Orthophoto-constrained mountain photovoltaic string point cloud segmentation and parameter extraction method

Abstract

The application relates to the technical field of three-dimensional point cloud processing, and discloses a mountain photovoltaic string point cloud segmentation and parameter extraction method constrained by orthographic images, which is used for solving the problems of insufficient point cloud segmentation precision and inaccurate parameter extraction of the prior art in a mountain photovoltaic scene. Three-dimensional point cloud and orthographic image data of a mountain photovoltaic station are obtained through unmanned aerial vehicle aerial survey, a photovoltaic string two-dimensional vector region is extracted in the orthographic image as a space constraint, candidate point cloud is obtained through projection screening of the three-dimensional point cloud, after voxel down-sampling and planarity determination, a double-scale height envelope is constructed, and hole compensation, connected domain size screening and normal consistency are combined to realize accurate photovoltaic string point cloud segmentation, effective point cloud is reflected to original high-density point cloud, and point cloud quantity, geometric area, spatial center and normal vector parameters are extracted. The application can effectively inhibit background interference, and improve the segmentation integrity of photovoltaic string point cloud and the parameter extraction precision of the photovoltaic string point cloud in a complex mountain environment.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional point cloud processing technology, specifically to a method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos. Background Technology

[0002] With the advancement of the national new energy development strategy, centralized and distributed photovoltaic power stations are widely built in complex terrain areas such as mountains and hills. Due to the influence of terrain conditions, the spatial distribution of photovoltaic modules is characterized by large differences in ground elevation, varying installation heights, and different array tilt angles and orientations. UAV aerial surveying combined with photogrammetry technology can acquire high-resolution images and 3D point cloud data, providing data support for the spatial positioning, geometric modeling, and operation and maintenance of photovoltaic modules. Through point cloud processing, module identification, parameter estimation, and feature extraction can be achieved, facilitating the digital management and intelligent operation and maintenance of power stations.

[0003] Existing photovoltaic point cloud processing methods are mainly based on single three-dimensional point cloud analysis, using methods such as fixed height threshold segmentation, overall ground model fitting, or plane segmentation to extract component point clouds. These methods are applicable to certain scenarios with flat terrain and uniform component installation height, but they have significant shortcomings in complex mountainous terrain.

[0004] Mountainous photovoltaic (PV) power plants experience significant variations in surface elevation, leading to unstable relative height distribution of modules. Relying solely on absolute elevation thresholds makes it difficult to distinguish between ground points, support points, and panel points, resulting in misjudgments or incorrect exclusions. This leads to high segmentation noise, missing regions, and compromised parameter extraction accuracy. Furthermore, the spatial proximity and similar geometric features of support and panel point clouds make them difficult to differentiate using existing geometric feature segmentation methods, especially when point cloud density is insufficient or obstructed, easily misidentifying support points as panel points and reducing segmentation accuracy. Inconsistent tilt angles and orientations among different arrays mean that a single planar model cannot accommodate modules with multiple orientations, increasing fitting errors, causing unstable region growth, and affecting point cloud integrity. Moreover, existing technologies process 2D orthophotos and 3D point clouds independently, failing to fully utilize the advantages of clear boundaries and precise positioning of orthophotos, using them only as auxiliary displays. This results in a large segmentation search space, a high probability of missegmentation, and low efficiency. Existing segmentation methods based solely on 3D point cloud geometric features are insufficient to meet the segmentation accuracy and stability requirements of mountainous PV power plants and cannot adapt to various problems in complex terrains. Therefore, there is an urgent need to propose a method that integrates spatial information from two-dimensional orthophotos with structural features from three-dimensional point clouds. This method introduces orthophoto regions as spatial prior constraints and combines them with a terrain-adaptive point cloud fine-tuning strategy to achieve accurate segmentation of component point clouds and automatic extraction of multiple parameters, thereby meeting the needs of digital management and intelligent operation and maintenance of mountain photovoltaic power stations. Summary of the Invention

[0005] This invention proposes a method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos, in order to solve the problems of insufficient point cloud segmentation accuracy and inaccurate parameter extraction in existing technologies for mountain photovoltaic scenarios.

[0006] This invention provides the following technical solution: a method for segmenting and extracting parameters of point clouds of mountain photovoltaic strings constrained by orthophotos, comprising the following steps: S1. Acquire and preprocess 3D point cloud data of mountain photovoltaic strings to generate UAV orthophotos with the same spatial reference system, and extract the 2D polygon vector region of the photovoltaic strings in the orthophotos as the spatial constraint for point cloud segmentation. S2. Orthographically project the 3D point cloud onto the corresponding 2D plane coordinate system, select the point clouds whose projection falls into the vector region as candidate point clouds, and perform voxelization downsampling to obtain downsampled point clouds; S3. Construct a local neighborhood for the downsampled point cloud, perform planarity determination, and filter to obtain planar candidate point clouds. Construct fine-scale and coarse-scale grids in two-dimensional space, statistically analyze the height distribution of the planar candidate point clouds, and construct dual-scale height envelope constraints. Filter to obtain height-effective point clouds, and perform hole compensation on points that do not pass the height envelope constraints to obtain complete photovoltaic point clouds. S4. Perform connected component analysis and geometric size screening on the complete photovoltaic point cloud to select the photovoltaic string region. Introduce normal consistency constraints to refine the selection of the photovoltaic string region. Map the selection results back to the original high-density point cloud and extract the number of points, geometric area, spatial center coordinates and normal vector parameters of the photovoltaic string and individual panel.

[0007] Further, S1 specifically includes acquiring three-dimensional point cloud data containing mountain photovoltaic strings, wherein each point in the point cloud contains three-dimensional spatial coordinate information: ; In the formula, This is the original point cloud; The original point cloud is denoted as the original high-density point cloud set. : ; The point cloud is preprocessed, including using a statistical outlier detection method to construct a neighborhood for each point and calculate the average distance. When the difference between the average distance and the average distance of the neighborhood exceeds a preset threshold, the point cloud is identified as an outlier and removed. The point cloud is then converted from a camera coordinate system or local coordinate system to a unified geographic coordinate system or engineering coordinate system. The point cloud data is then uniformly stored in a point cloud data structure containing three-dimensional coordinate attributes as input point cloud. The two-dimensional polygonal vector region of the photovoltaic string is obtained through automatic extraction or manual annotation of the photovoltaic module boundary in orthophoto image: ; In the formula, This represents the two-dimensional polygonal vector region of the photovoltaic string extracted from the orthophoto. The first vector region boundary The coordinates of each vertex in the two-dimensional coordinate system of the orthophoto image. Polygon vertex number, This represents the total number of boundary vertices of the two-dimensional polygon vector region.

