Power facility surrounding landform adaptability analysis method based on point cloud data

Abstract

The invention relates to the technical field of electric power facility safety monitoring, and discloses an electric power facility surrounding landform adaptability analysis method based on point cloud data, which is applied to a point cloud processing system comprising data acquisition, processing, analysis and communication modules. Three-dimensional coordinates, point cloud density, textures and other data of landforms around the electric power facility are collected through the laser radar and the photogrammetry unit, after filtering, coordinate conversion and other processing, topographic changes are judged according to the point cloud density, vegetation / water landforms are recognized through the textures, and risk areas such as landslide / settlement are evaluated through the three-dimensional coordinates. Target parameters such as noise level, topographic features and settlement rate are obtained and then output to a user terminal through the communication module according to a preset strategy. The method can accurately analyze the influence of the landform around the power facility on the line safety, provides data support for operation and maintenance, risk early warning and disaster prevention and reduction of the power transmission line, and guarantees the safe and stable operation of a power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power facility safety monitoring, in particular to a power facility surrounding terrain and landform adaptability analysis method based on point cloud data. BACKGROUND

[0002] In the power system, the safe and stable operation of power facilities is of great importance to social and economic development and people's life. Power facilities are usually distributed in complex and diverse terrain and landform environments. The changes, characteristics and potential risk areas of the surrounding terrain and landform will have a significant impact on the safety, reliability and operation efficiency of power facilities.

[0003] Traditional power facility surrounding terrain and landform analysis methods have many limitations. Most traditional methods rely on manual field survey, which not only consumes a lot of manpower, material resources and time cost, but also is limited by the experience and technical level of survey personnel, making it difficult to achieve accurate measurement and comprehensive analysis of large-scale and complex terrain and landform. Manual survey is inefficient and cannot meet the needs of real-time monitoring and rapid response of power facilities. Traditional methods have shortcomings in data collection and processing. For example, for the monitoring of terrain changes, traditional methods may not be able to capture small but critical terrain changes in time, resulting in failure to warn potential safety hazards in advance; in terms of landform feature recognition, it is difficult to accurately distinguish different landform types such as vegetation and water bodies and their distribution, affecting the assessment of the environment around power facilities; for risk area assessment, there is a lack of scientific and quantitative analysis methods, often relying on qualitative judgment, lacking accuracy and reliability.

[0004] With the development of technology, point cloud data technology has been gradually applied in the field of geographic information. Point cloud data can quickly obtain three-dimensional coordinate information, density information, texture information and other rich data of terrain and landform through laser radar scanning, photogrammetry and other technologies, providing new ideas and methods for power facility surrounding terrain and landform analysis. However, the current application of point cloud data technology in power facility surrounding terrain and landform adaptability analysis is not mature enough, and there are some problems to be solved. For example, how to effectively use the multi-dimensional information in point cloud data for accurate judgment of terrain and landform analysis types, how to obtain relevant target parameters according to different analysis types and perform reasonable processing and output, and how to ensure the stability and reliability of point cloud data collection and processing system.

[0005] In addition, the terrain and landform environment around the power facility is complex and changeable, and the terrain and landform features of different regions are quite different, and the influence on the power facility is also different. Therefore, an analysis method capable of adapting to different terrain and landform conditions and having high flexibility and accuracy is needed to meet the needs of power facility safety monitoring in different regions. The existing point cloud data processing method lacks systematic and targeted analysis process and strategy in the specific application scene of the terrain and landform around the power facility, and cannot fully play the advantages of point cloud data technology, and it is difficult to realize comprehensive, in-depth and real-time adaptive analysis of the terrain and landform around the power facility. SUMMARY

[0006] The purpose of the present application is to provide a point cloud data-based adaptive analysis method for the terrain and landform around the power facility to solve the problems raised in the background art.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical solution: a point cloud data-based adaptive analysis method for the terrain and landform around the power facility, comprising: The method is applied to a point cloud processing system, which comprises a data acquisition module, a data processing module, an analysis module and a communication module; the method comprises: acquiring point cloud data around the power facility; wherein the point cloud data comprises three-dimensional coordinate information of the terrain and landform collected by the data acquisition module; determining the terrain and landform analysis type according to the point cloud data; wherein the terrain and landform analysis type comprises: terrain change monitoring type, landform feature recognition type, risk area assessment type; acquiring target parameters related to the terrain and landform analysis type; outputting the target parameters to the user terminal through the communication module according to the preset output strategy.

[0008] Preferably, the data acquisition module comprises a laser radar scanning unit and a photogrammetry unit; the data processing module comprises a point cloud filtering unit and a coordinate conversion unit; the analysis module comprises a feature extraction unit and a classification unit; The point cloud data comprises point cloud density information collected by the laser radar scanning unit and point cloud texture information collected by the photogrammetry unit; The target parameters comprise point cloud noise level information processed by the point cloud filtering unit, point cloud coordinate system information processed by the coordinate conversion unit, terrain feature information extracted by the feature extraction unit, and landform category information output by the classification unit.

[0009] Preferably, the determination of the terrain and landform analysis type according to the point cloud data comprises: determine whether the topography has a significant change according to point cloud density information in the point cloud data; when the topography has a significant change, determine that the topography analysis type includes a topography change monitoring type; determine whether the topography contains vegetation or water according to point cloud texture information in the point cloud data; when the topography contains vegetation or water, determine that the topography analysis type includes a topographic feature identification type; determine whether there is a potential risk area around the power facility according to three-dimensional coordinate information in the point cloud data; when there is a potential risk area, determine that the topography analysis type includes a risk area evaluation type.

[0010] Preferably, before the point cloud data around the power facility is acquired, the method further comprises: activate the data processing module in the dormant state according to the start instruction sent by the user terminal, or automatically activate the data processing module in the dormant state according to a preset acquisition period; when the data processing module is activated, calibrate and detect the laser radar scanning unit and the photogrammetry unit; when there is a calibration deviation in the laser radar scanning unit or the photogrammetry unit, report first alarm information indicating abnormal data acquisition to the user terminal; when there is no calibration deviation in the laser radar scanning unit and the photogrammetry unit, start the laser radar scanning unit and the photogrammetry unit to acquire point cloud data.

[0011] Preferably, the determination of whether the topography has a significant change according to the point cloud density information in the point cloud data comprises: stability check of the real-time acquired point cloud density information to determine whether the point cloud density information is out of the preset threshold range; if yes, determine that the topography has a significant change; if no, determine that the topography has no significant change; the acquisition of the target parameter related to the analysis type comprises: when the topography has a significant change, determine a preset output strategy according to the deviation degree of the point cloud density information, and acquire the point cloud density information as the target parameter according to the sampling frequency indicated by the preset output strategy; when the topography has no significant change, determine that the preset output strategy is periodic output, and acquire the point cloud density information acquired in the current period as the target parameter.

