Structural surface identification method based on rock mass outcrop point cloud data

Abstract

The invention provides a structural surface identification method based on rock mass outcrop point cloud data, and relates to the technical field of rock mass engineering, and the method comprises the steps: obtaining the rock mass outcrop point cloud data, carrying out the partitioning of the rock mass outcrop point cloud data based on an octree principle, obtaining points in each partition, traversing the points in each partition, and carrying out the recognition of the structural surface of the rock mass outcrop point cloud data. The method comprises the steps of obtaining a normal vector of each point based on principal component analysis, determining a clustering center of the normal vectors based on kernel density estimation, classifying the normal vectors of the points to clustering centers based on the clustering centers of the normal vectors by adopting a fuzzy C mean value method, and obtaining point cloud data forming each group of structural surfaces by representing each group of structural surfaces. Secondary clustering is carried out on the point cloud data of the structural plane based on the spatial density algorithm, the boundary of the structural plane is recognized, the problem that existing structural plane recognition is low in efficiency is solved, and the method is suitable for structural plane recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of rock mass engineering, and particularly to a method for identifying structural planes based on rock mass outcrop point cloud data. Background Art

[0002] The geomechanical behavior of rock masses is determined by the overall structure of structural planes and intact rocks. Obtaining information about structural planes is very important for understanding the distribution of structural planes within rock masses and analyzing the deformation and instability of rock masses.

[0003] The existing extraction of information about structural planes requires directly measuring each part of the structural plane using a tape measure and a compass. This method requires a large amount of manpower, material resources, and time, and has low efficiency. At the same time, this method also faces many limitations. For example, in the face of high-steep rocky slopes that are inaccessible to surveyors, it is difficult to carry out the interpretation work of structural planes. Such slopes of rock masses often have high disaster risks.

[0004] Unmanned aerial vehicle (UAV) photogrammetry technology can be used as a non-contact photogrammetry technology to obtain three-dimensional slope surface information from the terrain with high precision and high spatial resolution. Using non-contact measurement methods can more easily obtain point clouds with three-dimensional information on rock masses than traditional methods. Therefore, combining UAV photogrammetry to obtain point clouds and carrying out rock mass structure analysis based on this has good prospects. Summary of the Invention

[0005] The technical problem to be solved by the present invention: Provide a method for identifying structural planes based on rock mass outcrop point cloud data to solve the problem of low efficiency in existing structural plane identification.

[0006] The technical solution adopted by the present invention to solve the above technical problem: A method for identifying structural planes based on rock mass outcrop point cloud data, comprising the following steps:

[0007] S1. Obtain rock mass outcrop point cloud data;

[0008] S2. Based on the octree principle, partition the rock mass outcrop point cloud data to obtain the points in each partition;

[0009] S3. Traverse the points in each partition and obtain the normal vector of each point based on principal component analysis;

[0010] S4. Determine the clustering center of the normal vectors based on kernel density estimation;

[0011] S5. Based on the normal vector clustering center, use the fuzzy C-means method to classify the normal vectors of the points to the clustering center. Each clustering center represents a set of structural planes, thereby obtaining the point cloud data constituting each set of structural planes;

[0012] S6. Perform secondary clustering on the point cloud data of the structural plane based on the spatial density algorithm to identify the boundaries of the structural plane.

[0013] Further, in S1, obtaining the point cloud data of the rock mass outcrop includes the following steps:

[0014] S11. Use a drone to preliminarily photograph the area to be studied to obtain the basic engineering geological conditions of the area to be studied, and the basic geological engineering conditions include the rock mass outcrop;

[0015] S12. Use the drone lens to be perpendicular to the rock mass outcrop to collect point cloud data and obtain the point cloud data of the rock mass outcrop.

[0016] Further, in S12, it also includes bilateral filtering optimization of the point cloud data of the rock mass outcrop.

[0017] Further, in S2, the number of points in each partition after partitioning is between the first threshold and the second threshold, and the average distance from the points in each partition to the fitting plane of the points in the partition is lower than the coplanarity threshold.