[0008] Further, S2 specifically includes orthographically projecting the three-dimensional point cloud onto the corresponding two-dimensional plane coordinate system: ; In the formula, This represents the projection transformation relationship from three-dimensional space to two-dimensional orthophoto coordinates. For the third point cloud The two-dimensional coordinates of a point projected onto the two-dimensional orthophoto coordinate system The x-axis coordinates of the two-dimensional plane. The y-axis coordinate of the two-dimensional plane. For the third point cloud The coordinates of a point in a three-dimensional coordinate system The x-axis coordinate in three-dimensional space. Let y be the y-axis coordinate in three-dimensional space. The z-axis coordinate in three-dimensional space; The projection relationship is determined by the camera model parameters and the space registration relationship; When satisfied ,Will Add photovoltaic string candidate point cloud subset : ; The point cloud of the photovoltaic string is obtained based on two-dimensional orthophoto constraints; Subset of candidate point cloud The three-dimensional space is divided into voxel meshes: ; In the formula, Let i, j, and k be the voxel mesh elements in three-dimensional space. , , Let x, y, and z be the starting coordinates of the point cloud space on the x, y, and z axes, Let the voxel side length be , for The coordinate range covered in the x-axis direction. for The coordinate range covered in the y-axis direction. for The coordinate range covered in the z-axis direction; Subset of candidate point cloud After dividing the three-dimensional space into a voxel mesh, the geometric center of all points within each voxel is calculated, and the geometric center is used as the representative point of the voxel: ; In the formula, For the first The coordinates of the geometric center of all points within an individual element. For the first The total number of point cloud points contained within an individual element. For the j-th voxel, the first voxel is the first voxel. The three-dimensional spatial coordinates of each point; Point cloud downsampling is performed based on voxel centers to obtain a downsampled point cloud set: ; In the formula, This is the set of downsampled point clouds obtained after voxelization downsampling. For the first The three-dimensional coordinates of the geometric center point of an individual element, i.e. , This represents the total number of points in the point cloud after voxel downsampling. The total number of points in the original point cloud before voxel downsampling.

[0009] Furthermore, the process of constructing a local neighborhood for the downsampled point cloud, performing planarity determination, and filtering to obtain planar candidate point clouds includes, for the set of downsampled point clouds... Each point in ,by Construct a spatial neighborhood with a fixed radius around the center: ; In the formula, For the point Centered on, with radius The set of points within the spatial neighborhood, For the neighboring region The three-dimensional coordinates of the points For the downsampled point cloud set, the first The three-dimensional coordinates of the points For the point The search radius of the neighborhood centered on; Calculate the mean of the neighborhood point set: ; In the formula, For neighborhood point set The coordinates of the geometric mean point, For neighborhood point set The number of points contained therein. This is to sum the coordinates of all points in the neighborhood; Planarity is determined by eigenvalue decomposition of the covariance matrix, and the covariance matrix is ​​constructed as follows: ; In the formula, For the first The covariance matrix of the neighborhood corresponding to each point Let be the coordinate deviation vector of points in the neighborhood relative to the mean point. This is a vector transpose operation; Perform eigenvalue decomposition on the covariance matrix: ; In the formula, , , These are the covariance matrices. The smallest, second smallest, and largest eigenvalues ​​obtained from the decomposition; When the following two conditions are met, the point will be... Determined as a candidate point in the plane: ; ; In the formula, For the planarity scale threshold, The value range is 0.08 to 0.15. For thickness threshold, satisfying , The average point spacing in the point cloud. This is the thickness threshold scaling factor. The value range is 1.5 to 3.0, under point cloud conditions without a transfer unit. The value range is 0.1 to 0.3 m.

[0010] Furthermore, the process of constructing fine-scale and coarse-scale grids in a two-dimensional space, statistically analyzing the height distribution of candidate point clouds in the plane, constructing dual-scale height envelope constraints, and filtering to obtain highly effective point clouds includes constructing a coarse-scale grid in a two-dimensional space. With fine-scale mesh The fine-scale mesh size is smaller than the coarse-scale mesh size. Within each grid cell, statistically analyze the height distribution of candidate points on the plane and calculate the high quantile height value: ; ; In the formula, This represents the high quantile height value of the candidate points in the plane within a coarse-scale grid cell. This represents the high quantile height value of the candidate points in the plane within the fine-scale grid cell. This function performs high quantile statistical calculations on elevation data z within a coarse-scale grid cell. A function for performing high quantile statistical calculations on elevation data z within a fine-scale grid cell; For each planar candidate point, when the height of the planar candidate point... satisfy or The height is then determined to be a valid point cloud height. For fine-scale height tolerance, the value is 0.3–0.8 m. The height tolerance is a coarse-scale tolerance, ranging from 0.8 to 1.5 m.

[0011] Furthermore, the step of performing hole compensation on points that fail to pass the height envelope constraint to obtain a complete photovoltaic point cloud includes, for points that fail to pass the height constraint , construction point The corresponding spatial neighborhood point set : ; In the formula, For point The corresponding spatial neighborhood point set, For the neighboring region One point, These are the points in the original point cloud that failed the height constraint screening after being filtered by the dual-scale height envelope constraint. For the point The search radius of the neighborhood centered on; Calculate the proportion of effective points within the neighborhood: ; In the formula, For point The proportion of effective points within the neighborhood. For point The total number of all points in the neighborhood. This represents the number of points within the neighborhood that have been determined to be valid. when When, then the determination point The local area is a continuous and valid region, and the point is... Re-evaluated as a valid photovoltaic point The threshold for hole compensation ratio, The value range is 0.6 to 0.7.

[0012] Furthermore, the step of performing connectivity analysis and geometric size screening of the photovoltaic string region on the complete photovoltaic point cloud includes constructing spatial adjacency relationships for all valid photovoltaic points, performing spatial connectivity analysis, and extracting a set of connected regions from the point cloud. : ; Calculate the geometric dimensions of each connected region in the horizontal projection direction: ; ; In the formula, This represents the length of the current connected region along the horizontal projection x-axis. This represents the maximum x-coordinate of all points within the current connected region. It is the minimum x-coordinate of all points within the current connected region. This represents the width of the current connected region along the horizontal projection y-axis. This represents the maximum y-coordinate of all points within the current connected region. It is the minimum value of the y-coordinate of all points within the current connected region; When the geometric dimensions meet the preset minimum length and width constraints of the photovoltaic module, the connected region is determined to be a valid photovoltaic string region; The spatial connectivity analysis employs a clustering algorithm or a region growing algorithm based on Euclidean distance, and the geometric constraints include the length range, width range, and area range of the photovoltaic string.