[0012] Preferably, the determining whether the terrain feature contains vegetation or water body according to the point cloud texture information in the point cloud data comprises: when the point cloud texture information indicates the presence of irregular surface, starting the feature extraction unit to perform texture feature detection to determine whether the terrain feature contains vegetation cover; when vegetation cover is detected, determining that the terrain feature contains vegetation; when no vegetation cover is detected, determining that the terrain feature contains bare ground surface; the obtaining of the target parameter related to the analysis type comprises: when the terrain feature contains vegetation, calculating a vegetation height parameter according to the point cloud texture information within a first set time window before the current time, and taking the vegetation height parameter as the target parameter.

[0013] Preferably, the determining whether the power facility surrounding area contains potential risk area according to the three-dimensional coordinate information in the point cloud data comprises: when the three-dimensional coordinate information indicates that the slope value exceeds a safety threshold, determining that there is a landslide risk area; when the three-dimensional coordinate information indicates that the elevation change value exceeds a stability threshold, determining that there is a subsidence risk area; the obtaining of the target parameter related to the analysis type comprises: when there is a landslide risk area, taking the currently detected slope value as the target parameter; when there is a subsidence risk area, calculating a subsidence rate parameter according to the three-dimensional coordinate information of a plurality of elevation sampling points, and taking the subsidence rate parameter as the target parameter.

[0014] Preferably, the calculating of the vegetation height parameter according to the point cloud texture information within a first set time window before the current time comprises: performing outlier removal processing on the point cloud texture information within the first set time window; calculating a texture mean value and a texture variance value according to the processed point cloud texture information; calculating a feature mean value and a feature variance value according to the texture features extracted by the feature extraction unit within the first set time window; constructing a feature matrix based on multi-dimensional feature distribution according to the processed point cloud texture information and texture features within the first set time window, the texture mean value, the texture variance value, the feature mean value and the feature variance value; performing vegetation height prediction through a pre-trained random forest model according to the feature matrix to obtain the vegetation height parameter.

[0015] Preferably, the multiple elevation sampling points are distributed at different positions around the power facility; and the calculation of the subsidence rate parameter according to the three-dimensional coordinate information of the multiple elevation sampling points comprises: For each of the elevation sampling points, a corresponding elevation change time series graph is constructed according to the three-dimensional coordinate information of the corresponding elevation sampling point within a second set time window before the current time; The elevation change time series graphs corresponding to the multiple elevation sampling points are normalized and fused to obtain a comprehensive elevation change time series graph; According to the comprehensive elevation change time series graph, a subsidence trend is identified through a pre-trained support vector machine model to obtain the subsidence rate parameter and risk level information.

[0016] Preferably, the identification of the subsidence trend according to the comprehensive elevation change time series graph through the pre-trained support vector machine model comprises: The comprehensive elevation change time series graph is segmented by a sliding window to generate multiple subsequence segments; For each subsequence segment, a time series feature vector is extracted, including a change rate mean value, a change rate range, and a change rate stability index; The time series feature vector is input into the support vector machine model to output a subsidence trend classification result and a subsidence rate prediction value; According to the subsidence trend classification result and the subsidence rate prediction value, the subsidence rate parameter is updated. In terms of data acquisition and processing, the system obtains point cloud data containing point cloud density information and texture information through a laser radar scanning unit and a photogrammetric unit, processes the data through a point cloud filtering unit and a coordinate conversion unit, effectively improves the quality and availability of the data, and ensures the accuracy of the laser radar scanning unit and the photogrammetric unit through a calibration detection mechanism before data acquisition, avoids data errors caused by equipment deviation, and guarantees the reliability of subsequent analysis from the source. At the same time, the system can automatically activate the data processing module according to user terminal instructions or a preset acquisition period, realizes the automation and intelligentization of data acquisition, reduces manual intervention, and improves work efficiency.

[0017] In terms of terrain and landform analysis type determination, through point cloud density information, texture information and three-dimensional coordinate information in the point cloud data, terrain changes, landform features and potential risk areas are judged respectively, realizing the diversification and precision of analysis types. The stability check based on point cloud density information to determine whether the terrain has changed significantly can timely capture the slight changes of the terrain, providing a scientific basis for terrain change monitoring; using point cloud texture information to detect vegetation coverage can accurately identify landform features, which helps to assess the impact of vegetation on power facilities; through three-dimensional coordinate information to analyze slope value and elevation change value, potential risk areas such as landslides and subsidence can be accurately located, providing quantitative indicators for risk assessment.

[0018] In terms of target parameter acquisition and processing, different algorithms and models are used for parameter calculation and prediction according to different analysis types. For example, in terrain change monitoring, the sampling frequency is determined according to the deviation degree of point cloud density information, realizing dynamic adjustment of data collection and improving the timeliness and pertinence of data; in vegetation height parameter calculation, through outlier removal, multi-dimensional feature matrix construction and random forest model prediction, the accuracy of vegetation height parameter is ensured; in subsidence rate parameter calculation, time series data fusion of elevation sampling points and support vector machine model analysis are used to realize accurate identification of subsidence trend and risk level assessment, providing scientific decision basis for maintenance and management of power facilities.

[0019] In terms of data output and interaction, the system outputs target parameters to user terminal according to preset output strategy through communication module, realizing timely transmission and sharing of analysis results. Users can understand the changes of terrain and landform around power facilities and potential risks in time according to real-time acquired target parameters, so as to take corresponding measures for prevention and treatment, improving the real-time performance and response speed of power facility safety monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The working principle diagram of the power facility surrounding terrain and landform adaptive analysis method based on point cloud data is described. Figure 2 The flowchart for starting point cloud data acquisition is described. Figure 3 The flowchart for terrain change monitoring processing is described. Figure 4 The flowchart for vegetation height parameter calculation is described. DETAILED DESCRIPTION

[0021] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0022] Please refer to Figures 1-4 The present application relates to a power facility surrounding terrain adaptability analysis method based on point cloud data, which is applied to a point cloud processing system including a data acquisition module, a data processing module, an analysis module and a communication module. The specific implementation steps are as follows: The three-dimensional coordinate information of the power facility surrounding terrain is collected by the data acquisition module of the point cloud processing system to form point cloud data. The data acquisition module can realize the stereoscopic acquisition of spatial data based on a laser radar scanning unit, a photogrammetry unit and other devices to ensure that the point cloud data covers the geometric coordinates and texture features of the terrain surface.

