshold was set at 20 points per cubic meter. The density change of 15 points was compared with this threshold, and the density change comparison result was deemed safe.

[0126] Finally, based on the comparison results from the three directions, a comprehensive judgment was made. Since all growth parameters were within the safe threshold range, the vegetation growth assessment result was determined to be in a safe state. This assessment result will be directly used to guide the formulation of vegetation maintenance work plans for the transmission line corridor.

[0127] The aforementioned S105 overall solution, through the establishment of a scientific vegetation growth model and a safety threshold comparison mechanism, achieves precise quantitative assessment of vegetation growth status, forming a complete technical chain from multi-source characteristic data to safety decision-making. This process not only enables timely detection of abnormal vegetation growth but also provides specific directional guidance, enhancing the scientific rigor and predictability of vegetation management along transmission line corridors. It provides reliable technical support for ensuring the safe and stable operation of power facilities and offers a basis for optimizing the allocation of vegetation maintenance resources.

[0128] The following is a complete example for steps S101 to S105, such as Figure 2 As shown, in the B-section transmission line corridor monitoring project managed by the power grid company of Province A, firstly, a UAV lidar system was used to collect data in spring and autumn for a 50-kilometer-long corridor area. The flight altitude was set to 80 meters, and the scanning frequency was set to 100 Hz. The total amount of point cloud data obtained in spring was calculated by multiplying the scanning frequency by the scanning time. A scanning frequency of 100 Hz means 100 scans per second. Scanning for 4 hours yielded 144 million point cloud points (100 x 3600 x 4 hours x 10 scan lines). After noise reduction, 120 million points were retained. In autumn, under the same parameters, scanning for 5 hours yielded 180 million point cloud points (100 x 3600 x 5 hours x 10 scan lines). After noise reduction, 150 million points were retained. Simultaneously, temperature, humidity, wind speed, and precipitation data were collected from four meteorological stations along the line. When extracting features from the meteorological data, the formula for calculating the average temperature was... ,in, Indicates average temperature. This represents the i-th temperature sample value, and N represents the number of sampling points, where N equals 288. The calculated average temperature for spring is 18.6 degrees Celsius, and the average temperature for autumn is 25.3 degrees Celsius. The formula for calculating average humidity is similar. ,in, Indicates average humidity. Let represent the i-th humidity sample value. The calculated average humidity for spring is 65.2%, and for autumn it is 58.7%. The formula for calculating cumulative precipitation is... ,in, This indicates the cumulative precipitation. Let M represent the precipitation in hour j, and M represent the number of hours, where M equals 24. The calculated cumulative precipitation for spring is 45.6 mm, and for autumn it is 32.1 mm. These meteorological characteristic data are correlated with the point cloud data for the corresponding period through timestamps, forming a dataset with spatiotemporal consistency, which is used for subsequent point cloud stitching and segmentation processing.

[0129] Next, based on the point cloud data and meteorological features obtained in embodiment S101, point cloud stitching processing is performed on the multi-view laser point cloud data from spring and autumn. First, feature points in the overlapping areas of point cloud data from different views are identified, and coordinate transformation parameters are calculated. The rotation matrix is ​​calculated through singular value decomposition of the covariance matrix of the feature point set, and the translation vector is calculated using the following formula: Where t represents the translation vector, and Let R represent the centroid coordinates of the two point cloud sets, and let R represent the rotation matrix. After unifying the coordinates of the point clouds from different viewpoints according to the transformation parameters, the spring point cloud data was stitched together to form a complete point cloud dataset containing 68.5 million points, and the autumn point cloud data was stitched together to form a complete point cloud dataset containing 72.6 million points. Subsequent segmentation was performed, setting the spatial height range to 0.5 meters to 25 meters above the ground and the reflection intensity range to 0.2 to 0.8. Each point in the complete point cloud data was then filtered. 15.2 million vegetation points were selected from the spring data to form a vegetation point cloud distribution dataset, and 16.8 million vegetation points were selected from the autumn data to form a vegetation point cloud distribution dataset. These data were used for subsequent 3D reconstruction processing.