[0018] Further, in S3, obtaining the normal vector of each point based on principal component analysis includes the following steps:

[0019] S31. Set a fixed radius, with the target point as the center of the sphere, and find all the points inside the sphere to form a neighborhood;

[0020] S32. Calculate the covariance matrix of the neighborhood points;

[0021] S33. Perform eigenvalue decomposition on the covariance matrix, and the eigenvector corresponding to the minimum eigenvalue is the normal vector of the target point.

[0022] Further, in S4, determining the clustering center of the normal vector based on kernel density estimation includes the following steps:

[0023] S41. Smooth the normal vector of each point through a kernel function to obtain a density function;

[0024] S42. Take the normal vector corresponding to the local maximum of the density function as the clustering center of the normal vector.

[0025] Further, the density function is: where f(x) represents the density corresponding to the target point, n represents the number of normal vectors, h represents the bandwidth, K represents the kernel function, x represents the normal vector of the target point, and x i represents the normal vector of the i-th point.

[0026] Further, in S5, it includes the following steps:

[0027] S51. Optimize the clustering centers and membership degrees by minimizing the objective function, where the objective function is:

[0028] where J represents the value of the objective function, A represents the number of normal vectors, B represents the number of clustering centers of normal vectors, u ab represents the membership degree of the a-th normal vector to the b-th clustering center, m represents the fuzzy factor, and ||x a - P b || represents the Euclidean distance from the a-th normal vector to the b-th clustering center;

[0029] S52. Update the membership degrees according to the membership degree update formula, where the membership degree update formula is:

[0030] where represents the weighted sum of the relative distances of the a-th normal vector to all clustering centers;

[0031] S53. Update the clustering centers according to the clustering center update formula, where the clustering center update formula is: where P b represents the b-th clustering center, represents the weighted sum of all normal vectors to the b-th clustering center, represents the weighted sum of the membership degrees of all normal vectors to the b-th clustering center;

[0032] S54. Repeat S51 to S53 until the objective function converges.

[0033] Further, in S6, the minimum number of neighboring points of the clustering center in the spatial density algorithm is 4, and the radius is the average value of the distances of 4 neighboring points plus the standard deviation.

[0034] Advantages of the present invention: The present invention provides a method for identifying structural planes based on rock outcrop point cloud data. By acquiring rock outcrop point cloud data, partitioning the rock outcrop point cloud data based on the octree principle to obtain the points in each partition, traversing the points in each partition, obtaining the normal vector of each point based on principal component analysis, determining the clustering center of the normal vectors based on kernel density estimation, and using the fuzzy C-means method based on the normal vector clustering center to classify the normal vectors of the points to the clustering centers, where each clustering center represents a set of structural planes, so as to obtain the point cloud data constituting each structural plane, and performing secondary clustering on the point cloud data of the structural planes based on the spatial density algorithm to identify the structural plane boundaries, solving the problem of low efficiency of existing structural plane identification. Description of the Drawings

[0035] Figure 1 is a schematic flow chart of a method for identifying structural planes based on rock outcrop point cloud data provided by the present invention. Detailed implementation mode

[0036] In view of the problem of low efficiency in obtaining existing structural planes, the present invention provides a method for identifying structural planes based on rock mass outcrop point cloud data, as Figure 1 shown, including the following steps:

[0037] S1. Obtain the rock mass outcrop point cloud data.

[0038] Specifically, obtaining the rock mass outcrop point cloud data includes the following steps:

[0039] S11. Use a drone to preliminarily photograph the area to be studied, and obtain the basic engineering geological conditions of the area to be studied. The basic geological engineering conditions include rock mass outcrops; the basic engineering geological conditions also include topography, vegetation, and engineering buildings.

[0040] S12. Use the drone lens to be perpendicular to the rock mass outcrop for point cloud data acquisition, and obtain the point cloud data of the rock mass outcrop.

[0041] Specifically, determine the drone flight mission through the rock mass outcrop to ensure that the drone lens is perpendicular to the rock mass outcrop for point cloud data acquisition, and determine the flight height of the drone when acquiring outcrop point cloud data according to the height of the rock mass outcrop. To improve the accuracy of the point cloud data, it also includes bilateral filtering optimization of the rock mass outcrop point cloud data. Bilateral filtering combines the information in the spatial domain and the value domain, identifies the noise points in the point cloud from two angles and retains the edge features. In the spatial domain, mainly consider the distance between point clouds; in the value domain, compare the attribute value differences between different points, and use the RGB value of the point cloud as the value domain parameter.