[0023] According to the obtained point cloud data, the terrain and geomorphology analysis types are identified, including terrain change monitoring type, geomorphology feature identification type and risk area assessment type. The determination of the analysis type needs to combine the multi-dimensional features of the point cloud data, such as density, texture, three-dimensional coordinates, etc., to provide direction for subsequent parameter extraction.

[0024] According to the determined analysis type, the corresponding target parameters are extracted from each functional module of the point cloud processing system. For example, the point cloud filtering unit in the data processing module can output noise level information, and the coordinate conversion unit can provide coordinate system information; the feature extraction unit and the classification unit in the analysis module can respectively output terrain feature information and geomorphology category information.

[0025] According to the preset output strategy, the target parameters are transmitted to the user terminal by the communication module for the operator to evaluate and decide the power facility surrounding environment. The preset output strategy can be dynamically adjusted according to the analysis type, such as real-time early warning, periodic report and other modes.

[0026] The present application will be further described below in combination with Examples 1 to 5: The hardware architecture of the point cloud processing system in Example 1 is composed of a data acquisition module, a data processing module, an analysis module, and a communication module. Each module realizes data flow and function coordination through standardized interfaces. The data acquisition module includes a laser radar scanning unit and a photogrammetry unit. The former obtains three-dimensional coordinates and point cloud density information of the terrain surface based on the laser pulse ranging principle by emitting laser beams at a high frequency. The spatial resolution can be accurate to the centimeter level. For example, within a 200-meter range around power facilities, high-density point cloud data is collected at a point spacing of 0.5 meters to completely capture the terrain micro-topographic features. The latter uses a multi-spectral camera-equipped unmanned aerial vehicle platform to take multiple-angle photographs of the ground surface and generates point cloud data containing texture information using stereo image matching algorithms. It can identify details such as vegetation leaf texture and water reflection characteristics. For example, for farmland areas under power lines, the texture characteristics of crop growth status are obtained through photogrammetry.

[0027] The data processing module includes a point cloud filtering unit and a coordinate conversion unit. The core function of the point cloud filtering unit is to remove noise points in the original point cloud data. Its processing process is based on statistical analysis or machine learning algorithms. For example, through the radius filtering algorithm, set the neighborhood radius to 1 meter, calculate the distance between each point and its neighborhood points, and determine the points that deviate from the average distance by more than 2 times the standard deviation as noise points. Output point cloud noise level information, which is represented as the percentage of noise points in the total number of points. For example, after filtering, the noise point ratio decreased from 15% to less than 5%. The coordinate conversion unit is responsible for converting the original coordinate data (such as WGS84 geographic coordinate system) obtained by laser radar and photogrammetry into a projection coordinate system suitable for power industry applications (such as Gauss-Kruger projection coordinate system). Through a seven-parameter conversion model or a Bursa model, the coordinate system is translated, rotated, and scaled to output point cloud coordinate system information containing plane coordinates (X, Y) and elevation coordinates (Z), ensuring that point cloud data from different data sources is fused under a unified spatial reference.

[0028] The analysis module is composed of a feature extraction unit and a classification unit. The feature extraction unit extracts terrain feature information based on the geometric and texture features of point cloud data through a digital terrain analysis (DTA) algorithm. Specifically, the point cloud data is used to calculate parameters such as surface slope, slope direction, and roughness. For example, for the area where the power tower is located, the slope value within a 5x5 meter grid is calculated using a moving window algorithm to reflect the steepness of the terrain. For the power transmission line corridor, the elevation profile features along the line are extracted to identify potential terrain undulation hazards. The classification unit uses supervised or unsupervised classification algorithms to automatically identify landform types. For example, based on the support vector machine (SVM) algorithm, multi-dimensional features such as point cloud density, texture, and elevation are used to train a classification model to divide landform categories into forest land, grassland, water area, bare land, and building land, and output landform category information including the area proportion and spatial distribution of each category. For example, within a 500-meter range around a substation, the classification results show that forest land accounts for 40%, bare land accounts for 35%, and water area accounts for 25%.

[0029] The composition of point cloud data has a clear module correspondence relationship. The point cloud density information generated by the laser radar scanning unit directly reflects the spatial sampling density of the terrain surface. In densely vegetated areas, the point cloud density may be reduced due to limited laser penetration, while in bare ground or building surfaces, the point cloud density is relatively uniform. The point cloud texture information collected by the photogrammetry unit reflects the material differences of the ground objects through image gray value or spectral reflectance. For example, the point cloud texture of water areas presents low gray value and high spectral reflectance characteristics, while the point cloud texture of vegetation areas presents medium gray value and specific spectral absorption characteristics.

[0030] The target parameter acquisition process runs through the processing links of each module. The original point cloud data first enters the data processing module. The point cloud filtering unit removes noise points through filtering algorithms and outputs point cloud noise level information, which is used to evaluate data quality. If the noise proportion is too high (e.g., more than 10%), the data resampling mechanism is triggered. The coordinate conversion unit outputs point cloud coordinate system information containing accurate spatial positions after coordinate system conversion, providing a geographical reference for subsequent analysis. The processed point cloud data enters the analysis module. The feature extraction unit calculates terrain parameters such as slope and elevation difference based on geometric features to form terrain feature information. The classification unit outputs landform category information through feature space clustering or model prediction. The above target parameters are sent to the user terminal in a preset data format (such as JSON, CSV) through wired or wireless transmission links (such as 4G / 5G, optical fiber) of the communication module. Users can view the values and spatial distribution patterns of each parameter in real time through a dedicated software interface, such as displaying terrain feature information and landform category information on a GIS platform to intuitively evaluate the terrain and landform adaptability around power facilities.

[0031] In practical application scenarios, such as a certain mountainous power transmission line project, the data acquisition module scans the line corridor by moving the vehicle-mounted laser radar, and simultaneously uses a UAV to carry a photogrammetric device to obtain high-resolution images, generating point cloud data containing density and texture information. The data processing module filters and converts the original data, obtaining high-quality point cloud data with a noise level below 3%, and converting it to the local engineering coordinate system. The analysis module extracts the slope characteristics along the line and finds that a certain section of the line passes through an area with a slope of more than 35°. At the same time, the classification unit identifies the area as shrub land, which poses a risk of insufficient conductor-to-ground distance due to vegetation growth. The communication module transmits target parameters such as slope values and landform categories to the user terminal of the project management department in real time, providing data support for line path optimization and vegetation clearance decision-making.

