Photovoltaic land slope stability partition evaluation method

Abstract

The invention relates to a photovoltaic land slope stability partition evaluation method, which comprises the following steps that: an unmanned aerial vehicle flies in an oblique photography manner according to a path planning algorithm, and acquires an image and point cloud data of a slope region; correcting the collected images, including distortion removal, noise processing, image splicing, denoising, filtering and point cloud splicing of the point cloud data; creating a real three-dimensional terrain model by combining the texture information of the image and the height data of the point cloud, and extracting terrain height information from the point cloud data to generate a digital elevation model; performing feature extraction on the three-dimensional terrain model, and analyzing a stability index of the slope; establishing a slope stability evaluation model and performing zoning evaluation, performing stability zoning evaluation according to the feature data and the clustering result, and optimizing a zoning standard; and visualizing a slope stability partition result, generating a stability partition map and a report, establishing a dynamic monitoring system, and regularly using the unmanned aerial vehicle to repeatedly fly to obtain updated slope data.

Description

Technical Field

[0001] The present application relates to the field of geological exploration technology, and in particular to a method for evaluating the stability of photovoltaic land slopes by zoning. Background Art

[0002] Photovoltaic power generation has been widely used in recent years. The selection and preparation of PV sites is crucial. When installing PV systems on sloping terrain, slope stability must be considered to prevent geological hazards from threatening the system. Slope stability is a key factor. Traditional slope stability assessment methods often rely on ground surveys, which are costly, inefficient, inaccurate, and incapable of covering large areas. Summary of the Invention

[0003] The present application provides a photovoltaic land slope stability zoning evaluation method, which can improve the accuracy and efficiency of the evaluation and provide stronger protection for the safety of photovoltaic land.

[0004] The photovoltaic land slope stability zoning evaluation method provided in this application includes the following steps:

[0005] S1. Data collection: The UAV performs oblique photography flight based on the path planning algorithm to collect images and point cloud data of the slope area;

[0006] S2. Image and point cloud data preprocessing: Correct the collected images, including distortion removal, noise processing, and image stitching; perform denoising, filtering, and point cloud stitching on the point cloud data;

[0007] S3, 3D modeling and digital elevation model generation: Combine the texture information of the image and the height data of the point cloud to create a realistic 3D terrain model, and extract the terrain height information from the point cloud data to generate a digital elevation model;

[0008] S4. Feature extraction and analysis: Extract features from the 3D terrain model and analyze slope stability indicators, including slope, soil type, and vegetation cover;

[0009] S5. Slope stability zoning evaluation: Establish a slope stability evaluation model and conduct zoning evaluation. Perform stability zoning evaluation based on characteristic data and clustering results, and optimize zoning criteria.

[0010] S6. Results presentation and application: Visualize the slope stability zoning results, generate stability zoning maps and reports, and establish a dynamic monitoring system. Regularly use drones for repeated flights to obtain updated slope data.

[0011] In addition, the photovoltaic land slope stability zoning evaluation method provided in this application also has the following additional technical features:

[0012] In an optional solution, in step S1, the path planning algorithm of the UAV includes the following steps:

[0013] Initialize the population: first randomly generate a D-dimensional vector of [0,1] as the initial chaotic sequence, and then substitute each dimension of the D-dimensional vector into the following formula in turn:

[0014]

[0015] Calculate and generate a new D-dimensional vector as the second chaotic sequence. Repeat the above steps until N chaotic sequences are generated. Map the generated N chaotic sequences to the parameter range of the UAV flight path to form the initial dung beetle population.

[0016] Then calculate the population diversity by cosine similarity:

[0017]

[0018] When Diver is greater than 0.5, the population diversity is too low and may fall into a local optimum. The number of rolling dung beetles and foraging dung beetles determines the algorithm's ability to explore the solution space and the speed of convergence. Therefore, when the diversity is lower than the threshold, the algorithm's global search ability is enhanced and individuals are mutated into other behavioral strategies to improve search efficiency.

[0019] Using reverse learning strategies to enhance the global search capabilities of reproduction and stealing behaviors:

[0020]

[0021]

[0022] Where, For the reverse solution, lb and ub are D-dimensional vectors representing the lower and upper bounds of each dimension, and rand() is a D-dimensional random vector. is the current feasible solution; by generating the reverse solution of the current solution and comparing its fitness with the original solution, a better solution can be selected to enhance the global search capability.

[0023] In an optional solution, the image captured by the drone is corrected using PhotoScan software in step S2, and the point cloud data is denoised by the following steps:

[0024] First, determine the number of points k in the surrounding area and the standard deviation multiplier a s , for point p in the point cloud data i (i=1,……,N), N represents the number of points in the point cloud, and determines its k nearest neighbor points s j (j=1,……,k), and calculate the k closest points and pi The average distance d between i :

[0025]

[0026] Calculate the average distance d of all points in the point cloud i The expected μ d and standard deviation σ d and the discrimination threshold, according to the expected μ d and standard deviation σ d Calculate the discrimination threshold T:

[0027] T=μ d +α S ·σ d ;

[0028] Finally, the threshold is less than the average distance d in the point cloud i The points are eliminated.