[0013] Furthermore, the refinement and screening process by introducing normal consistency constraints on the photovoltaic string region includes calculating the overall principal normal vector of each photovoltaic string region. and the local normal vectors of each point within the region and calculate and The included angle : ; when Points with inconsistent normals are retained, while those with inconsistent normals are discarded, thereby improving the geometric consistency of the point cloud of photovoltaic modules. The threshold of the angle between the normal vectors, The value ranges from 10° to 30°.

[0014] Furthermore, the step of backmapping the screening results to the original high-density point cloud and extracting the point cloud quantity, geometric area, spatial center coordinates, and normal vector parameters of the photovoltaic string and individual panel includes, based on spatial neighborhood search, backmapping the points identified as photovoltaic modules in the downsampled point cloud to the original high-density point cloud to obtain a complete and continuous photovoltaic string and panel point cloud. : ; In the formula, For a complete and continuous point cloud of photovoltaic string panels, This is the original high-density point cloud collection. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point, This is the set of downsampling point clouds of photovoltaic modules obtained after judgment. For the downsampled point cloud set, the first The three-dimensional coordinates of the points The search radius for the back-mapping; Calculate the number of points, geometric area, spatial center coordinates, and normal vector parameters of the point cloud for each photovoltaic string; The number of point clouds : ; In the formula, Point cloud for photovoltaic string panels The total number of points contained therein; The geometric area The point cloud of the photovoltaic string panels is triangulated, and the total projected area is calculated. ; In the formula, After triangulation of the point cloud of the photovoltaic string panel, the first The area of ​​each triangle; The coordinates of the space center : ; In the formula, The spatial center coordinates of the point cloud of the photovoltaic string panel. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point; The normal vector parameters : ; In the formula, The global normal vector of the point cloud of the photovoltaic string panel. , , These are the components of the normal vector along the x-axis, y-axis, and z-axis, respectively. Then, the installation orientation and azimuth angle of the string are calculated.

[0015] Compared with the prior art, the present invention has the following beneficial effects: By introducing two-dimensional vector regions extracted from orthophotos as spatial prior constraints for three-dimensional point cloud segmentation, the collaborative utilization of two-dimensional image information and three-dimensional point cloud data is achieved, significantly reducing the search space for point cloud segmentation. A height reference surface is constructed within a local area of ​​the photovoltaic string, and point cloud elevation is normalized, effectively eliminating the influence of complex terrain undulations on the point cloud elevation distribution. Combined with planar consistency constraints, the normalized point cloud is refined and screened, fully utilizing the engineering structural characteristics of the photovoltaic panels to improve the accuracy and stability of the segmentation results. Automatic grouping of photovoltaic panel string-level point clouds and extraction of multiple parameters are achieved, providing reliable data support for refined modeling, operation and maintenance inspection, and digital management of photovoltaic power plants. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process of the method of the present invention; Figure 2 To create a histogram of slope distribution in the study area; Figure 3 This is a schematic diagram of the two-dimensional orthophoto vector constraint region of a photovoltaic string; Figure 4 Top view of the photovoltaic string point cloud after orthophoto constraint; Figure 5 Front view of the photovoltaic string point cloud after orthophoto constraint; Figure 6 The above are top views of the photovoltaic string point cloud after processing by the method of the present invention, wherein (a) is the top view of the first group of photovoltaic string point clouds, (b) is the top view of the second group of photovoltaic string point clouds, and (c) is the top view of the third group of photovoltaic string point clouds. Figure 7 This is a front view of the photovoltaic string point cloud after processing by the method of the present invention; Figure 8 The left view of the photovoltaic string point cloud after processing by the method of the present invention is shown in (a), which is a left view of the photovoltaic string point cloud with a double string structure, and (b) is a left view of the photovoltaic string point cloud with a single string structure. Detailed Implementation

[0017] The present invention will be further illustrated below with reference to embodiments. These embodiments are for illustrative purposes only and are not intended to limit the invention in any way. It should be understood that the described embodiments are merely some, not all, of the embodiments described in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0018] Example 1 Example 1 uses the Zibo Dazhangzhuang Mountain Centralized Photovoltaic Power Station as an application scenario. It uses an UAV oblique photogrammetry system to acquire photovoltaic power station image data and generate corresponding three-dimensional point cloud data and orthophoto data to achieve fine segmentation and parameter extraction of photovoltaic string point cloud.

[0019] A method for point cloud segmentation and parameter extraction of mountain photovoltaic arrays constrained by orthophotos includes the following steps: like Figure 1 As shown, a drone platform equipped with a visible light camera was used to conduct multi-view aerial surveys of a mountain photovoltaic power station, acquiring continuous image data. Three-dimensional point cloud data was then reconstructed using photogrammetry. Each point in the point cloud contains three-dimensional spatial coordinate information. ; In the formula, This is the original point cloud; The original point cloud is denoted as the original high-density point cloud set. : ; The point cloud is preprocessed, which includes constructing a neighborhood for each point and calculating the average distance using a statistical outlier detection method. When the difference between the point and the average distance of the neighborhood exceeds a preset threshold, the point is identified as an outlier and removed. Unify the point cloud coordinate system by transforming the point cloud from the camera coordinate system or local coordinate system to a unified geographic coordinate system or engineering coordinate system; The point cloud data format is standardized and stored uniformly as a point cloud data structure containing three-dimensional coordinate attributes, which serves as the input point cloud for subsequent processing.

[0020] like Figure 2 The histogram of slope distribution in the study area is shown, which can intuitively reflect the topographic undulation characteristics of the mountain photovoltaic power station and provide a priori topographic reference for subsequent dual-scale height envelope screening.

[0021] Based on image data collected by UAVs, photogrammetric modeling methods are used to generate orthophotos that are in the same spatial reference frame as the three-dimensional point cloud data.