[0032] In Example 2, through multi-dimensional analysis of point cloud density information, point cloud texture information, and three-dimensional coordinate information, the automatic determination of terrain change monitoring types, landform feature recognition types, and risk area assessment types is achieved. The entire process is based on the physical characteristics of point cloud data and spatial analysis algorithms, combined with the actual needs of power facility surrounding environment monitoring, to build a hierarchical analysis process.

[0033] For the determination of terrain change monitoring types, the core basis is the stability characteristics of point cloud density information. Point cloud density reflects the distribution density of laser radar sampling points in a unit space, and its changes can directly represent structural changes in the terrain surface. For example, when a landslide causes a sudden change in the surface morphology, the point cloud density will show abnormal fluctuations in local density or sparsity due to the intensification of terrain undulations. The specific processing process is as follows: time series analysis is performed on the real-time collected point cloud density information, the density mean and variance in the current window (such as a window size of 1 hour) are calculated through sliding window technology, and compared with the statistical values of the historical baseline window (such as the same time period of the previous 7 days). If the current density mean deviates from the historical mean by more than a pre-set threshold range (such as ±20%), and the variance significantly increases (indicating an increase in the dispersion of density distribution), it is determined that there has been a significant change in the terrain and landscape, triggering the terrain change monitoring type; if the deviation is within the threshold range and the variance is stable, it is determined that there has been no significant change. This process does not require human intervention and achieves dynamic monitoring through real-time data stream computing. For example, during high-risk periods such as the rainy season, the system can automatically shorten the sliding window time to 15 minutes, improving the timeliness of change detection.

[0034] The determination of the landform feature recognition type relies on the pattern recognition of point cloud texture information. The point cloud texture information collected by the photogrammetric unit represents the surface characteristics of the ground objects through parameters such as image gray value and spectral reflectance, and different ground objects (such as vegetation, water, and bare land) have unique texture patterns. The specific determination steps are as follows: when irregular surface features (such as large gray value fluctuation amplitude and multi-peak spectral curve) are detected in the point cloud texture information, the system automatically starts the texture feature detection module of the feature extraction unit. This module calculates parameters such as contrast, correlation, energy, and entropy value of the texture through the gray level co-occurrence matrix (GLCM) algorithm. For example, the texture contrast of the vegetation coverage area is high (the gray difference of the leaf edge is significant), and the entropy value is large (the texture pattern is complex and diverse); while the water surface shows low contrast and low entropy value due to the mirror reflection characteristics. If the texture parameters meet the feature threshold of vegetation coverage (such as contrast > 80 and entropy > 0.6), it is determined that the terrain and landform contain vegetation; if the texture parameters show low complexity and high regularity (such as contrast < 30 and entropy < 0.3), it is further determined to be bare land or water. At this time, the system can combine with the point cloud elevation information to assist in identification - the regular texture of the low elevation area is usually water, while the higher elevation area is determined to be bare land. Through the discrimination logic of multi-feature fusion, vegetation and other ground types can be effectively distinguished, for example, in hilly areas, the system can accurately identify the vegetation boundary of the terrace edge and the bare field.

[0035] The risk area evaluation type determination is based on the three-dimensional coordinate information of the point cloud data, and the potential risks around the power facilities are identified through spatial geometric calculation. The three-dimensional coordinate information contains the X, Y, and Z coordinates of each point on the ground, which can be used to construct a digital elevation model (DEM) and derive key indicators such as slope and elevation change rate. The specific analysis process is as follows: the slope value is calculated using the coordinate difference of adjacent point clouds, and the formula is: wherein is the elevation difference between two points, is the horizontal distance. When the calculated slope value exceeds the safety threshold (such as 45°, which can be adjusted according to the type of power facilities, for example, the threshold for the transmission line corridor can be set to 35°), it is determined that there is a landslide risk area; for the monitoring of elevation change, the system compares the point cloud coordinates of the same area at different times to calculate the elevation change value, and if the cumulative elevation change of a certain area within a certain time exceeds the stability threshold, it is determined that there is a settlement risk area. For example, in the backfilling area around a substation, if continuous monitoring finds that the elevation decreases by 15 cm per month and the slope value gradually increases, the system can comprehensively determine that there is a combined risk of landslide and settlement in this area.

[0036] In practical application scenarios, the above determination logic can form a linkage monitoring mechanism. For example, when the system detects that the density near a mountainous power transmission line abnormally decreases (indicating that the ground surface may collapse), it immediately triggers the analysis of the slope and elevation changes in this area. If it is detected that the slope exceeds 40° and the elevation change rate reaches 5 cm / day, the system automatically determines that the terrain and landform analysis type is a composite type of terrain change monitoring and risk area assessment, and starts high-frequency data acquisition (e.g., once an hour) and multi-module collaborative analysis. For monitoring of vegetation-covered areas, the system can dynamically adjust the feature threshold for vegetation identification based on the seasonal variation of texture characteristics (e.g., increased texture complexity during the spring vegetation regrowth period), to avoid misjudgment due to phenological changes.

[0037] Embodiment 3: Based on Embodiment 2, this embodiment further clarifies the preprocessing process before "acquiring point cloud data around the power facility", covering the activation mechanism of the data processing module, device calibration detection logic, and abnormal alarm mechanism. The entire process is designed through a closed loop of state control, hardware verification, and communication feedback to ensure the accuracy of point cloud data acquisition and the stability of system operation. The core technical details are as follows: The data processing module of the system is in a dormant state by default to reduce energy consumption and prolong the service life of the hardware. There are two activation methods: Instruction-triggered activation: The user terminal sends a start instruction to the point cloud processing system through the communication module. The instruction content includes activation timestamp, task identification, and other information. For example, the maintenance personnel sends an activation instruction containing "task ID: 20250605-01" through the mobile terminal before inspecting the power facility. After receiving the instruction, the data processing module parses the task identification and wakes up the relevant hardware units.

[0038] Periodic automatic activation: The system has a built-in timer that automatically activates the data processing module according to the preset acquisition period. The acquisition period can be flexibly set according to monitoring needs, such as once a day, three times a week, etc. For example, for a mountainous power transmission line, the acquisition period is set to activate once every 7 days to periodically monitor terrain changes during the rainy season.