[0029] In an optional solution, in step S2, the point cloud data is filtered using a mean curvature flow filtering algorithm, including the following steps:

[0030]

[0031] Where G(L) is the mean curvature and n(L) is the direction of the external normal vector of the point. The final calculation formula is:

[0032] L t+1 =L t +λG(L)n(L);

[0033] Let the vertex be V i , then the corresponding updates of the vertices are as follows:

[0034]

[0035] Point cloud stitching of the point cloud data includes the following steps:

[0036] First, match the feature points selected at the overlap of adjacent measuring stations and observe the error value and ratio diagram during iteration. The error value after registration is used to determine whether the accuracy requirements are met. If the error value exceeds 3cm, re-registration should be performed until the accuracy standard is met.

[0037] The remaining station cloud data are cyclically processed according to the above steps, and the matched point cloud data are unified together to complete the registration of all station points.

[0038] In an optional solution, step S3 further includes the following steps:

[0039] The Kriging interpolation algorithm based on quantized pigeon flock optimization improves the accuracy and detail of DEM:

[0040] The basic pigeon swarm optimization algorithm consists of two parts: the map compass operator and the landmark operator. The map compass operator simulates the influence of the sun and the earth's magnetic field, two navigation tools, on pigeons. The update strategy for the speed and position information of the j-th pigeon in the pigeon swarm in the s-th generation is as follows:

[0041] V j (s+1)=e -Rt V j (s)+r u [(G gb (s)-G(s)];

[0042] G j (s+1)=G j (s)+V j (s+1);

[0043] Where R is the compass operator, ranging from 0 to 1, V j (s) is the speed information of the pigeon, G j (s) is the location information of the pigeon, r u Represents a random number, G gb (s) is the global optimal position of all pigeons in the pigeon flock;

[0044] When the number of iterations reaches a certain value, the map compass operator stops working and enters the iterative process of the landmark operator to continue working. The number of pigeons is halved in each iteration, and the position update strategy is:

[0045] N p (s)=N p (s-1) / 2;

[0046]

[0047] G i (s)=G i (s-1)+r u (G C (s)-G i (s-1);

[0048] Where N p is the current number of pigeons, F(G i (s) is the adaptive function, G C (s) is the center of the pigeon flock and is used as a landmark, r u is a random number between 0 and 1;

[0049] The quantized pigeon flock optimization algorithm introduces quantum representation of real numbers and quantum rotation gates. A quantum can be represented by the states of "0" and "1". The quantum bit state is represented as:

[0050]

[0051] Where, α i and β i Represent the linear probability of the two states, and satisfy

[0052] When the algorithm is initialized, α i and β i The corresponding probability amplitudes are If the global optimal solution changes after the iteration, then α i and β i Updated with the following quantum turnstile strategy:

[0053]

[0054]

[0055] In the formula, Δθ is the rotation angle, and ε is the constraint parameter that prevents the quantum from being trapped in the states "0" and "1".

[0056] In an optional solution, step S4 uses an adaptive KNN algorithm to determine the number of neighborhood point clouds for each target point during feature extraction and analysis to improve the accuracy of target point normal vector calculation, specifically comprising the following steps:

[0057] According to the maximum and minimum density distribution of the acquired three-dimensional point cloud, the maximum value K of the neighborhood K value is determined max and minimum value K min , so the value range of K is [K min ,K max ], the calculation formula of characteristic index entropy E is as follows:

[0058] E=λ L ln L+λ P ln P+λ S ln S;

[0059] Where λ L ,λ P and λ S are the weight coefficients of linear, planar and scattered features respectively, L, P and S are the corresponding eigenvalues ​​respectively, and the maximum value E of the characteristic index entropy E is calculated by traversing the K value range max , the optimal neighborhood K value can be determined, and then the normal vector calculation can be performed;

[0060] After the adaptive KNN search, the neighborhood of each target point is tested for coplanarity. Through the coplanarity test, points with high coplanarity in the neighborhood are screened out to eliminate the influence of interference on the normal vector calculation. The coplanarity is expressed by the deviation parameter (η), and the calculation formula is:

[0061]

[0062] Where λ1, λ2, and λ3 are the covariance characteristic roots of the target point neighborhood point cloud data. When the deviation value is less than the threshold η max When , the neighborhood of the target point is considered to be in the same plane; otherwise, the target point will be excluded;

[0063] After determining the coplanar points through coplanarity detection, the corresponding optimal plane is fitted using the coplanar points to calculate the normal vector of the corresponding target point. For the point cloud that meets the coplanarity requirements, the fitting plane equation is set as:

[0064] ax+by+cz+d=0;

[0065] The plane parameters a, b and c are the components of the unit normal vector corresponding to the plane in the directions of the coordinate axes, satisfying a 2 +b 2 +c 2 =1, d is the distance from the plane to the origin.

[0066] In an optional solution, step S4 further includes the following steps:

[0067] In order to make the fitted plane the best plane, the sum of the squares of the distances from the K points in the neighborhood point cloud set {P} to the plane must be minimized. In order to obtain the best plane, the sum of the squares of the distances from the neighborhood points to the plane must be minimized:

[0068]

[0069] Where, d i is any point p in the neighborhood point cloud set {P} m (x m ,y m ,z m ) to the plane distance, calculate the unit eigenvector μ3 corresponding to the minimum eigenvalue λ3 through the eigenvalue and eigenvector as the plane parameters a, b, c and normal vector;

[0070] The slope of each point is calculated using the normal vector. The slope θ is calculated using the z-direction component of the normal vector:

[0071]

[0072] Where n x 、n y and nz are the x, y, and z components of the normal vector;

[0073] The slope information obtained by normal vector calculation is integrated with the slope stability index to form a multidimensional feature data set.