[0022] like Figure 3 As shown, Figure 3 This is a schematic diagram of the two-dimensional orthorectified vector constraint region of the photovoltaic string; in the orthorectified image, the two-dimensional polygonal vector region of the photovoltaic string panel is extracted by automatic image segmentation algorithm or manual interactive annotation: ; In the formula, This represents the two-dimensional polygonal vector region of the photovoltaic string extracted from the orthophoto. The first vector region boundary The coordinates of each vertex in the two-dimensional coordinate system of the orthophoto image. Polygon vertex number, The total number of boundary vertices of the two-dimensional polygonal vector region; The two-dimensional vector region serves as a spatial constraint prior for the photovoltaic string in the horizontal projection direction, and is used to limit the search range of the three-dimensional point cloud segmentation.

[0023] Orthophoto projection of the 3D point cloud onto the corresponding 2D plane coordinate system: ; In the formula, This represents the projection transformation relationship from three-dimensional space to two-dimensional orthophoto coordinates. For the third point cloud The two-dimensional coordinates of a point projected onto the two-dimensional orthophoto coordinate system The x-axis coordinates of the two-dimensional plane. The y-axis coordinate of the two-dimensional plane. For the third point cloud The coordinates of a point in a three-dimensional coordinate system The x-axis coordinate in three-dimensional space. Let y be the y-axis coordinate in three-dimensional space. The z-axis coordinate in three-dimensional space; The projection relationship is determined by the camera model parameters and the space registration relationship; When satisfied ,Will Add photovoltaic string candidate point cloud subset : ; The point cloud of the photovoltaic string is obtained based on two-dimensional orthophoto constraints; like Figure 4 As shown, Figure 4 This is a top-down view of the photovoltaic string point cloud after orthophoto constraint, as shown below. Figure 5 As shown, Figure 5 This is a front view of the photovoltaic string point cloud after orthophoto constraint. It can be seen that a large number of terrain and background interference point clouds have been initially removed after orthophoto constraint.

[0024] Subset of candidate point cloud The three-dimensional space is divided into voxel meshes: ; In the formula, Let i, j, and k be the voxel mesh elements in three-dimensional space. , , Let x, y, and z be the starting coordinates of the point cloud space on the x, y, and z axes, Let the voxel side length be , for The coordinate range covered in the x-axis direction. for The coordinate range covered in the y-axis direction. for The coordinate range covered in the z-axis direction; Subset of candidate point cloud After dividing the three-dimensional space into a voxel mesh, the geometric center of all points within each voxel is calculated, and the geometric center is used as the representative point of the voxel: ; In the formula, For the first The coordinates of the geometric center of all points within an individual element. For the first The total number of point cloud points contained within an individual element. For the j-th voxel, the first voxel is the first voxel. The three-dimensional spatial coordinates of each point; Point cloud downsampling is performed based on voxel centers to obtain a downsampled point cloud set: ; In the formula, This is the set of downsampled point clouds obtained after voxelization downsampling. For the first The three-dimensional coordinates of the geometric center point of an individual element, i.e. , This represents the total number of points in the point cloud after voxel downsampling. The total number of points in the original point cloud before voxel downsampling; This step is used to reduce point cloud density, reduce computational load, and suppress local noise.

[0025] A local neighborhood is constructed for the downsampled point cloud, planarity is determined, and planar candidate point clouds are obtained through screening, including for the set of downsampled point clouds. Each point in ,by Construct a spatial neighborhood with a fixed radius around the center: ; In the formula, For the point Centered on, with radius The set of points within the spatial neighborhood, For the neighboring region The three-dimensional coordinates of the points For the downsampled point cloud set, the first The three-dimensional coordinates of the points For the point The search radius of the neighborhood centered on; Calculate the mean of the neighborhood point set: ; In the formula, For neighborhood point set The coordinates of the geometric mean point, For neighborhood point set The number of points contained therein. This is to sum the coordinates of all points in the neighborhood; Planarity is determined by eigenvalue decomposition of the covariance matrix, and the covariance matrix is ​​constructed as follows: ; In the formula, For the first The covariance matrix of the neighborhood corresponding to each point Let be the coordinate deviation vector of points in the neighborhood relative to the mean point. This is a vector transpose operation; Perform eigenvalue decomposition on the covariance matrix: ; In the formula, , , These are the covariance matrices. The smallest, second smallest, and largest eigenvalues ​​obtained from the decomposition; When the following two conditions are met, the point will be... Determined as a candidate point in the plane: ; ; In the formula, For the planarity scale threshold, The value range is 0.08 to 0.15. For thickness threshold, satisfying , The average point spacing in the point cloud. This is the thickness threshold scaling factor. The value range is 1.5 to 3.0, under point cloud conditions without a transfer unit. The value range is 0.1 to 0.3 m; Point The points are identified as planar candidate points, and thus ground undulation points, vegetation points, and support points are proposed.

[0026] Constructing coarse-scale meshes in two-dimensional space With fine-scale mesh The fine-scale mesh size is smaller than the coarse-scale mesh size. Within each grid cell, statistically analyze the height distribution of candidate points on the plane and calculate the high quantile height value: ; ; In the formula, This represents the high quantile height value of the candidate points in the plane within a coarse-scale grid cell. This represents the high quantile height value of the candidate points in the plane within the fine-scale grid cell. This function performs high quantile statistical calculations on elevation data z within a coarse-scale grid cell. A function for performing high quantile statistical calculations on elevation data z within a fine-scale grid cell; For each planar candidate point, when the height of the planar candidate point... satisfy or The height is then determined to be a valid point cloud height. For fine-scale height tolerance, the value is 0.3–0.8 m. For coarse-scale height tolerance, the value is 0.8 to 1.5 m; Fine scales are used to depict the upper limit of local height, while coarse scales are used to suppress the influence of large-scale terrain undulations.

[0027] Aperture compensation is performed on points that fail the height envelope constraint to obtain a complete photovoltaic point cloud, including points that fail the height constraint. , construction point The corresponding spatial neighborhood point set : ; In the formula, For point The corresponding spatial neighborhood point set, For the neighboring region One point, These are the points in the original point cloud that failed the height constraint screening after being filtered by the dual-scale height envelope constraint. For the point The search radius of the neighborhood centered on; Calculate the proportion of effective points within the neighborhood: ; In the formula, For point The proportion of effective points within the neighborhood. For point The total number of all points in the neighborhood. This represents the number of points within the neighborhood that have been determined to be valid. when When, then the determination point The local area is a continuous and valid region, and the point is... Re-evaluated as a valid photovoltaic point The threshold for hole compensation ratio, The value range is 0.6 to 0.7; Based on the principle of spatial neighborhood consistency, the distribution ratio of effective points in the local neighborhood is determined by statistical analysis, which is used to restore the continuous structure of the point cloud and suppress the influence of noise.