[0039] After the data processing module is activated, the calibration and detection process of the laser radar scanning unit and the photogrammetry unit is first performed. The core purpose of calibration and detection is to verify the spatial positioning accuracy and data acquisition consistency of the device. The specific steps are as follows: Laser radar calibration and detection: The system calls the built-in calibration target model, which is a standard geometric body (such as a cube) with known three-dimensional coordinates, placed in a fixed position around the power facility. The laser radar scanning unit scans the calibration target to obtain point cloud data of the target surface. By calculating the deviation of the three-dimensional coordinates of the target point cloud from the theoretical coordinates, the positioning accuracy of the laser radar is evaluated. Let the theoretical coordinates of a certain feature point of the target be The measured coordinate obtained by scanning is The single-point positioning deviation is The calculation formula is: wherein, is the theoretical coordinate value, in meters; is the measured coordinate value, in meters; is the spatial straight-line distance deviation, in meters. If the average deviation of all detection points exceeds the preset threshold (such as ±5 cm), it is determined that the laser radar has a calibration deviation.

[0040] Photogrammetry calibration detection: The camera of the photogrammetry unit is calibrated by a chessboard calibration plate. The camera captures an image of the calibration plate, and the focal length, optical center coordinates, distortion coefficient and other parameters of the camera are calculated by Zhang Zhengyou calibration method. During the calibration process, if the absolute value of the radial distortion coefficient of the camera is detected to exceed 0.1 or the absolute value of the tangential distortion coefficient exceeds 0.05, it is determined that the photogrammetry unit has a calibration deviation.

[0041] When any device of the laser radar scanning unit or the photogrammetry unit detects a calibration deviation, the system reports a first warning information to the user terminal through the communication module. The warning information includes the device type (such as "laser radar" or "photogrammetry"), the deviation type (such as "positioning deviation" or "distortion deviation"), the deviation value and the calibration suggestion (such as "please check the installation position of the device" or "suggest to recalibrate the camera"). For example, if the average single-point positioning deviation of the laser radar is 8 cm, which exceeds the threshold of 5 cm, the warning information will display "laser radar has a positioning deviation, the current average deviation is 8 cm, and it is suggested to calibrate the device".

[0042] If the laser radar scanning unit and the photogrammetry unit have no calibration deviation (i.e. all detection parameters are within the preset threshold range), the system starts the two devices to collect point cloud data. The starting process includes: Device initialization: send an initialization instruction to the laser radar to set the scanning range (such as 500 meters in radius), the point cloud density (such as 20 points per square meter), the scanning frequency (such as 10 Hz) and other parameters; at the same time, send a take-off instruction to the unmanned aerial vehicle of the photogrammetry unit to plan a flight route (such as a spiral route centered on the power facility) and set the shooting interval (such as shooting an image every 5 seconds).

[0043] Synchronous acquisition trigger: through hardware synchronous clock or software timestamp mechanism, ensure that the time of laser radar scanning and photogrammetry shooting is synchronized, avoid the spatial matching error of point cloud and image data caused by time difference. For example, after receiving the "synchronous acquisition" instruction, the system sends a trigger signal to the laser radar and the unmanned aerial vehicle at the same millisecond level time point to ensure that the two devices start data collection at the same time.

[0044] Data caching and transmission: The point cloud density information collected by the laser radar in real time and the point cloud texture information obtained by photogrammetry are temporarily stored in the local cache of the system (such as a solid state disk), and after the completion of a single collection task (such as the end of the flight route of the unmanned aerial vehicle), the data are transmitted in batches to the data processing module through the communication module for subsequent processing. For scenes with high real-time requirements, a streaming processing mode can also be used, in which the point cloud data are transmitted in real time to the cloud server through the 5G network.

[0045] In actual application scenarios, the reliability of the preprocessing procedure directly affects the accuracy of the subsequent analysis results. For example, a substation in a coastal area is equipped with a laser radar and an unmanned aerial vehicle photogrammetry system, and the data collection module is automatically activated every 3 days. After a certain activation, the system detects that the radial distortion coefficient of the camera of the photogrammetry unit is 0.12, which exceeds the threshold of 0.1, and immediately reports the first alarm information. After receiving the alarm, the operation and maintenance personnel recalibrate the unmanned aerial vehicle camera, and after calibration, the distortion coefficient is reduced to 0.08, which meets the detection requirements, and the system restarts the collection task to obtain high-quality point cloud data. Through this procedure, errors in landform classification (such as misjudging water bodies as vegetation) or terrain change false alarms (such as slope calculation errors caused by laser radar deviation) caused by device errors are avoided.

[0046] This embodiment establishes a quality control system for point cloud data collection through a standardized activation mechanism, a quantitative calibration detection method, and real-time alarm feedback. The calculation formula for laser radar calibration deviation provides a clear quantitative basis for device accuracy evaluation, ensuring that the detection results of different devices and different periods are comparable. At the same time, the dual activation mode (instruction triggering and periodic automatic) takes into account the needs of emergency monitoring and routine inspection, making the system flexible and adaptable to various application scenarios. The threshold settings (such as laser radar deviation ± 5 cm and camera distortion coefficient ± 0.1) of the calibration detection link can be adjusted according to the device model and monitoring accuracy requirements. For example, in high-precision mapping scenarios, the laser radar deviation threshold can be tightened to ± 3 cm to meet the millimeter-level positioning requirements.

[0047] The entire preprocessing procedure does not require full-time manual supervision, and self-diagnosis and repair guidance of device status are achieved through automated logic, significantly reducing operation and maintenance costs. For power facilities in remote areas, this mechanism can reduce the frequency of manual inspection, especially during adverse weather conditions, the system can automatically extend the sleep cycle to avoid running the device in extreme environments, further improving the reliability and service life of the system.

[0048] Embodiment 4: This embodiment determines the terrain change state by the deviation degree of point cloud density information, and matches the differentiated data acquisition and output mode accordingly. The whole process closely surrounds the power facility surrounding terrain stability monitoring demand, combines the data transmission efficiency and storage resource optimization in the actual application scene, forms an adaptive parameter acquisition mechanism, and the specific implementation details are as follows: When the system determines that the terrain and landscape has changed significantly through the stability check of point cloud density information (such as real-time monitoring finds that the point cloud density of a certain power transmission line along the line suddenly decreases by 18%, exceeding the preset threshold of 15%), the deviation degree needs to be evaluated to determine the emergency level of data acquisition. The deviation degree is measured based on the deviation amplitude of the current point cloud density value from the historical benchmark value, for example, the historical benchmark value is 50 points per square meter, if the real-time value decreases to 40 points per square meter, the deviation amplitude is 20%, which belongs to the medium-high risk level; if it decreases to 30 points per square meter, the deviation amplitude is 40%, which is determined as high risk level. According to different deviation amplitudes, the system automatically calls the corresponding preset output strategy: High-risk scenario: When the deviation amplitude exceeds 30%, the preset output strategy triggers high-frequency data acquisition mode. For example, the scanning frequency of the laser radar is increased from the default 1 time / hour to 1 time / 10 minutes, and the flight frequency of the photogrammetry unit is adjusted synchronously to ensure that the latest point cloud density information is obtained every hour. At this time, the acquisition of target parameters gives priority to real-time, and the communication module pushes the point cloud density value collected each time to the user terminal in real time through the 5G network, and the operation and maintenance personnel can see the dynamic curve of the density value changing with time on the monitoring interface, such as the point cloud density in a certain area decreases from 35 points per square meter to 28 points per square meter within 2 hours, which indicates that the terrain may collapse rapidly.