[0074] In an optional solution, when performing the slope stability zoning evaluation in step S5, a SAGA-FCM cluster analysis algorithm is used to perform cluster analysis on the characteristic data, specifically comprising the following steps:

[0075] 1) Initialize control parameters: given the population size S, the maximum number of evolutions G max , the maximum number of iterations L max , mutation probability P m , initial annealing temperature T0, temperature cooling coefficient q, crossover probability P c , selection probability P a , termination temperature T end and the termination tolerance ΔJ of the objective function;

[0076] 2) Initialize the population: Initialize k cluster centers to generate the initial population C. For each cluster center, calculate the initial membership of each sample according to the following formula:

[0077]

[0078] Where n is the total number of point cloud normal vector data, u ji is the membership degree of the i-th point cloud to the j-th cluster center, the value range is [0, 1], m is the weighting parameter, take m = 2;

[0079] Calculate the objective function value J corresponding to the initial population individual according to the following formula: I , where I = 1, 2, 3, ..., S;

[0080]

[0081] Where, d 2 (X i ,V j ) is the similarity distance between the normal vector of the i-th point cloud and the j-th cluster center;

[0082] 3) Set the initial evolution times gen = 0;

[0083] 4) Genetic algorithm: Select, cross, and mutate the old individuals in the initial population C to generate new individuals. Calculate the k cluster centers and the membership of each sample for the generated new individuals using formula (1) and formula (3), and calculate the objective function value J' corresponding to the new population individuals according to formula (2) I ;

[0084]

[0085] Where x is the previous solution; f(x) is the objective function corresponding to x; f(x') is the objective function corresponding to x';

[0086] 5) Metropolis criterion: If the objective function J' I <J I , then a new individual is selected to replace the old individual, otherwise, the probability P calculated by the following formula accepts the new individual;

[0087]

[0088] Where PC is the separation degree, k is the total number of cluster centers, and cluster centers V = {V1,…,V j ,…,V k};

[0089] 6) Genetic algorithm termination condition: If the number of evolutions gen <G max , then gen=gen+1, go to step 3), otherwise continue to the next step;

[0090] 7) Termination condition of simulated annealing algorithm: If the current temperature Ti <T end , then the annealing process is completed and the current optimal cluster center V is output best Otherwise, continue the annealing process, Ti = qTi, and go to step 4);

[0091] According to the FCM clustering results, the slope area is divided into different stability levels.

[0092] In an optional solution, the result display and application performed in step S6 includes the following steps:

[0093] Use visualization tools to display the slope stability zoning results as clear stability zoning maps, showing the different stability areas of the slope, helping users to intuitively understand the slope stability distribution;

[0094] Generate detailed analysis reports, including zoning maps, stability evaluation results, and data analysis conclusions, to provide comprehensive support for decision-making; establish a dynamic monitoring system to obtain updated slope data through regular repeated drone flights; track changes in slope stability in real time, and automatically generate alarm notifications when potential risks are discovered to ensure timely action.

[0095] The beneficial effects of this application are:

[0096] The photovoltaic land slope stability zoning evaluation method in this application uses drones to collect images and point cloud data of slope areas. Drones can cover a large area in a short period of time and quickly collect slope images and point cloud data, greatly improving the efficiency of data collection. Then, the slope stability is evaluated by zoning after image and point cloud data preprocessing, three-dimensional modeling and digital elevation model generation, feature extraction and analysis. Such a zoning evaluation method can realize efficient and accurate stability detection and evaluation of photovoltaic land slopes, providing stronger protection for the safety of photovoltaic land.

[0097] It should be understood that the foregoing general description and the following detailed description are merely illustrative and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 The overall flow chart of the photovoltaic land slope stability zoning evaluation method provided in this application;

[0099] Figure 2 Flowchart of the SAGA-FCM cluster analysis algorithm provided in this application.

[0100] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application. DETAILED DESCRIPTION

[0101] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0102] It should be clear that the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other technical solutions obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0103] The terms used in the embodiments of the present application are for the purpose of describing specific embodiments only and are not intended to limit the present application. The singular forms "a", "an", "the" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise.

[0104] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0105] like Figure 1-2As shown, the embodiment of the present application provides a method for evaluating the stability of photovoltaic land slopes by zoning, which mainly includes the following steps:

[0106] S1. Data Collection: The drone conducts oblique photography flights based on a path planning algorithm, collecting images and point cloud data from the sloped area. The drone is equipped with a high-resolution camera and lidar to capture high-definition images and point cloud data. During flight, an adaptive flight path planning algorithm is used to plan the route, dynamically adjusting the drone's flight altitude and angle based on terrain characteristics to optimize data collection efficiency and image quality.