[0028] For all valid photovoltaic points, construct spatial adjacency relationships, perform spatial connectivity analysis, and extract the set of connected regions from the point cloud. : ; Calculate the geometric dimensions of each connected region in the horizontal projection direction: ; ; In the formula, This represents the length of the current connected region along the horizontal projection x-axis. This represents the maximum x-coordinate of all points within the current connected region. It is the minimum x-coordinate of all points within the current connected region. This represents the width of the current connected region along the horizontal projection y-axis. This represents the maximum y-coordinate of all points within the current connected region. It is the minimum value of the y-coordinate of all points within the current connected region; When the geometric dimensions meet the preset minimum length and width constraints of the photovoltaic module, the connected region is determined to be a valid photovoltaic string region; The minimum length and width constraints are adjusted as follows: ; ; In the formula, and The standard geometric dimensions of a single photovoltaic module. This is the size magnification factor; The spatial connectivity analysis employs a clustering algorithm or a region growing algorithm based on Euclidean distance, and the geometric constraints include the length range, width range, and area range of the photovoltaic string.

[0029] Calculate the overall principal normal vector of each photovoltaic string region. and the local normal vectors of each point within the region and calculate and The included angle : ; when The point is retained if the normal is not consistent; otherwise, points with inconsistent normals are discarded, thereby improving the geometric consistency of the photovoltaic module point cloud. The threshold of the angle between the normal vectors, The value ranges from 10° to 30°; when Points with an angle greater than 10° to 30° are identified as outliers and removed. This is determined by statistically analyzing the normal distribution characteristics of the photovoltaic module panel, and is used to remove outliers while maintaining the geometric consistency of the module.

[0030] Based on spatial neighborhood search, points identified as photovoltaic modules in the downsampled point cloud are backmapped to the original high-density point cloud to obtain a complete and continuous point cloud of photovoltaic string panels. : ; In the formula, For a complete and continuous point cloud of photovoltaic string panels, This is the original high-density point cloud collection. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point, This is the set of downsampling point clouds of photovoltaic modules obtained after judgment. For the downsampled point cloud set, the first The three-dimensional coordinates of the points The search radius for the back-mapping; like Figure 6 As shown, Figure 6 This is a top view of the photovoltaic string point cloud after processing by the method of the present invention, as shown below. Figure 7 As shown, Figure 7 This is a front view of the photovoltaic string point cloud after processing by the method of the present invention. Figure 8 The left view of the photovoltaic string point cloud after processing by the method of the present invention; by Figures 6 to 8 As can be seen, the point clouds of the three photovoltaic strings processed by the method of this invention are complete and continuous, free from noise interference, and have clear boundaries, with excellent segmentation effect.

[0031] Calculate the number of points, geometric area, spatial center coordinates, and normal vector parameters of the point cloud for each photovoltaic string; The number of point clouds : ; In the formula, Point cloud for photovoltaic string panels The total number of points contained therein; The geometric area The point cloud of the photovoltaic string panels is triangulated, and the total projected area is calculated. ; In the formula, After triangulation of the point cloud of the photovoltaic string panel, the first The area of ​​each triangle; The coordinates of the space center : ; In the formula, The spatial center coordinates of the point cloud of the photovoltaic string panel. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point; The normal vector parameters : ; In the formula, The global normal vector of the point cloud of the photovoltaic string panel. , , These are the components of the normal vector along the x-axis, y-axis, and z-axis, respectively. Then, the installation orientation and azimuth angle of the string are calculated.

[0032] Example 2 Example 2 is based on Example 1, and some parameters are adjusted for point cloud density, terrain undulation, and photovoltaic module deployment methods of different mountain photovoltaic power stations to verify the adaptability of the method of the present invention under different application conditions.

[0033] A method for point cloud segmentation and parameter extraction of mountain photovoltaic arrays constrained by orthophotos includes the following steps: A drone platform equipped with a visible light camera was used to conduct multi-view aerial surveys of a mountainous photovoltaic power station, acquiring continuous image data. Three-dimensional point cloud data was then reconstructed using photogrammetry. Each point in the point cloud contains three-dimensional spatial coordinate information. ; In the formula, This is the original point cloud; The original point cloud is denoted as the original high-density point cloud set. : ; The point cloud is preprocessed, which includes constructing a neighborhood for each point and calculating the average distance using a statistical outlier detection method. When the difference between the point and the average distance of the neighborhood exceeds a preset threshold, the point is identified as an outlier and removed. Unify the point cloud coordinate system by transforming the point cloud from the camera coordinate system or local coordinate system to a unified geographic coordinate system or engineering coordinate system; The point cloud data format is standardized and stored uniformly as a point cloud data structure containing three-dimensional coordinate attributes, which serves as the input point cloud for subsequent processing.

[0034] Based on image data collected by UAVs, photogrammetric modeling methods are used to generate orthophotos that are in the same spatial reference frame as the three-dimensional point cloud data.

[0035] In the orthophoto, the two-dimensional polygonal vector regions of the photovoltaic string panels are extracted using an automatic image segmentation algorithm or manual interactive annotation: ; In the formula, This represents the two-dimensional polygonal vector region of the photovoltaic string extracted from the orthophoto. The first vector region boundary The coordinates of each vertex in the two-dimensional coordinate system of the orthophoto image. Polygon vertex number, The total number of boundary vertices of the two-dimensional polygonal vector region; The two-dimensional vector region serves as a spatial constraint prior for the photovoltaic string in the horizontal projection direction, and is used to limit the search range of the three-dimensional point cloud segmentation.

[0036] Orthophoto projection of the 3D point cloud onto the corresponding 2D plane coordinate system: ; In the formula, This represents the projection transformation relationship from three-dimensional space to two-dimensional orthophoto coordinates. For the third point cloud The two-dimensional coordinates of a point projected onto the two-dimensional orthophoto coordinate system The x-axis coordinates of the two-dimensional plane. The y-axis coordinate of the two-dimensional plane. For the third point cloud The coordinates of a point in a three-dimensional coordinate system The x-axis coordinate in three-dimensional space. Let y be the y-axis coordinate in three-dimensional space. The z-axis coordinate in three-dimensional space; The projection relationship is determined by the camera model parameters and the space registration relationship; When satisfied ,Will Add photovoltaic string candidate point cloud subset : ; The point cloud of the photovoltaic string is obtained based on two-dimensional orthophoto constraints; Subset of candidate point cloud The three-dimensional space is divided into voxel meshes: ; In the formula, Let i, j, and k be the voxel mesh elements in three-dimensional space. , , Let x, y, and z be the starting coordinates of the point cloud space on the x, y, and z axes, Let the voxel side length be , for The coordinate range covered in the x-axis direction. for The coordinate range covered in the y-axis direction. for The coordinate range covered in the z-axis direction; Voxel side length in Example 2 The scale was set to 1.5 times that of Example 1 to accommodate field data with high point cloud density. By increasing the voxel scale, the computational load was further reduced, and the overall processing efficiency was improved.