[0049] Medium-risk scenario: When the deviation amplitude is between 15%-30%, the system adopts dynamic adjustment of sampling frequency. For example, the initial sampling frequency is 1 time / hour, if the density value deviates from the benchmark value continuously for 3 times and there is no convergence trend, the frequency is increased to 1 time / 30 minutes; if the subsequent monitoring shows that the density value tends to be stable, the default frequency is restored. The target parameters are output in packaged form every hour, including the current density value, deviation amplitude, trend prediction (such as "density continues to decrease, suggest attention") and other information, which is convenient for operation and maintenance personnel to master the stage characteristics of terrain change.

[0050] When it is determined that the terrain and landscape has not changed significantly (such as the point cloud density value is monitored for 7 consecutive days and the fluctuation amplitude is less than 10%), the system adopts periodic output strategy. The preset output period can be set according to the importance level of the power facility and the environmental stability, for example: Special facilities (such as hub substations): The output cycle is set to 12 hours, and a point cloud density report is generated at 0 and 12 o'clock every day, including the average, maximum, minimum, and standard deviation of the density during the cycle, reflecting the subtle fluctuations of the terrain surface (such as the slight deformation of the ground surface caused by the day-night temperature difference).

[0051] Primary facilities (such as high-voltage transmission lines): The output cycle is set to 24 hours, and the data of the day is summarized at 2 am every day to generate a report containing a density trend line chart. For example, the point cloud density of a certain line corridor remained between 48-52 points per square meter for a week, with a standard deviation of 1.2, indicating that the terrain state was stable, and the report automatically marked "no significant change."

[0052] In actual application scenarios, this strategy can effectively balance monitoring accuracy and system resource consumption. For example, a mountainous power transmission line passes through a mudslide-prone area. During the rainy season, the system detects that the density value in a certain area drops from 60 points per square meter to 42 points per square meter within 2 hours, with a deviation of 30%, which is determined as a high-risk scenario. The system immediately starts high-frequency acquisition mode, acquiring density data every 10 minutes, and sends a red alert information to the operation and maintenance center through the communication module, accompanied by a real-time density curve and a three-dimensional view of the risk area. The operation and maintenance personnel quickly organize on-site investigation based on the information, confirm that a small-scale landslide has occurred in the area, and timely start the emergency repair plan, avoiding the risk of line outage.

[0053] For scenarios where no significant changes have occurred, take a substation in a plain area as an example. The system has monitored that the surrounding point cloud density fluctuation amplitude is less than 5% for 30 consecutive days, and determines that it is in a stable state. According to the cycle of once a day, the report shows that the average density is 55 points per square meter, the standard deviation is 0.8, and the trend analysis is "terrain has no abnormalities". The operation and maintenance personnel can extend the inspection cycle of this area from once a week to once every two weeks, optimizing the allocation of manpower and resources.

[0054] The parameter configuration of the preset output strategy is extensible, supporting user customization according to regional characteristics. For example, in permafrost areas, seasonal freezing and thawing may cause periodic fluctuations in point cloud density (such as a slight decrease in density during the summer melting period and a recovery during the winter freezing period). By training historical data, the deviation threshold can be temporarily adjusted from 15% to 20% in summer to avoid false alarms caused by natural phenomena. In addition, the system also supports manual intervention to adjust the strategy, such as during major activity power protection, operation and maintenance personnel can manually shorten the output cycle of a certain area from 24 hours to 6 hours to increase the monitoring frequency.

[0055] The output content of the target parameter not only contains the point cloud density value, but also can be associated with the processing results of other modules. For example, in the area where the terrain changes significantly, the system automatically calls the analysis module to extract the slope change information of the area, and forms an associated report with the point cloud density data to assist the operation and maintenance personnel in analyzing the causes of terrain changes (such as density reduction accompanied by slope increase, which may indicate landslide; uniform density reduction accompanied by constant slope, which may indicate vegetation growth blocking laser radar signal). The communication module supports multiple output formats, such as JSON format for system interface, PDF format for report archiving, and visualization interface for real-time monitoring, to meet the needs of users with different roles.

[0056] Embodiment 5: Based on the risk area evaluation of embodiment 3, this embodiment details the parameter calculation process of subsidence risk. Through distributed monitoring of multiple elevation sampling points, time series data fusion and machine learning model analysis, quantitative evaluation of subsidence risk around power facilities is realized.

[0057] In the subsidence monitoring scenario around power facilities, the system uniformly arranges multiple elevation sampling points in the target area. Taking a 220kV substation as an example, 20 elevation sampling points are arranged in a grid-like layout within a 50-meter range outside the substation fence, with a 25-meter spacing between adjacent points, covering the main geological units around the substation (such as backfill soil area, natural foundation area). Each sampling point corresponds to a fixed ground marker (such as stainless steel leveling marker), ensuring that point cloud data at different times can be accurately matched to the same physical location.

[0058] For each elevation sampling point, the system constructs an elevation change time series curve based on three-dimensional coordinate information within the second set time window before the current time. Taking sampling point A as an example, assuming the second set time window is the past 90 days, the system collects the elevation value (unit: meters) of this point once a day, forming a time series containing 90 data points, with the horizontal axis representing the date (such as March 1, 2025 to May 30, 2025) and the vertical axis representing the elevation value. If the elevation value of sampling point A gradually decreases from the initial 50.00 meters to 49.85 meters, and the downward trend accelerates after the 60th day (such as a decrease of 0.12 meters in the last 30 days), the time series curve presents a gentle then steep downward curve.