[0107] Specifically, this method adopts the adaptive dung beetle algorithm for UAV flight path planning and introduces the Tent chaotic mapping as a method to improve the diversity of the dung beetle algorithm initialization population. The tent chaotic mapping can generate an initial population that is evenly distributed in the solution space and has strong correlation, thereby improving the solution accuracy and convergence speed of the intelligent algorithm.

[0108] First, initialize the population: first randomly generate a D-dimensional vector of [0,1] as the initial chaotic sequence; then substitute each dimension of the D-dimensional vector into the following formula in turn:

[0109]

[0110] A new D-dimensional vector is calculated and generated as the second chaotic sequence. The above steps are repeated until N chaotic sequences are generated. The generated chaotic sequences are mapped to the parameter range of the UAV's flight path to form the initial dung beetle population.

[0111] Then calculate the population diversity by cosine similarity:

[0112]

[0113] When Diver is greater than 0.5, the population diversity is too low, potentially leading to a local optimum. The number of ball-rolling and foraging dung beetles determines the algorithm's ability to explore the solution space and its convergence speed. Therefore, when diversity falls below the threshold, the algorithm's global search capability is enhanced, and individuals are mutated into alternative behavioral strategies to improve search efficiency.

[0114] This method uses a reverse learning strategy to enhance the global search capabilities of reproduction and stealing behaviors:

[0115]

[0116]

[0117] Where, For the reverse solution, lb and ub are D-dimensional vectors representing the lower and upper bounds of each dimension, and rand() is a D-dimensional random vector. The current feasible solution is generated by generating the reverse solution of the current solution and comparing its fitness with the original solution to select the better solution and enhance the global search capability.

[0118] S2. Image and point cloud data preprocessing: Correct the collected oblique images, including distortion removal, noise processing, image stitching, etc.; denoise, filter and stitch the point cloud data.

[0119] This method uses PhotoScan software, a Russian-made photogrammetry software package used to automatically generate orthophotos from static images, to process drone-captured data. The collected point cloud data undergoes preprocessing, including denoising, filtering, and point cloud stitching.

[0120] This method uses the clustering method of outlier processing to perform denoising. First, the number of points selected in the surrounding area k and the standard deviation multiplier a are determined. s ; For point p in the point cloud data i (i=1,……,N), N represents the number of points in the point cloud; determine its k nearest neighbor points s j (j=1,……,k), and calculate the k nearest points and p i The average distance d between i ;

[0121]

[0122] Calculate the average distance d of all points in the point cloud i The expected μ d and standard deviation σ d and discrimination threshold; according to the expected μ d and standard deviation σ d Calculate the discrimination threshold T:

[0123] T=μ d +α S ·σ d

[0124] Finally, the threshold is less than the average distance d in the point cloud i The points are eliminated.

[0125] The filtering process uses the mean curvature flow filtering algorithm, which is based on grid denoising. It moves the vertex coordinates along the tangent and normal directions of the point coordinates at the speed of the curvature mean, thereby removing noise points in the point cloud data. Specifically:

[0126]

[0127] Where G(L) is the mean curvature and n(L) is the direction of the external normal vector of the point. The final calculation formula is:

[0128] L t+1 =L t +λG(L)n(L)

[0129] Let the vertex be V i , then the corresponding updates of the vertices are as follows:

[0130]

[0131] The point cloud data stitching uses rigid body transformation to unify the coordinate system. Using the unique point cloud data stitching method in Realworks, the scanning target is matched to complete the stitching of cloud data from different measuring stations. The implementation process is as follows:

[0132] First, feature points selected at the overlapped locations of adjacent stations are matched, and the resulting error values ​​and ratio plots are observed during iterations. The accuracy requirements are determined based on the magnitude of the registered error values. If the error exceeds 3 cm, re-registration is performed until the accuracy standard is met. The remaining station cloud data is then looped through the above steps, and the matched point cloud data is unified to complete the registration of all station points.

[0133] S3. 3D modeling and digital elevation model (DEM) generation: Combine the texture information of the image and the height data of the point cloud to create a realistic 3D terrain model, and extract the terrain height information from the point cloud data to generate a digital elevation model (DEM).

[0134] The texture information of the image is integrated with the height data of the point cloud to create a high-precision 3D terrain model. Texture mapping technology is used to map the texture information of the image data onto the 3D mesh generated by the point cloud data, ensuring the realism and detailed representation of the model. Improved multi-resolution fusion algorithms (such as deep fusion technology based on convolutional neural networks) are used to generate a detailed and accurate 3D terrain model. The DEM provides detailed height information of slopes, facilitating the analysis of slope and aspect.

[0135] In addition, this method adopts an improved interpolation algorithm: the Kriging interpolation algorithm based on quantized pigeon flock optimization to improve the accuracy and detail of DEM.

[0136] The basic pigeon swarm optimization algorithm consists of two parts: the map compass operator and the landmark operator. The map compass operator simulates the influence of the sun and the earth's magnetic field, two navigation tools, on pigeons. The update strategy for the speed and position information of the j-th pigeon in the pigeon swarm in the s-th generation is as follows:

[0137] V j (s+1)=e -Rt Vj (s)+r u [(G gb (s)-G(s)]

[0138] G j (s+1)=G j (s)+V j (s+1)

[0139] Where R is the compass operator, and its value range is 0 to 1; V j (s) is the speed information of the pigeon; G j (s) is the location information of the pigeon; r u Represents a random number; G gb (s) is the global optimal position of all pigeons in the pigeon flock.

