Rock discontinuous surface identification method based on NRLC enhanced two-stage DBSCAN clustering

Abstract

The invention discloses an NRLC enhanced two-stage DBSCAN clustering rock discontinuous surface identification method, and relates to the field of rock mass structural surface intelligent identification, and the method comprises the steps: S1, identifying concave, convex and boundary feature points in a rock point cloud; s2, calculating a rock point cloud normal vector based on local fitting; s3, based on the rock point cloud normal vector, identifying a group discontinuous surface through DBSCAN clustering; s4, DBSCAN clustering is carried out based on the group discontinuous surfaces to obtain all discontinuous surfaces in the point clouds, automatic recognition and accurate representation of the rock mass discontinuous surfaces are achieved through normal vector clustering analysis, a set of efficient and accurate rock mass structural surface intelligent recognition technology system is established, and reliable three-dimensional data support is provided for rock mass stability evaluation.

Description

Technical Field

[0001] This application relates to the field of intelligent recognition of rock mass structural surfaces, and in particular to a rock discontinuity surface recognition method using NRLC-enhanced two-stage DBSCAN clustering. Background Technology

[0002] Accurate identification and characterization of rock mass discontinuities are of great significance for rock mass stability evaluation and engineering geological analysis. Traditional methods for measuring discontinuities, such as compass measurement, line surveying, and photogrammetry, typically rely on manual on-site sampling and two-dimensional planar data analysis.

[0003] While these methods offer some reliability in terms of local accuracy, they suffer from limitations such as low efficiency, limited sampling range, and difficulty in acquiring three-dimensional spatial information. Three-dimensional laser scanning technology can rapidly acquire high-precision point cloud data of rock surfaces, providing a new technical means for discontinuity surface identification. However, this method still faces many challenges in practical applications: First, clustering performance is highly sensitive to hyperparameters and easily affected by noise, especially at the boundaries between discontinuous surfaces. Shared boundaries and surface roughness often pose significant challenges to accurate identification. Shared boundaries between discontinuities integrate the geometric features of two surfaces, which usually leads to abrupt changes in normal direction and curvature, resulting in a large number of noise clusters. Surface irregularities also negatively affect the accuracy of plane fitting.

[0004] Secondly, the massive point cloud data generated on the surface of complex rock masses has high requirements for computing resources, which increases the threshold for data processing. Finally, the surface roughness and occlusion effect of the rock mass can affect the quality of the point cloud, making it difficult to extract structural surface features.

[0005] In recent years, point cloud normal vector-based analysis methods have demonstrated advantages in discontinuity identification due to their sensitivity to local geometric features. Compared to traditional two-dimensional profile analysis, three-dimensional point cloud normal vector clustering methods can accurately identify the spatial distribution characteristics of discontinuities, providing more comprehensive data support for the three-dimensional characterization of rock mass structures. Summary of the Invention

[0006] The purpose of this application is to provide a rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering. A highly efficient and fully automated method for plane detection in rock mass point clouds is proposed, using 3D laser point cloud data. Before clustering, boundary points, concave points, and convex points in the point cloud are identified and initially removed. After clustering, boundary points are reassigned to corresponding discontinuities based on spatial proximity. Concave and convex points that negatively impact plane fitting accuracy are completely excluded from the analysis. Then, discontinuities are identified using a two-stage clustering method.

[0007] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering, characterized in that the method comprises: S1. Identify concave, convex, and boundary feature points in rock point clouds; S2. Calculate the normal vector of the rock point cloud based on local fitting; S3. Identify group discontinuities using DBSCAN clustering based on rock point cloud normal vectors; S4. Perform DBSCAN clustering based on group discontinuities to obtain all discontinuities in the point cloud.

[0008] Furthermore, the identification of concave, convex, and boundary feature points in the rock point cloud in step S1 also includes: S101. Obtain the local centroid of the current point and its neighborhood; S102. Calculate the distance weights of the current point and its neighborhood. ; S103. Calculate the shape change weights of the current point and its neighborhood. ; S104. Based on distance weights and shape change weight Optimize the feature vector of the current point's neighborhood to obtain the second-level feature vector of the current point's neighborhood; S105. Based on the secondary feature vectors of all points in the point cloud, the points in the rock point cloud are classified to obtain multiple types of feature points.