[0059] After completing the construction of the single sampling point time series curve, the system performs normalization fusion processing on the data of multiple sampling points. Since the elevation benchmarks of each sampling point may differ (such as affected by equipment installation errors), the dimension difference needs to be eliminated first through a benchmark unification algorithm. For example, taking the elevation value of the center of the substation as the benchmark, the elevation values of all sampling points are converted to the difference relative to this benchmark (unit: meters). ), so that the data of different sampling points are in the same dimensional space. After normalization, the time series curves of the 20 sampling points are fused by principal component analysis (PCA) method to retain the main components reflecting the overall subsidence trend of the region (such as the cumulative contribution rate of the first three principal components is more than 85%), and a comprehensive elevation change time series curve graph is generated. This curve graph no longer shows the fluctuations of individual sampling points, but reflects the subsidence trend of the entire monitoring area with a smooth curve, such as a certain regional comprehensive curve shows that the cumulative subsidence in the past 90 days is 0.2 meters, and the subsidence rate in the past 30 days has accelerated to 2 mm / day.

[0060] The system inputs the comprehensive elevation change time series curve graph into a pre-trained support vector machine (SVM) model for subsidence trend identification. In the model training stage, historical subsidence data (such as artificially simulated subsidence scenarios with known subsidence rates) are used to construct a sample set, and each sample contains a time series feature vector of a certain time period and a corresponding subsidence rate label. In actual application, the model divides the comprehensive curve graph into multiple subsequence segments by sliding window segmentation technology, and each segment corresponds to an analysis period (such as 15 days). Taking a subsequence segment as an example, it contains the values of 15 time points, and the system calculates the time series feature vector of the segment, including: average rate of change: the average value of daily elevation change values in 15 days, reflecting the average subsidence rate; range of change rate: the difference between the maximum and minimum daily change values in 15 days, reflecting the subsidence fluctuation amplitude; stability index of change rate: the standard deviation of change values in 15 days, reflecting the stability of subsidence trend.

[0061] After inputting the feature vector into the SVM model, the model outputs two key results: subsidence trend classification result: divided into "stable", "slow subsidence" and "accelerated subsidence". For example, if the average rate of change of a segment is 0.1 mm / day, the range is 0.3 mm, and the standard deviation is 0.05 mm, it is determined to be "stable"; if the average is 1.5 mm / day, the range is 1.0 mm, and the standard deviation is 0.2 mm, it is determined to be "accelerated subsidence".

[0062] subsidence rate prediction value: based on the time series pattern learned by the model training, the subsidence rate of the next period is predicted. For example, according to the characteristics of the current segment, the model predicts that the subsidence rate in the next 15 days will increase to 2.0 mm / day.

[0063] The system updates the subsidence rate parameters and risk level information according to the above results. The risk level is divided into low, medium and high levels, corresponding to different subsidence rate intervals: low risk: subsidence rate ≤0.5 mm / day, such as a certain regional comprehensive curve shows stable change, and the model determines it to be "stable", with a low risk level; Medium risk: 0.5 mm / day < settlement rate < 2.0 mm / day, e.g. a segment is determined as “slow settlement” with a prediction value of 1.2 mm / day, the risk level is medium; High risk: settlement rate > 2.0 mm / day, e.g. the model outputs “accelerating settlement” with a prediction value of 3.0 mm / day, the risk level is high.

[0064] In a practical application scenario, within half a year after the operation of a substation, the system found through elevation sampling points that multiple sampling points in the backfill soil area on the northeast side of the substation had a sustained elevation drop. The integrated time series curve showed that the cumulative settlement in this area was 18 cm in the past 60 days, and the settlement rate increased from 1.5 mm / day to 2.5 mm / day in the past 30 days. The analysis of the latest sub-sequence segment by the SVM model showed that the average change rate was 2.5 mm / day, the range was 1.2 mm, and the standard deviation was 0.3 mm. It was determined as “accelerating settlement”, and the prediction rate would reach 3.0 mm / day in the next 15 days, and the risk level was upgraded to high. The operation and maintenance personnel immediately started the foundation reinforcement project according to the results, avoiding the risk of equipment foundation cracking caused by settlement.

[0065] Another example is a farmland area crossed by a power transmission line. After the end of the rainy season, the system monitored the periodic fluctuations in the elevation values of multiple sampling points (such as a temporary increase in elevation after rain and a drop after drying), but the integrated curve showed that the overall trend was stable, with an average change rate of 0.2 mm / day. The model determined it as “stable” and the risk level was low. The operation and maintenance personnel excluded the risk of surface settlement based on this, avoiding unnecessary inspection work.

[0066] The technical solution of this embodiment combines multi-sampling point cooperative monitoring, data fusion, and machine learning models to solve the possible one-sidedness of single measurement point data, realizing the leap from local single-point monitoring to regional overall evaluation. Its core advantages are: Distributed monitoring: through high-density sampling points, the spatial difference of regional settlement is captured, such as distinguishing the different settlement performances of backfill soil areas and undisturbed soil areas; Time series feature mining: using sliding windows and multi-dimensional feature vectors, the dynamic evolution law of the settlement process is captured, and natural fluctuations and abnormal settlement are distinguished; Risk classification and early warning: through the quantitative parameters and qualitative classification results output by the model, clear risk level guidance is provided for operation and maintenance decision-making, facilitating the priority allocation of resources.

[0067] The system supports flexible configuration of sampling point layout, time window length, feature vector parameters, etc. For example, in complex geological conditions, the sampling points can be densified (with a reduced interval of 10 meters), and in sensitive settlement periods (such as after excavation of a foundation pit), the time window can be shortened to 15 days to improve monitoring sensitivity. In addition, the model can be incrementally trained based on new data on a regular basis to adapt to changes in settlement patterns under different seasons and different geological conditions, ensuring the timeliness and accuracy of the prediction results.

[0068] Through the method of this embodiment, the power industry can achieve early identification and trend prediction of the settlement risk around the facility, transforming the traditional "post-repair" mode into a "prevention" mode, which is especially suitable for areas prone to surface subsidence such as soft soil foundations and mined-out areas. This solution does not require complex on-site instrument deployment and can complete risk assessment through remote analysis of point cloud data, significantly reducing monitoring costs and improving the safety and reliability of power grid operation.

[0069] It should be noted that, in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0070] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives, and variations can be made in the embodiments without departing from the spirit and scope of the present application as defined by the appended claims and their equivalents.