[0130] Subsequently, based on the vegetation point cloud distribution data obtained in embodiment S102, three-dimensional reconstruction processing was performed. First, point cloud density analysis was conducted, setting a density threshold of 50 points per cubic meter. By calculating the number of points per unit volume, it was identified that the core growth area in spring contained 8.6 million points, and the core growth area in autumn contained 9.2 million points. Next, a spatial clustering algorithm was used to divide the point cloud within the core growth area into independent point clusters, with a cluster radius set to 0.5 meters. 235 point clusters were formed in spring, and 248 point clusters were formed in autumn. Taking one typical point cluster as an example, this cluster contained 32,500 points in spring. By calculating the maximum vertical height difference of the point cloud, the overall height was found to be 7.8 meters. The calculation formula is as follows: Where h represents the overall height and z represents the vertical coordinate of the point. The point cloud is vertically divided into 8 levels, with each level representing one meter. The point cloud density of each level is calculated using the following formula: ,in, Let represent the point cloud density, n represent the number of points in that layer, and V represent the layer volume. After determining that the 4th layer is the main canopy region, the point cloud in this region is divided into 5 sub-regions based on the local density gradient. The volume of each sub-region is estimated using the following formula: ,in, This represents the estimated volume, where n represents the number of point clouds in the sub-region. This represents the average point cloud density, with a value of 100 points per cubic meter. The total volume of the fused model is 12.6 cubic meters. The error between this volume and the theoretical total volume of 12.3 cubic meters is corrected by scaling transformation to obtain the final 3D vegetation model, which is used for subsequent feature extraction.

[0131] Next, vegetation point cloud features were extracted from the 3D vegetation model obtained in embodiment S103. Height features were calculated by measuring the maximum vertical distance of the model; the average height was 7.2 meters in spring and 7.8 meters in autumn. Crown width features were calculated using the horizontal projected area of ​​the model; the average crown width was 10.5 square meters in spring and 11.8 square meters in autumn. Volume features were calculated using a voxelization method, with the formula as follows: ,in, Indicates volume characteristics, Indicates the number of voxels. The volume of a single feature is represented as 0.001 cubic meters. The average volume in spring is 25.3 cubic meters, and the average volume in autumn is 28.7 cubic meters. These features are arranged in chronological order to form a vegetation point cloud feature sequence. This sequence is then fused with the meteorological features obtained in embodiment S101 using spatiotemporal coding. The fusion feature unit is obtained by combining vegetation point cloud features and meteorological features from the same period. The time series features are obtained by arranging the fusion feature units in chronological order. Feature change trend information is obtained by calculating the difference in feature values ​​between adjacent periods. For example, the formula for calculating height growth is... ,in, Indicates height growth. The height characteristics of period t are represented, and the spatiotemporal fusion features are finally generated for subsequent growth model calculations.

[0132] Finally, based on the spatiotemporal fusion feature vector X obtained in embodiment S104, vegetation growth change data are calculated using a pre-constructed vegetation growth model. This model employs a multiple regression algorithm, where the formula for calculating height growth is: ,in, Indicates height growth. The weight matrix represents the high-level branch, and X represents the input feature vector. This represents the bias term; in the specific calculation, it is assumed that the eigenvector X contains three eigenvalues. Weight matrix Bias term =0.1, then The height growth was calculated to be 0.6 meters after model optimization and rounding; the formula for calculating crown growth was... ,in, Indicates crown growth. This represents the weight matrix of the crown branch. Represents the bias term; the ReLU function indicates that it takes a positive value and is zero otherwise. Assume... If , = 0.2, then the linear output is... =1.11, after applying the ReLU function =ReLU(1.11)=1.11 square meters, and after model adjustment, the crown growth is obtained as 1.3 square meters; the formula for calculating the density change is... ,in, Indicates the change in density. The weight matrix represents the density branch. Indicates the bias term. This represents the Sigmoid function, and its calculation formula is: , assuming , =0.3, then =σ(1.44)≈0.809. Since the density change needs to be converted to the actual point cloud density unit, the density change obtained after model scaling is 18 points per cubic meter. Finally, the calculated growth change data is compared with the preset safety threshold. The vertical safety threshold is 0.8 meters, the horizontal safety threshold is 1.5 square meters, and the density safety threshold is 25 points per cubic meter. Since the height growth of 0.6 meters is less than the vertical safety threshold of 0.8 meters, the crown growth of 1.3 square meters is less than the horizontal safety threshold of 1.5 square meters, and the density change of 18 points / cubic meter is less than the density safety threshold of 25 points / cubic meter, all growth change data do not exceed the safety threshold. Therefore, the vegetation growth assessment result is judged to be in a safe state.

[0133] Figure 3 This application provides a schematic diagram of a specific implementation of a safety inspection, identification, and assessment system for overhead power lines based on laser point clouds, as illustrated in the embodiments of this application. Figure 3 The system may include:

[0134] The acquisition module 31 is used to acquire multi-view laser point cloud data and corresponding meteorological environment data of the same transmission line corridor area at different times, and extract meteorological features from the meteorological environment data.

[0135] The separation module 32 is used to perform point cloud stitching processing on the multi-view laser point cloud data of each period to obtain complete point cloud data, and to perform segmentation processing on the complete point cloud data to separate the area containing vegetation points to obtain vegetation point cloud distribution data.