[0009] Furthermore, the calculation of the rock point cloud normal vector based on local fitting also includes: S201. Obtain the K neighboring points of the current point in the point cloud; By combining the current point with the local plane formed by the obtained K neighboring points, a fitted plane is obtained; ; in, Let the normal vector and the distance to the origin be... Defined plane, The unit normal vector of the plane. The sign distance from the plane to the origin of the coordinate system. For the first The coordinate vectors of the adjacent points; S202. Obtain the covariance matrix T of the point set formed by the point and its neighboring points based on principal component analysis, and obtain the normal vector of the plane P based on the covariance matrix T of the point set formed by the point and its neighboring points.

[0010] Furthermore, the step of obtaining the normal vector of plane P based on the covariance matrix T of the point set composed of the point and its neighboring points also includes: The covariance matrix T can be represented as follows: ; ; Eigenvalue decomposition of the coequation matrix T yields the three eigenvalues ​​of M. , , and the corresponding feature vectors , and ; By comparing the obtained eigenvalues, we obtain the eigenvector corresponding to the smallest eigenvalue, which is the normal vector of the current point.

[0011] Furthermore, the step S3, which identifies group discontinuities based on rock point cloud normal vectors using DBSCAN clustering, also includes: Based on the distribution of normal vectors of all points in the acquired point cloud in 3D space, DBSCAN clustering is performed on the normal vector space to obtain group discontinuities.

[0012] Furthermore, S4, which involves using DBSCAN clustering based on group discontinuities to obtain all discontinuities in the point cloud, also includes: Based on DBSCAN clustering, individual discontinuities in the group discontinuities are obtained. All group discontinuities are clustered separately to obtain all discontinuities in the point cloud.

[0013] This application provides a rock discontinuity identification method using NRLC-enhanced two-stage DBSCAN clustering, which has the following technical advantages compared with existing technologies: Improved identification efficiency: Compared with traditional manual measurement methods such as compass measurement and line surveying, this method uses point cloud normal clustering technology, which can quickly process a large amount of point cloud data, significantly improving the identification efficiency of rock mass discontinuities and reducing the time and workload of manual measurement.

[0014] Obtaining more comprehensive spatial information: Traditional manual measurement methods have limited sampling range and are difficult to obtain three-dimensional spatial information. However, this method is based on three-dimensional point cloud data, which can comprehensively and three-dimensionally analyze the rock mass structure, provide more complete data support, and help to more accurately understand the mechanical properties and stability of the rock mass.

[0015] Improving fitting accuracy: By removing concave, convex, and boundary points from the point cloud, the interference of these special points on the fitting of discontinuities can be reduced, making the fitting of discontinuities more accurate, thereby improving the accuracy of describing the geometric features of rock mass discontinuities.

[0016] Reduced computational complexity: The combination of normal vector estimation algorithm and cluster analysis method effectively reduces the computational complexity of point cloud data on complex rock mass surfaces, enabling the method to maintain high computational efficiency when processing large-scale point cloud data.

[0017] Hierarchical identification of complex structures: A dual clustering strategy is adopted. First, the point cloud normal vectors are clustered to identify group discontinuities. Then, the group discontinuities are clustered to further subdivide individual discontinuities. This method can efficiently extract the complex discontinuous areas of rocks with a simple approach, realizing hierarchical identification of discontinuities in complex rock structures and improving the accuracy and reliability of identification. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart for obtaining group discontinuities in one embodiment of this application; Figure 2 This is an image showing the result of feature point extraction from a rock point cloud provided in an embodiment of this application. Figure 3 This is a schematic diagram of group discontinuity surface extraction provided in an embodiment of this application; Figure 4 The image shows the individual discontinuity surface identification result and the filtered noise point map provided in one embodiment of this application. Figure 5 A diagram showing the annotation and identification results of a reference data discontinuity surface provided in an embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] like Figure 1 As shown in the embodiment of this application, a rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering is provided, comprising: S1. Identify concave, convex, and boundary feature points in rock point clouds; S2. Calculate the normal vector of the rock point cloud based on local fitting; S3. Identify group discontinuities using DBSCAN clustering based on rock point cloud normal vectors; S4. Perform DBSCAN clustering based on group discontinuities to obtain all discontinuities in the point cloud.