[0140] When the number of iterations reaches a certain value, the map compass operator stops working and enters the iterative process of the landmark operator to continue working. The number of pigeons is halved in each iteration, and the position update strategy is:

[0141] N p (s)=N p (s-1) / 2

[0142]

[0143] G i (s)=G i (s-1)+r u (G C (s)-G i (s-1)

[0144] Where N p is the current number of pigeons; F(G i (s) is the adaptability function; G C (s) is the center of the pigeon flock and is used as a landmark; r u is a random number between 0 and 1.

[0145] The quantized pigeon flock optimization algorithm introduces quantum representation of real numbers and quantum rotation gates. A quantum can be represented by the states of "0" and "1". The quantum bit state is represented as:

[0146]

[0147] Where: α i and β i Represent the linear probability of the two states, and satisfy

[0148] When the algorithm is initialized, α i and β iThe corresponding probability amplitudes are If the global optimal solution changes after the iteration, then α i and β i Updated with the following quantum turnstile strategy:

[0149]

[0150] Where: Δθ is the rotation angle; ε is the constraint parameter that prevents the quantum from being trapped in states "0" and "1".

[0151] S4. Feature extraction and analysis: Extract features from the 3D model and analyze the stability indicators of the slope, such as slope, soil type, vegetation cover, etc.

[0152] The adaptive KNN algorithm is used to calculate the normal vector of the point cloud to improve the accuracy of feature extraction. To improve the accuracy of the target point normal vector calculation, the adaptive KNN algorithm is used to determine the number of neighborhood point clouds for each target point. The maximum value K of the neighborhood K value is determined based on the maximum and minimum density distribution of the acquired three-dimensional point cloud. max and minimum value K min , so the value range of K is [K min ,K max ]. The calculation formula of characteristic index entropy E is as follows:

[0153] E=λ L ln L+λ P ln P+λ S ln S

[0154] Where λ L ,λ P and λ S are the weight coefficients of linear, planar and scattered features respectively, and L, P and S are the corresponding eigenvalues ​​respectively. By traversing the K value range, the maximum value E of the characteristic index entropy E is calculated. max , the optimal neighborhood K value can be determined, and then the normal vector calculation can be performed.

[0155] After the adaptive KNN search, it is necessary to perform a coplanarity test on the neighborhood of each target point. This test can filter out points with high coplanarity in the neighborhood and eliminate the influence of interference on the normal vector calculation. Coplanarity is represented by the deviation parameter (η), and the calculation formula is:

[0156]

[0157] Among them, λ1, λ2, and λ3 are the covariance characteristic roots of the target point neighborhood point cloud data. When the deviation value is less than the threshold η max , the neighborhood of the target point is considered to be in the same plane; otherwise, the target point will be excluded.

[0158] After determining the coplanar points through coplanarity detection, the corresponding optimal plane is fitted using the coplanar points to calculate the normal vector of the corresponding target point. For the point cloud that meets the coplanarity requirements, the fitting plane equation is set as:

[0159] ax+by+cz+d=0

[0160] The plane parameters a, b and c are the components of the unit normal vector corresponding to the plane in the directions of the coordinate axes, satisfying a 2 +b 2 +c 2 =1; d is the distance from the plane to the origin.

[0161] At this time, in order to make the fitted plane the best plane, it is necessary to satisfy the minimum sum of the squares of the distances from the K points in the neighborhood point cloud set {P} to the plane. In order to obtain the best plane, it is necessary to minimize the sum of the squares of the distances from the neighborhood points to the plane:

[0162]

[0163] Among them, d i is any point p in the neighborhood point cloud set {P} m (x m ,y m ,z m ) to the plane distance, and calculate the unit eigenvector μ3 corresponding to the minimum eigenvalue λ3 through the eigenvalue and eigenvector as the plane parameters a, b, c and normal vector.

[0164] The slope of each point can be calculated using the normal vector. The slope θ can be calculated using the z-direction component of the normal vector:

[0165]

[0166] Among them, n x 、n y and n z are the x, y, and z components of the normal vector.

[0167] Vegetation classification: If the point cloud data contains color or reflectance information, use these attributes to identify vegetation areas. If not, it is necessary to combine other data sources such as remote sensing imagery. Soil type analysis: Fusion of point cloud data with existing soil information data (such as GIS data) to identify and label soil type areas.

[0168] The slope information obtained by normal vector calculation is integrated with other feature data (such as vegetation cover and soil type) to form a multidimensional feature dataset.

[0169] S5. Slope stability zoning evaluation: Establish a slope stability evaluation model for zoning evaluation, and conduct stability zoning evaluation based on characteristic data and clustering results, and optimize the zoning standards.

[0170] The data is classified and partitioned, and the characteristic data of the slope is clustered and analyzed by SAGA-FCM to identify the stability level area. The SAGA-FCM cluster analysis algorithm is used to cluster the characteristic data, which can converge to the global optimal solution more effectively and quickly. The specific process is as follows (e.g. Figure 2 shown):

[0171] 1) Initialize control parameters: given the population size S, the maximum number of evolutions G max , the maximum number of iterations L max , mutation probability P m , initial annealing temperature T0, temperature cooling coefficient q, crossover probability P c , selection probability P a , termination temperature T end and the termination tolerance ΔJ of the objective function.