[0042] S2. Based on the octree principle, partition the rock mass outcrop point cloud data to obtain the points in each partition.

[0043] Specifically, divide the three-dimensional point cloud data into equal-sized cubes. If the number of points in the cube is between the first threshold and the second threshold, then take this cube as a partition. If the number of points in the cube is lower than the first threshold, then delete the points in this cube. If the number of points in the cube is higher than the second threshold, then divide this cube into 8 equal-sized sub-cubes until the number of points in each sub-cube is between the first threshold and the second threshold, that is, the number of points in each partition is between the first threshold and the second threshold. To ensure the accuracy of the structural plane, it is also necessary to perform a coplanarity test on the points in each partition. Specifically: the average distance from the points in each partition to the fitting plane of the points in the partition is lower than the coplanarity threshold.

[0044] S3. Traverse the points in each partition, and obtain the normal vector of each point based on principal component analysis.

[0045] Specifically, the normal vector of each point is obtained based on principal component analysis, including the following steps:

[0046] S31. Set a fixed radius, take the target point as the center of the sphere, and find all the points inside the sphere to form a neighborhood.

[0047] S32. Calculate the covariance matrix of the neighborhood points.

[0048] S33. Perform eigenvalue decomposition on the covariance matrix, and the eigenvector corresponding to the smallest eigenvalue is the normal vector of the target point.

[0049] In this way, the normal vector of each point in each partition is obtained.

[0050] S4. Determine the clustering center of the normal vectors based on kernel density estimation.

[0051] Specifically, determining the clustering center of the normal vectors based on kernel density estimation includes the following steps:

[0052] S41. Smooth the normal vector of each point through a kernel function to obtain a density function, and the density function is: where f(x) represents the density corresponding to the target point, n represents the number of normal vectors, h represents the bandwidth, K represents the kernel function, x represents the normal vector of the target point, and x i represents the normal vector of the i-th point.

[0053] S42. Take the normal vector corresponding to the local maximum of the density function as the clustering center of the normal vectors.

[0054] Specifically, for the normal vector that is the clustering center of the normal vectors, normalize it for easy calculation.

[0055] S5. Based on the normal vector clustering center, use the fuzzy C-means method to classify the normal vectors of the points into the clustering centers. Each clustering center represents a set of structural planes, thereby obtaining the point cloud data constituting each set of structural planes.

[0056] Specifically, it includes the following steps:

[0057] S51. Optimize the clustering center and membership degree by minimizing the objective function, and the objective function is: where J represents the value of the objective function, A represents the number of normal vectors, B represents the number of normal vector clustering centers, u ab represents the membership degree of the a-th normal vector to the b-th clustering center, m represents the fuzzy factor, and ||x a -P b || represents the Euclidean distance from the a-th normal vector to the b-th clustering center.

[0058] S52. Update the membership degree according to the membership degree update formula, and the membership degree update formula is as follows: Wherein, represents the weighted sum of the relative distances of the a-th normal vector to all cluster centers.

[0059] S53. Update the cluster centers according to the cluster center update formula, and the cluster center update formula is as follows: Where P b represents the b-th cluster center, represents the weighted sum of all normal vectors to the b-th cluster center, represents the weighted sum of the membership degrees of all normal vectors to the b-th cluster center.

[0060] S54. Repeat S51 to S53 until the objective function converges.

[0061] Through the above steps, the points in each cluster center are obtained, and each cluster center corresponds to a set of structural planes, so as to obtain the point cloud data constituting each set of structural planes.

[0062] S6. Perform secondary clustering on the point cloud data of the structural planes based on the spatial density algorithm to identify the boundaries of the structural planes.

[0063] Specifically, the minimum number of neighboring points of the cluster center in the spatial density algorithm is 4, and the radius is the average value of the distances of 4 neighboring points plus the standard deviation.