[0136] The reconstruction module 33 is used to perform three-dimensional reconstruction processing based on the vegetation point cloud distribution data and combined with the multi-dimensional indicators of vegetation to obtain a three-dimensional vegetation model.

[0137] The fusion module 34 is used to extract vegetation point cloud features from the vegetation 3D model, and to perform spatiotemporal feature encoding and fusion of the vegetation point cloud features and the meteorological features to obtain spatiotemporal fusion features.

[0138] The comparison module 35 is used to calculate the growth change data of vegetation based on the spatiotemporal fusion features and a pre-constructed vegetation growth model, compare the growth change data with a preset safety threshold, and generate a vegetation growth assessment result based on the comparison result.

[0139] The safety inspection, identification, and evaluation system for overhead transmission lines based on laser point clouds in this application is used to implement the aforementioned safety inspection, identification, and evaluation method for overhead transmission lines based on laser point clouds. Therefore, the specific implementation of the safety inspection, identification, and evaluation system for overhead transmission lines based on laser point clouds can be found in the embodiment section of the safety inspection, identification, and evaluation method for overhead transmission lines based on laser point clouds mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0140] like Figure 4 As shown, this application also provides an electronic device, including: a memory 41 for storing a computer program; and a processor 42 for executing the computer program to implement the steps of any of the above-described methods for safety inspection, identification, and evaluation of overhead power transmission lines based on laser point clouds.

[0141] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described methods for safety inspection, identification, and evaluation of overhead power transmission lines based on laser point clouds.

[0142] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0143] The embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the safety inspection, identification, and evaluation method for overhead power transmission lines based on laser point clouds.

[0144] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0145] The foregoing has provided a detailed description of the safety inspection, identification, and evaluation method and system for overhead power transmission lines based on laser point clouds, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for safety inspection, identification, and assessment of overhead power transmission lines based on laser point clouds, characterized in that, include: Collect multi-view laser point cloud data and corresponding meteorological environmental data of the same transmission line corridor area at different times, and extract meteorological features from the meteorological environmental data; The multi-view laser point cloud data from various periods are stitched together to obtain complete point cloud data. The complete point cloud data is then segmented to separate the regions containing vegetation points, thus obtaining vegetation point cloud distribution data. Based on the vegetation point cloud distribution data, combined with the multi-dimensional indicators of vegetation, a three-dimensional reconstruction process is performed to obtain a three-dimensional vegetation model. Vegetation point cloud features are extracted from the three-dimensional vegetation model, and the vegetation point cloud features are fused with the meteorological features through spatiotemporal feature encoding to obtain spatiotemporal fused features. Based on the spatiotemporal fusion features, vegetation growth change data is calculated using a pre-constructed vegetation growth model. The growth change data is then compared with a preset safety threshold, and a vegetation growth assessment result is generated based on the comparison result. The process of performing three-dimensional reconstruction based on the vegetation point cloud distribution data, combined with multi-dimensional vegetation indicators, yields a three-dimensional vegetation model, including: Point cloud density analysis is performed on the vegetation point cloud distribution data to obtain the core growth area with a point cloud density greater than a first preset density threshold. Based on spatial proximity, the point cloud in the core growth area is clustered to obtain multiple point groups, each point group corresponding to an independent vegetation. For each point group, the outer contour shape of the point group is calculated to obtain a surface structure that reflects the external morphology of the vegetation. Combined with the multi-dimensional indicators of the vegetation, the detailed features of the surface structure are supplemented and optimized to obtain a three-dimensional model of the vegetation. The method combines multi-dimensional indicators of vegetation to supplement and optimize the detailed features of the surface structure, resulting in a three-dimensional vegetation model, including: The overall height of vegetation is determined based on the vertical distribution range of the point cloud, and the point cloud is divided into several layers in the vertical direction according to the overall height. Calculate the point cloud density within each layer, and identify the layer with a point cloud density greater than the second preset density threshold and the largest horizontal coverage as the main area of ​​the vegetation canopy. The point cloud of the main canopy region is divided into multiple sub-regions based on the local density gradient. For each sub-region, the sub-volume corresponding to the sub-region is estimated based on the number of point clouds in the sub-region. Based on the point cloud density and sub-volume of the sub-region, the parameters of the surface reconstruction algorithm are adaptively adjusted. Based on the adjusted parameters, the sub-region is reconstructed in three dimensions to generate the sub-model corresponding to the sub-region. The sum of the sub-volumes corresponding to all sub-regions is the theoretical total volume, and the theoretical total volume, overall height, and point cloud density within the layer are used to constitute the multi-dimensional indicators of vegetation. Merge the sub-models corresponding to all sub-regions and calculate the total volume of the merged model; When the error between the total volume of the fused model and the theoretical total volume is greater than a preset error, the total volume of the fused model is corrected by scaling transformation until the error between the total volume of the fused model and the theoretical total volume is less than or equal to the preset error. During the correction process, the overall height remains unchanged, and finally a three-dimensional vegetation model is obtained.