[0037] Subset of candidate point cloud After dividing the three-dimensional space into a voxel mesh, the geometric center of all points within each voxel is calculated, and the geometric center is used as the representative point of the voxel: ; In the formula, For the first The coordinates of the geometric center of all points within an individual element. For the first The total number of point cloud points contained within an individual element. For the j-th voxel, the first voxel is the first voxel. The three-dimensional spatial coordinates of each point; Point cloud downsampling is performed based on voxel centers to obtain a downsampled point cloud set: ; In the formula, This is the set of downsampled point clouds obtained after voxelization downsampling. For the first The three-dimensional coordinates of the geometric center point of an individual element, i.e. , This represents the total number of points in the point cloud after voxel downsampling. The total number of points in the original point cloud before voxel downsampling.

[0038] For downsampling point cloud ensemble Each point in ,by Construct a spatial neighborhood with a fixed radius around the center: ; In the formula, For the point Centered on, with radius The set of points within the spatial neighborhood, For the neighboring region The three-dimensional coordinates of the points For the downsampled point cloud set, the first The three-dimensional coordinates of the points For the point The search radius of the neighborhood centered on; Calculate the mean of the neighborhood point set: ; In the formula, For neighborhood point set The coordinates of the geometric mean point, For neighborhood point set The number of points contained therein. This is to sum the coordinates of all points in the neighborhood; Planarity is determined by eigenvalue decomposition of the covariance matrix, and the covariance matrix is ​​constructed as follows: ; In the formula, For the first The covariance matrix of the neighborhood corresponding to each point Let be the coordinate deviation vector of points in the neighborhood relative to the mean point. This is a vector transpose operation; Perform eigenvalue decomposition on the covariance matrix: ; In the formula, , , These are the covariance matrices. The smallest, second smallest, and largest eigenvalues ​​obtained from the decomposition; When the following two conditions are met, the point will be... Determined as a candidate point in the plane: ; ; In the formula, For the planarity scale threshold, For thickness threshold, satisfying , The average point spacing in the point cloud. This is the thickness threshold scaling factor; Point If a point is determined to be a candidate point for a planar surface, it is considered a non-panel point and is removed if the conditions are not met.

[0039] Constructing coarse-scale grids in two-dimensional space With fine-scale mesh The fine-scale mesh size is smaller than the coarse-scale mesh size. Within each grid cell, statistically analyze the height distribution of candidate points on the plane and calculate the high quantile height value: ; ; In the formula, This represents the high quantile height value of the candidate points in the plane within a coarse-scale grid cell. This represents the high quantile height value of the candidate points in the plane within the fine-scale grid cell. This function performs high quantile statistical calculations on elevation data z within a coarse-scale grid cell. A function for performing high quantile statistical calculations on elevation data z within a fine-scale grid cell; For each planar candidate point, when the height of the planar candidate point... satisfy or The height is then determined to be a valid point cloud height. For fine-scale height tolerance, the value is 0.3–0.8 m. The height tolerance is a coarse-scale tolerance, ranging from 0.8 to 1.5 m.

[0040] For points that fail the height constraint , construction point The corresponding spatial neighborhood point set : ; In the formula, For point The corresponding spatial neighborhood point set, For the neighboring region One point, These are the points in the original point cloud that failed the height constraint screening after being filtered by the dual-scale height envelope constraint. For the point The search radius of the neighborhood centered on; Calculate the proportion of effective points within the neighborhood: ; In the formula, For point The proportion of effective points within the neighborhood. For point The total number of all points in the neighborhood. This represents the number of points within the neighborhood that have been determined to be valid. when When, then the determination point The local area is a continuous and valid region, and the point is... Re-evaluated as a valid photovoltaic point This is the threshold for the hole compensation ratio.

[0041] After hole compensation, connected component analysis is performed on all valid points to obtain the set of connected regions of the extracted point cloud. : ; Calculate the geometric dimensions of each connected region in the horizontal projection direction: ; ; In the formula, This represents the length of the current connected region along the horizontal projection x-axis. This represents the maximum x-coordinate of all points within the current connected region. It is the minimum x-coordinate of all points within the current connected region. This represents the width of the current connected region along the horizontal projection y-axis. This represents the maximum y-coordinate of all points within the current connected region. It is the minimum value of the y-coordinate of all points within the current connected region; When the geometric dimensions meet the preset minimum length and width constraints of the photovoltaic module, the connected region is determined to be a valid photovoltaic string region; The minimum length and width constraints are adjusted as follows: ; ; In the formula, and The standard geometric dimensions of a single photovoltaic module. This is the size magnification factor; Only connected regions that meet the following conditions are retained: size meets the threshold requirement, point cloud within the region is continuous, and region orientation is consistent with the arrangement of photovoltaic modules. All other regions are discarded.

[0042] The spatial connectivity analysis employs a clustering algorithm or a region growing algorithm based on Euclidean distance, and the geometric constraints include the length range, width range, and area range of the photovoltaic string.

[0043] Calculate the overall principal normal vector of each photovoltaic string region. and the local normal vectors of each point within the region and calculate and The included angle : ; when The point is retained if the normal is not consistent; otherwise, points with inconsistent normals are discarded, thereby improving the geometric consistency of the photovoltaic module point cloud. The threshold of the angle between the normal vectors, The value ranges from 10° to 30°.