[0172] 2) Initialize the population: Initialize k cluster centers to generate the initial population C. For each cluster center, calculate the initial membership of each sample according to the following formula:

[0173]

[0174] Where n is the total number of point cloud normal vector data; u ji is the membership degree of the i-th point cloud to the j-th cluster center, and its value range is [0, 1]; m is the weighting parameter, usually m = 2.

[0175] Calculate the objective function value J corresponding to the initial population individual according to the following formula: I , where I = 1, 2, 3,…, S.

[0176]

[0177] Where, d 2 (X i ,V j ) is the similarity distance between the normal vector of the i-th point cloud and the j-th cluster center.

[0178] 3) Set the initial evolution times gen=0.

[0179] 4) Genetic algorithm: Select, cross, and mutate the old individuals in the initial population C to generate new individuals. Calculate the k cluster centers and the membership of each sample for the generated new individuals using formula (1) and formula (3), and calculate the objective function value J' corresponding to the new population individuals according to formula (2) I .

[0180]

[0181] Where x is the previous solution; f(x) is the objective function corresponding to x; and f(x') is the objective function corresponding to x'.

[0182] 5) Metropolis criterion: If the objective function J' I <J I , then a new individual is selected to replace the old individual, otherwise, the probability P calculated by the following formula accepts the new individual.

[0183]

[0184] Where PC is the separation degree; k is the total number of cluster centers, and cluster centers V = {V1,…,V j ,…,V k};

[0185] 6) Genetic algorithm termination condition: If the number of evolutions gen <G max , then gen=gen+1, go to step 3), otherwise continue to the next step.

[0186] 7) Termination condition of simulated annealing algorithm: If the current temperature Ti <T end , then the annealing process is completed and the current optimal cluster center V is output best , otherwise continue the annealing process, Ti=qTi, go to step 4).

[0187] According to the FCM clustering results, the slope area is divided into different stability level areas, such as high stability area, potential unstable area and unstable area.

[0188] S6. Results presentation and application: Visualize the slope stability zoning results, generate stability zoning maps and reports, and establish a dynamic monitoring system. Regularly use drones for repeated flights to obtain updated slope data.

[0189] Advanced visualization tools are used to present slope stability zoning results as clear stability zoning maps, which display the different stability zones of the slope and help users intuitively understand the slope stability distribution. A detailed analysis report is generated, including the zoning map, stability assessment results, and data analysis conclusions, to provide comprehensive support for decision-making. A dynamic monitoring system is also established, using regular drone flights to obtain updated slope data. The monitoring system automatically processes new data and updates the stability zoning results. The monitoring system includes data acquisition, processing, and analysis modules, tracking changes in slope stability in real time and automatically generating alerts when potential risks are identified, ensuring timely action.

[0190] The zoning method for evaluating the stability of photovoltaic land slopes in this application uses drones to collect imagery and point cloud data of slope areas. Drones can cover large areas in a short period of time, rapidly collecting slope images and point cloud data, significantly improving data collection efficiency. Using drone oblique photography and lidar point cloud technology, high-resolution imagery and detailed topographic data of photovoltaic land slopes are obtained. The digital elevation model (DEM) provides accurate terrain height information, enabling more precise calculations of slope gradient and aspect.

[0191] This zoning assessment method establishes a dynamic monitoring system that regularly updates slope data, monitors slope changes in real time, and promptly adjusts stability zones to effectively manage long-term risks. Using high-precision digital models for slope stability analysis, it can identify potential problems early, reducing subsequent risk management and remediation costs. A detailed stability assessment report, including risk levels and recommended measures for each area, provides a scientific and comprehensive reference for decision makers. Through scientific slope stability analysis, the planning and management of photovoltaic land use can be optimized, supporting the green development and sustainable use of photovoltaic projects.

[0192] The above are merely preferred embodiments of the present application and are not intended to limit the present application. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for evaluating the stability of photovoltaic land slopes by zoning, characterized in that: The following steps are involved: S1. Data collection: The UAV performs oblique photography flight based on the path planning algorithm to collect images and point cloud data of the slope area; S2. Image and point cloud data preprocessing: Correct the collected images, including distortion removal, noise processing, and image stitching; perform denoising, filtering, and point cloud stitching on the point cloud data; S3, 3D modeling and digital elevation model generation: Combine the texture information of the image and the height data of the point cloud to create a realistic 3D terrain model, and extract the terrain height information from the point cloud data to generate a digital elevation model; S4. Feature extraction and analysis: Extract features from the 3D terrain model and analyze slope stability indicators, including slope, soil type, and vegetation cover; S5. Slope stability zoning evaluation: Establish a slope stability evaluation model and conduct zoning evaluation. Perform stability zoning evaluation based on characteristic data and clustering results, and optimize zoning criteria. S6. Results presentation and application: Visualize the slope stability zoning results, generate stability zoning maps and reports, and establish a dynamic monitoring system. Regularly use drones for repeated flights to obtain updated slope data.