Claims

1.A method for analyzing adaptability of a power facility surrounding topography based on point cloud data, characterized by, The method is applied to a point cloud processing system, and the point cloud processing system comprises a data acquisition module, a data processing module, an analysis module and a communication module; the method comprises: acquiring point cloud data around a power facility; wherein the point cloud data comprises three-dimensional coordinate information of a topography and a geomorphology collected by the data acquisition module; determining a topography and geomorphology analysis type according to the point cloud data; wherein the topography and geomorphology analysis type comprises a topography change monitoring type, a geomorphology feature recognition type and a risk area evaluation type; acquiring target parameters related to the topography and geomorphology analysis type; outputting the target parameters to a user terminal according to a preset output strategy through the communication module. 2.The power facility surrounding topographic feature adaptability analysis method based on point cloud data according to claim 1, wherein, The data acquisition module comprises a laser radar scanning unit and a photogrammetry unit; the data processing module comprises a point cloud filtering unit and a coordinate conversion unit; and the analysis module comprises a feature extraction unit and a classification unit. The point cloud data comprises point cloud density information collected by the laser radar scanning unit and point cloud texture information collected by the photogrammetry unit. The target parameters comprise point cloud noise level information processed by the point cloud filtering unit, point cloud coordinate system information processed by the coordinate conversion unit, topography feature information extracted by the feature extraction unit and geomorphology category information output by the classification unit. 3.The power facility surrounding topographic feature adaptability analysis method based on point cloud data according to claim 2, wherein, The determination of the topography and geomorphology analysis type according to the point cloud data comprises: determining whether the topography and geomorphology have undergone significant changes according to point cloud density information in the point cloud data; when the topography and geomorphology have undergone significant changes, determining that the topography and geomorphology analysis type comprises a topography change monitoring type; determining whether the topography and geomorphology contain vegetation or water bodies according to point cloud texture information in the point cloud data; when the topography and geomorphology contain vegetation or water bodies, determining that the topography and geomorphology analysis type comprises a geomorphology feature recognition type; determining whether there is a potential risk area around the power facility according to three-dimensional coordinate information in the point cloud data; when there is a potential risk area, determining that the topography and geomorphology analysis type comprises a risk area evaluation type. 4.The power facility surrounding topography adaptability analysis method based on point cloud data according to claim 2, wherein, Before the point cloud data around the power facility is acquired, the method further comprises: activating the data processing module in a dormant state according to a start instruction sent by the user terminal, or automatically activating the data processing module in a dormant state according to a preset acquisition period; when the data processing module is activated, calibrating and detecting the laser radar scanning unit and the photogrammetry unit; when the laser radar scanning unit or the photogrammetry unit has a calibration deviation, reporting first alarm information indicating that data acquisition is abnormal to the user terminal; when the laser radar scanning unit and the photogrammetry unit have no calibration deviation, starting the laser radar scanning unit and the photogrammetry unit to collect point cloud data. 5.The power facility surrounding topography adaptability analysis method based on point cloud data according to claim 3, wherein, The determination of whether the topography and geomorphology have undergone significant changes according to the point cloud density information in the point cloud data comprises: performing stability verification on the point cloud density information collected in real time to determine whether the point cloud density information exceeds a preset threshold range; If yes, it is determined that the topography has a significant change; If no, it is determined that the topography has no significant change; The target parameter related to the analysis type is obtained, including: When the topography has a significant change, a preset output strategy is determined according to a deviation degree of the point cloud density information, and the point cloud density information is obtained as the target parameter according to a sampling frequency indicated by the preset output strategy; When the topography has no significant change, a preset output strategy is determined as periodic output, and point cloud density information collected in a current period is obtained as the target parameter. 6.The power facility surrounding topographic feature adaptability analysis method based on point cloud data according to claim 3, wherein, The point cloud texture information in the point cloud data is used to determine whether the topography contains vegetation or water, including: When the point cloud texture information indicates the presence of an irregular surface, the feature extraction unit is started to detect texture features to determine whether the topography contains vegetation coverage; When vegetation coverage is detected, it is determined that the topography contains vegetation; When no vegetation coverage is detected, it is determined that the topography contains bare ground; The target parameter related to the analysis type is obtained, including: When the topography contains vegetation, a vegetation height parameter is calculated according to point cloud texture information in a first set time window before the current time, and the vegetation height parameter is taken as the target parameter. 7.The power facility surrounding topography adaptability analysis method based on point cloud data according to claim 3, wherein, The three-dimensional coordinate information in the point cloud data is used to determine whether there is a potential risk area around the power facility, including: When the three-dimensional coordinate information indicates that a slope value exceeds a safety threshold, it is determined that there is a landslide risk area; When the three-dimensional coordinate information indicates that an elevation change value exceeds a stability threshold, it is determined that there is a subsidence risk area; The target parameter related to the analysis type is obtained, including: When there is a landslide risk area, a currently detected slope value is taken as the target parameter; When there is a subsidence risk area, a subsidence rate parameter is calculated according to three-dimensional coordinate information of a plurality of elevation sampling points, and the subsidence rate parameter is taken as the target parameter. 8.The power facility surrounding topography adaptability analysis method based on point cloud data according to claim 6, wherein, The vegetation height parameter is calculated according to point cloud texture information in a first set time window before the current time, including: Outlier point removal processing is performed on the point cloud texture information in the first set time window; Texture mean value and texture variance value are calculated according to the processed point cloud texture information; Feature mean value and feature variance value are calculated according to texture features extracted by the feature extraction unit in the first set time window; A feature matrix based on multi-dimensional feature distribution is constructed according to the processed point cloud texture information and texture features in the first set time window, the texture mean value, the texture variance value, the feature mean value, and the feature variance value; The vegetation height parameter is obtained by performing vegetation height prediction on the feature matrix through a pre-trained random forest model. 9.The power facility surrounding topography adaptability analysis method based on point cloud data according to claim 7, wherein, The plurality of elevation sampling points are distributed at different positions around the power facility; and the subsidence rate parameter is calculated according to three-dimensional coordinate information of the plurality of elevation sampling points, including: For each of the elevation sampling points, a corresponding elevation change time series curve is constructed according to three-dimensional coordinate information of the corresponding elevation sampling point within a second set time window before the current time; The elevation change time series curves corresponding to the plurality of elevation sampling points are normalized and fused to obtain a comprehensive elevation change time series curve; According to the comprehensive elevation change time series curve, a settlement trend is identified through a pre-trained support vector machine model to obtain a settlement rate parameter and risk level information. 10.The power facility surrounding topography adaptability analysis method based on point cloud data according to claim 9, wherein, The settlement trend identification through the pre-trained support vector machine model according to the comprehensive elevation change time series curve comprises: The comprehensive elevation change time series curve is subjected to sliding window segmentation processing to generate a plurality of sub-sequence segments; For each sub-sequence segment, a time series feature vector is extracted, including a change rate mean, a change rate range, and a change rate stability index; The time series feature vector is input into the support vector machine model to output a settlement trend classification result and a settlement rate prediction value; The settlement rate parameter is updated according to the settlement trend classification result and the settlement rate prediction value.

Images