[0044] Points identified as photovoltaic modules in the downsampled point cloud are backmapped to the original high-density point cloud to obtain a complete and continuous point cloud of photovoltaic string panels. : ; In the formula, For a complete and continuous point cloud of photovoltaic string panels, This is the original high-density point cloud collection. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point, This is the set of downsampling point clouds of photovoltaic modules obtained after judgment. For the downsampled point cloud set, the first The three-dimensional coordinates of the points The search radius for the back-mapping; Calculate the number of points, geometric area, spatial center coordinates, and normal vector parameters of the point cloud for each photovoltaic string; The number of point clouds : ; In the formula, Point cloud for photovoltaic string panels The total number of points contained therein; The geometric area The point cloud of the photovoltaic string panels is triangulated, and the total projected area is calculated. ; In the formula, After triangulation of the point cloud of the photovoltaic string panel, the first The area of ​​each triangle; The coordinates of the space center : ; In the formula, The spatial center coordinates of the point cloud of the photovoltaic string panel. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point; The normal vector parameters : ; In the formula, The global normal vector of the point cloud of the photovoltaic string panel. , , These are the components of the normal vector along the x-axis, y-axis, and z-axis, respectively. Then, the installation orientation and azimuth angle of the string are calculated.

[0045] This embodiment effectively eliminates interference from terrain, vegetation, and noise point clouds through steps such as orthophoto vector constraints, dual-scale height envelope filtering, connected component size filtering, and point cloud backmapping, obtaining a continuous and complete photovoltaic string point cloud. Based on this, the projected area is calculated using triangulation, and key parameters such as the number of points, spatial center, and normal vector are extracted. This verifies the applicability and reliability of the method in complex mountainous environments, meeting the practical needs of high-precision, automated point cloud processing for digital surveying, installed capacity calculation, and intelligent operation and maintenance of photovoltaic power plants.

[0046] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for point cloud segmentation and parameter extraction of mountain photovoltaic arrays constrained by orthophotos, characterized in that, Includes the following steps: S1. Acquire and preprocess 3D point cloud data of mountain photovoltaic strings to generate UAV orthophotos with the same spatial reference system, and extract the 2D polygon vector region of the photovoltaic strings in the orthophotos as the spatial constraint for point cloud segmentation. S2. Orthographically project the 3D point cloud onto the corresponding 2D plane coordinate system, select the point clouds whose projection falls into the vector region as candidate point clouds, and perform voxelization downsampling to obtain downsampled point clouds; S3. Construct a local neighborhood for the downsampled point cloud, perform planarity determination, and filter to obtain planar candidate point clouds. Construct fine-scale and coarse-scale grids in two-dimensional space, statistically analyze the height distribution of the planar candidate point clouds, and construct dual-scale height envelope constraints. Filter to obtain height-effective point clouds, and perform hole compensation on points that do not pass the height envelope constraints to obtain complete photovoltaic point clouds. S4. Perform connected component analysis and geometric size screening on the complete photovoltaic point cloud to select the photovoltaic string region. Introduce normal consistency constraints to refine the selection of the photovoltaic string region. Map the selection results back to the original high-density point cloud and extract the number of points, geometric area, spatial center coordinates and normal vector parameters of the photovoltaic string and individual panel.

2. The method for segmenting and extracting parameters of point clouds of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, S1 specifically includes acquiring three-dimensional point cloud data containing mountain photovoltaic strings, wherein each point in the point cloud contains three-dimensional spatial coordinate information: ； In the formula, This is the original point cloud; The original point cloud is denoted as the original high-density point cloud set. : ； The point cloud is preprocessed, including using a statistical outlier detection method to construct a neighborhood for each point and calculate the average distance. When the difference between the average distance and the average distance of the neighborhood exceeds a preset threshold, it is identified as an outlier and removed. The point cloud is then transformed from a camera coordinate system or local coordinate system to a unified geographic coordinate system or engineering coordinate system. The point cloud data is then uniformly stored in a point cloud data structure containing three-dimensional coordinate attributes as input point cloud. The two-dimensional polygonal vector region of the photovoltaic string is obtained through automatic extraction or manual annotation of the photovoltaic module boundary in orthophoto image: ； In the formula, This represents the two-dimensional polygonal vector region of the photovoltaic string extracted from the orthophoto. The first vector region boundary The coordinates of each vertex in the two-dimensional coordinate system of the orthophoto image. Polygon vertex number, This represents the total number of boundary vertices of the two-dimensional polygon vector region.

3. The method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, Specifically, S2 includes orthographically projecting the three-dimensional point cloud onto the corresponding two-dimensional plane coordinate system: ； In the formula, This represents the projection transformation relationship from three-dimensional space to two-dimensional orthophoto coordinates. For the third point cloud The two-dimensional coordinates of a point projected onto the two-dimensional orthophoto coordinate system The x-axis coordinates of the two-dimensional plane. The y-axis coordinate of the two-dimensional plane. For the third point cloud The coordinates of a point in a three-dimensional coordinate system The x-axis coordinate in three-dimensional space. Let y be the y-axis coordinate in three-dimensional space. The z-axis coordinate in three-dimensional space; The projection relationship is determined by the camera model parameters and the space registration relationship; When satisfied ,Will Add photovoltaic string candidate point cloud subset : ； The point cloud of the photovoltaic string is obtained based on two-dimensional orthophoto constraints; Subset of candidate point cloud The three-dimensional space is divided into voxel meshes: ； In the formula, Let i, j, and k be the voxel mesh elements in three-dimensional space. , , Let x, y, and z be the starting coordinates of the point cloud space on the x, y, and z axes, Let the voxel side length be , for The coordinate range covered in the x-axis direction. for The coordinate range covered in the y-axis direction. for The coordinate range covered in the z-axis direction; Subset of candidate point cloud After dividing the three-dimensional space into a voxel mesh, the geometric center of all points within each voxel is calculated, and the geometric center is used as the representative point of the voxel: ； In the formula, For the first The coordinates of the geometric center of all points within an individual element. For the first The total number of point cloud points contained within an individual element. For the j-th voxel, the first voxel is the first voxel. The three-dimensional spatial coordinates of each point; Point cloud downsampling is performed based on voxel centers to obtain a downsampled point cloud set: ； In the formula, This is the set of downsampled point clouds obtained after voxelization downsampling. For the first The three-dimensional coordinates of the geometric center point of an individual element, i.e. , This represents the total number of points in the point cloud after voxel downsampling. The total number of points in the original point cloud before voxel downsampling.