2. The photovoltaic land slope stability zoning evaluation method according to claim 1 is characterized in that: In step S1, the path planning algorithm of the UAV includes the following steps: Initialize the population: first randomly generate a D-dimensional vector of [0,1] as the initial chaotic sequence, and then substitute each dimension of the D-dimensional vector into the following formula in turn: Calculate and generate a new D-dimensional vector as the second chaotic sequence. Repeat the above steps until N chaotic sequences are generated. Map the generated N chaotic sequences to the parameter range of the UAV flight path to form the initial dung beetle population. Then calculate the population diversity by cosine similarity: When Diver is greater than 0.5, the population diversity is too low and may fall into a local optimum. The number of rolling dung beetles and foraging dung beetles determines the algorithm's ability to explore the solution space and the speed of convergence. Therefore, when the diversity is lower than the threshold, the algorithm's global search ability is enhanced and individuals are mutated into other behavioral strategies to improve search efficiency. Using reverse learning strategies to enhance the global search capabilities of reproduction and stealing behaviors: Where, For the reverse solution, lb and ub are D-dimensional vectors representing the lower and upper bounds of each dimension, and rand() is a D-dimensional random vector. is the current feasible solution; By generating the reverse solution of the current solution and comparing its fitness with the original solution, a better solution can be selected to enhance the global search capability.

3. The photovoltaic land slope stability zoning evaluation method according to claim 1 or 2, characterized in that: In step S2, the image captured by the drone is corrected using PhotoScan software. The denoising of the point cloud data includes the following steps: First, determine the number of points k in the surrounding area and the standard deviation multiplier a s , for point p in the point cloud data i (i=1,……,N), N represents the number of points in the point cloud, and determines its k nearest neighbor points s j (j=1,……,k), and calculate the average distance d between these k nearest points and p i : Calculate the average distance d of all points in the point cloud i The expected μ d and standard deviation σ d and the discrimination threshold, according to the expected μ d and standard deviation σ d Calculate the discrimination threshold T: T=μ d +a S ·s d ; Finally, the threshold is less than the average distance d in the point cloud i The points are eliminated.

4. The photovoltaic land slope stability zoning evaluation method according to claim 3 is characterized in that: In step S2, the point cloud data is filtered using a mean curvature flow filtering algorithm, which includes the following steps: Where G(L) is the mean curvature and n(L) is the direction of the external normal vector of the point. The final calculation formula is: L t+1 =L t +λG(L)n(L); Let the vertex be V i , then the corresponding updates of the vertices are as follows: Point cloud stitching of the point cloud data includes the following steps: First, match the feature points selected at the overlap of adjacent measuring stations and observe the error value and ratio diagram during iteration. The error value after registration is used to determine whether the accuracy requirements are met. If the error value exceeds 3cm, re-registration should be performed until the accuracy standard is met. The remaining station cloud data are cyclically processed according to the above steps, and the matched point cloud data are unified together to complete the registration of all station points.

5. The photovoltaic land slope stability zoning evaluation method according to claim 1, 2 or 4, characterized in that: The step S3 further comprises the following steps: The Kriging interpolation algorithm based on quantized pigeon flock optimization improves the accuracy and detail of DEM: The basic pigeon swarm optimization algorithm consists of two parts: the map compass operator and the landmark operator. The map compass operator simulates the influence of the sun and the earth's magnetic field, two navigation tools, on pigeons. The update strategy for the speed and position information of the j-th pigeon in the pigeon swarm in the s-th generation is as follows: V j (s+1)=e -Rt V j (s)+r u [(G gb (s)-G(s)]; G j (s+1)=G j (s)+V j (s+1); Where R is the compass operator, ranging from 0 to 1, V j (s) is the speed information of the pigeon, G j (s) is the location information of the pigeon, r u Represents a random number, G gb (s) is the global optimal position of all pigeons in the pigeon flock; When the number of iterations reaches a certain value, the map compass operator stops working and enters the iterative process of the landmark operator to continue working. The number of pigeons is halved in each iteration, and the position update strategy is: N p (s)=N p (s-1) / 2; G i (s)=G i (s)-1)+r u (G C (s)-G i (s-1); Where N p is the current number of pigeons, F(G i (s) is the adaptive function, G C (s) is the center of the pigeon flock and is used as a landmark, r u is a random number between 0 and 1; The quantized pigeon flock optimization algorithm introduces quantum representation of real numbers and quantum rotation gates. A quantum can be represented by "0" and "1" states, and the quantum bit state is represented as: Where, α i and β i Represent the linear probability of the two states, and satisfy When the algorithm is initialized, α i and β i The corresponding probability amplitudes are If the global optimal solution changes after the iteration, then α i and β i Updated with the following quantum turnstile strategy: Where Δθ is the rotation angle and ε is the constraint parameter that prevents the quantum from being trapped in states "0" and "1".