Claims

1. A method for identifying structural planes based on rock outcrop point cloud data, characterized in that, It includes the following steps: S1. Obtain the point cloud data of the rock mass outcrop; S2. Based on the octree principle, partition the point cloud data of the rock mass outcrop to obtain the points in each partition; S3. Traverse the points in each partition and obtain the normal vector of each point based on principal component analysis; S4. Determine the clustering center of the normal vectors based on kernel density estimation; S5. Based on the normal vector clustering center, use the fuzzy C-means method to classify the normal vectors of the points into the clustering centers. Each clustering center represents a set of structural planes, thereby obtaining the point cloud data constituting each set of structural planes; S6. Perform secondary clustering on the point cloud data of the structural planes based on the spatial density algorithm to identify the boundaries of the structural planes.

2. The structural plane recognition method based on rock mass outcrop point cloud data according to claim 1, wherein In S1, to obtain the point cloud data of the rock mass outcrop, it includes the following steps: S11. Use a drone to preliminarily photograph the area to be studied to obtain the basic engineering geological conditions of the area to be studied. The basic geological engineering conditions include the rock mass outcrop; S12. Use the drone camera to be perpendicular to the rock mass outcrop for point cloud data acquisition to obtain the point cloud data of the rock mass outcrop.

3. The structural plane recognition method based on the rock outcrop point cloud data according to claim 2, wherein, In S12, it also includes bilateral filtering optimization of the point cloud data of the rock mass outcrop.

4. The structural plane recognition method based on rock mass outcrop point cloud data according to claim 1, characterized in that In S2, the number of points in each partition is between the first threshold and the second threshold, and the average distance from the points in each partition to the fitting plane of the points in the partition is lower than the coplanarity threshold.

5. The structural plane recognition method based on rock mass outcrop point cloud data according to claim 1, characterized in that, In S3, to obtain the normal vector of each point based on principal component analysis, it includes the following steps: S31. Set a fixed radius, with the target point as the center of the sphere, and find all the points inside the sphere to form a neighborhood; S32. Calculate the covariance matrix of the neighborhood points; S33. Perform eigenvalue decomposition on the covariance matrix, and the eigenvector corresponding to the minimum eigenvalue is the normal vector of the target point.

6. The structural plane recognition method based on rock outcrop point cloud data according to claim 1, characterized in that, In S4, to determine the clustering center of the normal vectors based on kernel density estimation, it includes the following steps: S41. Smooth the normal vector of each point through a kernel function to obtain a density function; S42. Take the normal vector corresponding to the local maximum of the density function as the clustering center of the normal vectors.

7. The structural plane recognition method based on rock outcrop point cloud data according to claim 6, characterized in that The density function is as follows: where f(x) represents the density corresponding to the target point, n represents the number of normal vectors, h represents the bandwidth, K represents the kernel function, x represents the normal vector of the target point, and x i represents the normal vector of the i-th point.

8. The structural plane recognition method based on rock mass outcrop point cloud data according to claim 1, characterized in that, In S5, it includes the following steps: S51. Optimize the cluster centers and membership degrees by minimizing the objective function, where the objective function is: where J represents the objective function value, A represents the number of normal vectors, B represents the number of cluster centers of normal vectors, u ab represents the membership degree of the a-th normal vector to the b-th cluster center, m represents the fuzzy factor, ||x a -P b || represents the Euclidean distance from the a-th normal vector to the b-th cluster center; S52. Update the membership degree according to the membership degree update formula, and the membership degree update formula is as follows: wherein, represents the weighted sum of the relative distances of the a-th normal vector to all cluster centers; S53. Update the cluster center according to the cluster center update formula, where the cluster center update formula is: where P b represents the b-th cluster center, represents the weighted sum of all normal vectors for the b-th cluster center, represents the weighted sum of membership degrees of all normal vectors for the b-th cluster center; S54. Repeat S51 to S53 until the objective function converges.

9. The structural plane recognition method based on rock mass outcrop point cloud data according to claim 1, characterized in that In S6, the minimum number of neighboring points of the clustering center in the spatial density algorithm is 4, and the radius is the average value of the distances of 4 neighboring points plus the standard deviation.

Images