4. The method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, The process of constructing a local neighborhood for the downsampled point cloud, performing planarity determination, and filtering to obtain planar candidate point clouds includes, for the set of downsampled point clouds... Each point in ,by Construct a spatial neighborhood with a fixed radius around the center: ； In the formula, For the point Centered on, with radius The set of points within the spatial neighborhood, For the neighboring region The three-dimensional coordinates of the points For the downsampled point cloud set, the first The three-dimensional coordinates of the points For the point The search radius of the neighborhood centered on; Calculate the mean of the neighborhood point set: ； In the formula, For neighborhood point set The coordinates of the geometric mean point, For neighborhood point set The number of points contained therein. This is to sum the coordinates of all points in the neighborhood; Planarity is determined by eigenvalue decomposition of the covariance matrix, and the covariance matrix is ​​constructed as follows: ； In the formula, For the first The covariance matrix of the neighborhood corresponding to each point Let be the coordinate deviation vector of points in the neighborhood relative to the mean point. This is a vector transpose operation; Perform eigenvalue decomposition on the covariance matrix: ； In the formula, , , These are the covariance matrices. The smallest, second smallest, and largest eigenvalues ​​obtained from the decomposition; When the following two conditions are met, the point will be... Determined as a candidate point in the plane: ； ； In the formula, For the planarity scale threshold, The value range is 0.08 to 0.

15. For thickness threshold, satisfying , The average point spacing in the point cloud. This is the thickness threshold scaling factor. The value range is 1.5 to 3.0, under point cloud conditions without a transfer unit. The value range is 0.1 to 0.3 m.

5. The method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, The process involves constructing fine-scale and coarse-scale grids in a two-dimensional space, statistically analyzing the height distribution of candidate point clouds in the plane, constructing dual-scale height envelope constraints, and filtering to obtain point clouds with effective heights. This includes constructing a coarse-scale grid in a two-dimensional space. With fine-scale mesh The fine-scale mesh size is smaller than the coarse-scale mesh size. Within each grid cell, statistically analyze the height distribution of candidate points on the plane and calculate the high quantile height value: ； ； In the formula, This represents the high quantile height value of the candidate points in the plane within a coarse-scale grid cell. This represents the high quantile height value of the candidate points in the plane within the fine-scale grid cell. This function performs high quantile statistical calculations on elevation data z within a coarse-scale grid cell. A function for performing high quantile statistical calculations on elevation data z within a fine-scale grid cell; For each planar candidate point, when the height of the planar candidate point... satisfy or The height is then determined to be a valid point cloud height. For fine-scale height tolerance, the value is 0.3–0.8 m. The height tolerance is a coarse-scale tolerance, ranging from 0.8 to 1.5 m.

6. The method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, The process involves performing hole compensation on points that do not pass the height envelope constraint to obtain a complete photovoltaic point cloud. This includes points that fail to meet height constraints. , construction point The corresponding spatial neighborhood point set : ； In the formula, For point The corresponding spatial neighborhood point set, For the neighboring region One point, These are the points in the original point cloud that failed the height constraint screening after being filtered by the dual-scale height envelope constraint. For the point The search radius of the neighborhood centered on; Calculate the proportion of effective points within the neighborhood: ； In the formula, For point The proportion of effective points within the neighborhood. For point The total number of all points in the neighborhood. This represents the number of points within the neighborhood that have been determined to be valid. when When, then the determination point The local area is a continuous and valid region, and the point is... Re-evaluated as a valid photovoltaic point The threshold for hole compensation ratio, The value range is 0.6 to 0.

7.

7. The method for point cloud segmentation and parameter extraction of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, The process of performing connectivity analysis and geometric size screening of the complete photovoltaic point cloud for photovoltaic string regions includes: constructing spatial adjacency relationships for all valid photovoltaic points, performing spatial connectivity analysis, and extracting a set of connected regions from the point cloud. : ； Calculate the geometric dimensions of each connected region in the horizontal projection direction: ； ； In the formula, This represents the length of the current connected region along the horizontal projection x-axis. This represents the maximum x-coordinate of all points within the current connected region. It is the minimum x-coordinate of all points within the current connected region. This represents the width of the current connected region along the horizontal projection y-axis. This represents the maximum y-coordinate of all points within the current connected region. It is the minimum value of the y-coordinate of all points within the current connected region; When the geometric dimensions meet the preset minimum length and width constraints of the photovoltaic module, the connected region is determined to be a valid photovoltaic string region; The spatial connectivity analysis employs a clustering algorithm or a region growing algorithm based on Euclidean distance, and the geometric constraints include the length range, width range, and area range of the photovoltaic string.

8. The method for segmenting and extracting parameters of point clouds of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, The refinement and screening process by introducing normal consistency constraints on the photovoltaic string region includes calculating the overall principal normal vector of each photovoltaic string region. and the local normal vectors of each point within the region and calculate and The included angle : ； when Points with inconsistent normals are retained, while those with inconsistent normals are discarded, thereby improving the geometric consistency of the point cloud of photovoltaic modules. The threshold of the angle between the normal vectors, The value ranges from 10° to 30°.

9. The method for segmenting and extracting parameters of point clouds of mountain photovoltaic strings constrained by orthophotos according to claim 1, characterized in that, The process of back-mapping the screening results to the original high-density point cloud and extracting the point cloud quantity, geometric area, spatial center coordinates, and normal vector parameters of photovoltaic strings and individual panels includes, based on spatial neighborhood search, back-mapping points identified as photovoltaic modules in the downsampled point cloud to the original high-density point cloud to obtain a complete and continuous photovoltaic string and panel point cloud. : ； In the formula, For a complete and continuous point cloud of photovoltaic string panels, This is the original high-density point cloud collection. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point, This is the set of downsampling point clouds of photovoltaic modules obtained after judgment. For the downsampled point cloud set, the first The three-dimensional coordinates of the points The search radius for the back-mapping; Calculate the number of points, geometric area, spatial center coordinates, and normal vector parameters of the point cloud for each photovoltaic string; The number of point clouds : ； In the formula, Point cloud for photovoltaic string panels The total number of points contained therein; The geometric area The point cloud of the photovoltaic string panels is triangulated, and the total projected area is calculated. ； In the formula, After triangulation of the point cloud of the photovoltaic string panel, the first The area of ​​each triangle; The coordinates of the space center : ； In the formula, The spatial center coordinates of the point cloud of the photovoltaic string panel. Original high-density point cloud In the process, after back-mapping filtering, the data are incorporated into the final photovoltaic string panel point cloud. The One point; The normal vector parameters : ； In the formula, The global normal vector of the point cloud of the photovoltaic string panel. , , These are the components of the normal vector along the x-axis, y-axis, and z-axis, respectively. Then, the installation orientation and azimuth angle of the string are calculated.

Images