6. The photovoltaic land slope stability zoning evaluation method according to claim 1, 2 or 4, characterized in that: The step S4 uses an adaptive KNN algorithm to determine the number of neighborhood point clouds for each target point during feature extraction and analysis to improve the accuracy of the target point normal vector calculation, specifically including the following steps: According to the maximum and minimum density distribution of the acquired three-dimensional point cloud, the maximum value K of the neighborhood K value is determined max and minimum value K min , so the value range of K is [K min ,K max ], the calculation formula of characteristic index entropy E is as follows: E=λ L lnL+λ P lnP+λ S lnS; Where λ L ,λ P and λ S are the weight coefficients of linear, planar and scattered features respectively, L, P and S are the corresponding eigenvalues ​​respectively, and the maximum value E of the characteristic index entropy E is calculated by traversing the K value range max , the optimal neighborhood K value can be determined, and then the normal vector calculation can be performed; After the adaptive KNN search, the neighborhood of each target point is tested for coplanarity. Through the coplanarity test, points with high coplanarity in the neighborhood are screened out to eliminate the influence of interference on the normal vector calculation. The coplanarity is expressed by the deviation parameter (η), and the calculation formula is: Where λ1, λ2, and λ3 are the covariance characteristic roots of the target point neighborhood point cloud data. When the deviation value is less than the threshold η max When , the neighborhood of the target point is considered to be in the same plane; otherwise, the target point will be excluded; After determining the coplanar points through coplanarity detection, the corresponding optimal plane is fitted using the coplanar points to calculate the normal vector of the corresponding target point. For the point cloud that meets the coplanarity requirements, the fitting plane equation is set as: ax+by+cz+d=0; The plane parameters a, b and c are the components of the unit normal vector corresponding to the plane in the directions of the coordinate axes, satisfying a 2 +b 2 +c 2 =1, d is the distance from the plane to the origin.

7. The photovoltaic land slope stability zoning evaluation method according to claim 6, characterized in that: The step S4 further comprises the following steps: In order to make the fitted plane the best plane, the sum of the squares of the distances from the K points in the neighborhood point cloud set {P} to the plane must be minimized. In order to obtain the best plane, the sum of the squares of the distances from the neighborhood points to the plane must be minimized: Where, d i is any point p in the neighborhood point cloud set {P} m (x m ,y m ,z m ) to the plane distance, calculate the unit eigenvector μ3 corresponding to the minimum eigenvalue λ3 through the eigenvalue and eigenvector as the plane parameters a, b, c and normal vector; The slope of each point is calculated using the normal vector. The slope θ is calculated using the z-direction component of the normal vector: Where n x 、n y and n z are the x, y, and z components of the normal vector; The slope information obtained by normal vector calculation is integrated with the slope stability index to form a multidimensional feature data set.

8. The photovoltaic land slope stability zoning evaluation method according to any one of claims 1-2 or 4 or 7, characterized in that: When performing the slope stability zoning evaluation in step S5, the SAGA-FCM cluster analysis algorithm is used to perform cluster analysis on the characteristic data, which specifically includes the following steps: 1) Initialize control parameters: given the population size S, the maximum number of evolutions G max , the maximum number of iterations L max , mutation probability P m , initial annealing temperature T0, temperature cooling coefficient q, crossover probability P c , selection probability P a , termination temperature T end and the termination tolerance ΔJ of the objective function; 2) Initialize the population: Initialize k cluster centers to generate the initial population C. For each cluster center, calculate the initial membership of each sample according to the following formula: Where n is the total number of point cloud normal vector data, u ji is the membership degree of the i-th point cloud to the j-th cluster center, the value range is [0, 1], m is the weighting parameter, take m = 2; Calculate the objective function value J corresponding to the initial population individual according to the following formula: I , where I = 1, 2, 3, ..., S; Where, d 2 (X i ,V j ) is the similarity distance between the normal vector of the i-th point cloud and the j-th cluster center; 3) Set the initial evolution times gen = 0; 4) Genetic algorithm: Select, cross, and mutate the old individuals in the initial population C to generate new individuals. Calculate the k cluster centers and the membership of each sample for the generated new individuals using formula (1) and formula (3), and calculate the objective function value J' corresponding to the new population individuals according to formula (2) I ; Where x is the previous solution; f(x) is the objective function corresponding to x; f(x') is the objective function corresponding to x'; 5) Metropolis criterion: If the objective function J' I <J I , then a new individual is selected to replace the old individual, otherwise, the probability P calculated by the following formula accepts the new individual; Where PC is the separation degree, k is the total number of cluster centers, and cluster centers V = {V1,…,V j ,…,V k }; 6) Genetic algorithm termination condition: If the number of evolution gem <G max , then gen=gen+1, go to step 3), otherwise continue to the next step; 7) Termination condition of simulated annealing algorithm: If the current temperature Ti <T end , then the annealing process is completed and the current optimal cluster center V is output best Otherwise, continue the annealing process, Ti = qTi, and go to step 4); According to the FCM clustering results, the slope area is divided into different stability levels.

9. The photovoltaic land slope stability zoning evaluation method according to claim 8, characterized in that: The result display and application in step S6 includes the following steps: Use visualization tools to display the slope stability zoning results as clear stability zoning maps, showing the different stability areas of the slope, helping users to intuitively understand the slope stability distribution; Generate detailed analysis reports, including zoning maps, stability evaluation results, and data analysis conclusions, to provide comprehensive support for decision-making; establish a dynamic monitoring system to obtain updated slope data through regular repeated drone flights; track changes in slope stability in real time, and automatically generate alarm notifications when potential risks are discovered to ensure timely action.

Images