2. The method according to claim 1, characterized in that, The step of extracting vegetation point cloud features from the 3D vegetation model, and then performing spatiotemporal feature encoding and fusion of the vegetation point cloud features with the meteorological features to obtain spatiotemporal fused features includes: Calculate the vertical dimensions of the vegetation 3D model to obtain the height feature; calculate the horizontal coverage of the vegetation 3D model to obtain the crown width feature; calculate the spatial size of the vegetation 3D model to obtain the volume feature. Arrange the height feature, crown width feature, and volume feature in chronological order to obtain vegetation point cloud features; By using a spatiotemporal feature encoding method, the vegetation point cloud features and corresponding meteorological features from the same period are processed to obtain spatiotemporal fusion features.

3. The method according to claim 2, characterized in that, The process of processing the vegetation point cloud features and corresponding meteorological features from the same period using a spatiotemporal feature encoding method to obtain spatiotemporal fusion features includes: The vegetation point cloud features and meteorological features from the same period are combined to obtain multiple fused feature units; All the fused feature units are arranged in chronological order to obtain time series features; The fused feature units of adjacent periods in the time series features are analyzed to obtain feature change trend information; By using a spatiotemporal feature coding method, the time series features and the feature change trend information are integrated and processed to obtain spatiotemporal fusion features.

4. The method according to claim 1, characterized in that, The process involves calculating vegetation growth change data based on the spatiotemporal fusion features using a pre-constructed vegetation growth model, comparing the growth change data with a preset safety threshold, and generating a vegetation growth assessment result based on the comparison result, including: Based on the spatiotemporal fusion features, the height growth of vegetation in the vertical direction, the crown growth and density change in the horizontal direction are calculated using a pre-constructed vegetation growth model. The height growth, crown width growth, and density change are compared with their respective preset safety thresholds to obtain vertical comparison results, horizontal comparison results, and density change comparison results. The vertical comparison results, the horizontal comparison results, and the density change comparison results are comprehensively evaluated and processed to generate vegetation growth assessment results.

5. The method according to claim 1, characterized in that, The multi-view laser point cloud data from various periods is stitched together to obtain complete point cloud data. This complete point cloud data is then segmented to separate areas containing vegetation points, resulting in vegetation point cloud distribution data, including: Identify the corresponding points in the overlapping areas between laser point cloud data from different viewpoints, and calculate the transformation parameters for converting the laser point cloud data from each viewpoint to a unified coordinate system based on the corresponding points. Based on the conversion parameters, the multi-view laser point cloud data from each viewpoint is processed by point cloud stitching to obtain complete point cloud data. By using a preset spatial height range and reflection intensity range, each point in the complete point cloud data is segmented to separate the regions containing vegetation points that simultaneously satisfy the spatial height range and reflection intensity range, thus obtaining vegetation point cloud distribution data.

6. A safety inspection, identification, and assessment system for overhead power transmission lines based on laser point clouds, characterized in that, To implement the safety inspection, identification, and assessment method for overhead power transmission lines based on laser point clouds as described in claim 1, the method includes: The acquisition module is used to acquire multi-view laser point cloud data and corresponding meteorological environmental data of the same transmission line corridor area at different times, and extract meteorological features from the meteorological environmental data. The separation module is used to perform point cloud stitching processing on the multi-view laser point cloud data of each period to obtain complete point cloud data, and to perform segmentation processing on the complete point cloud data to separate the area containing vegetation points to obtain vegetation point cloud distribution data. The reconstruction module is used to perform three-dimensional reconstruction processing based on the vegetation point cloud distribution data and combined with the multi-dimensional indicators of vegetation to obtain a three-dimensional vegetation model. The fusion module is used to extract vegetation point cloud features from the vegetation 3D model, and to perform spatiotemporal feature encoding and fusion of the vegetation point cloud features with the meteorological features to obtain spatiotemporal fusion features. The comparison module is used to calculate vegetation growth change data based on the spatiotemporal fusion features and a pre-constructed vegetation growth model, compare the growth change data with a preset safety threshold, and generate vegetation growth assessment results based on the comparison results.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the safety inspection, identification, and evaluation method for overhead power transmission lines based on laser point clouds as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the safety inspection, identification, and evaluation method for overhead power transmission lines based on laser point clouds as described in any one of claims 1 to 5.

Images