[0023] Furthermore, the identification of concave, convex, and boundary feature points in the rock point cloud in step S1 also includes: S101. Obtain the local centroid of the current point and its neighborhood; Specifically, obtain the local centroids of the current point and its neighborhood. The local centroid of the i-th point in the point cloud is represented as: ; In the formula, This represents a feature vector containing information about curvature and boundaries. Decompose along the normal and parallel to the normal, and into components perpendicular to and parallel to the normal vector. and It can be represented as ; Suppose the neighborhood of the i-th point in the point cloud is , This represents the vector constructed from the neighboring points and the current point, and the sum of each neighboring vector is calculated. The components along the normal direction and perpendicular to the normal direction are the eigenvectors. The two components can be represented as ; For the first neighboring points and points The resulting vectors are parallel; For the first neighboring points and points The perpendicular to the point The component of the normal vector direction.

[0024] To enable the feature vectors to better reflect the local features of a point, two weights—distance weight and shape change weight—are introduced to scale the vectors, thereby increasing the differences in feature vectors for points with different characteristics.

[0025] S102. Calculate the distance weights of the current point and its neighborhood. ; Specifically, calculate the distance weights. Distance weights are determined by a factor... The length is defined as a function, and appropriate methods are used to exaggerate the differences in eigenvectors between different features, at the current point. With neighboring points The distance weight is defined as ; ; ; As can be seen from the above formula, The value of is between 1 and 2.

[0026] S103. Calculate the shape change weights of the current point and its neighborhood. ; Specifically, calculate the shape change weights. . Defined as a switch weight, taking into account the current point With neighboring points Local morphological changes, ; In the formula, It is an adjustable amplification factor; express and Estimation of local shape changes; A threshold indicates the extent to which two local surfaces are considered to be the same shape. Through two local standard deviations and To perform the calculation, the formula is as follows: ; ; In the formula, express The neighborhood point set, Neighboring points The normal vector, It is the average normal vector of the neighborhood points. Solution process and Consistent.

[0027] S104. Based on distance weights and shape change weight Optimize the feature vector of the current point's neighborhood to obtain the second-level feature vector of the current point's neighborhood; Specifically, in obtaining vectors of and Then, weights are applied to its two components, and all neighborhood vectors are traversed, finally... The two components of the eigenvector can be represented as ; Final point The eigenvectors are calculated as follows: ; S105. Based on the secondary feature vectors of all points in the point cloud, the points in the rock point cloud are classified to obtain multiple types of feature points.

[0028] Specifically, through the above calculations, the feature vectors of all points in the point cloud are obtained. Since feature points only account for a small portion of all points, and the feature vectors are exaggerated by two weighting coefficients, a larger feature vector length indicates a more drastic shape change at that point. Therefore, the 80th percentile of the feature vector length distribution is used as a threshold to determine whether each point is a key feature point. The number of feature points can be increased or decreased by adjusting the percentile of the threshold according to actual needs. If the length... Greater than the threshold Then the point These are considered as candidate feature points and forwarded to the second stage of feature recognition.

[0029] Four indicators were constructed to partition candidate features, among which Indicates quantities related to the concave point. Indicates quantities related to convexity. Indicates quantities related to the boundary. These represent non-characteristic quantities. The four indicator quantities can be represented as: ; ; ; ; In the formula, Point The angle between the vector and the normal vector; It is a custom separation factor, set to 2. The total of the four indicators. As a measure of probability, the four probability categories are defined as follows: ; Candidate features The probability of belonging to a category will form a set. The type with the highest probability is selected as the candidate. The feature labels are: ; After the above feature classification, the candidate points are divided into four sets: concave, convex, boundary, and featureless points.

[0030] like Figure 2 The image shows the extraction results of concave, convex, and boundary points of the rock point cloud.

[0031] Furthermore, the calculation of the rock point cloud normal vector based on local fitting also includes: S201. Obtain the K neighboring points of the current point in the point cloud; By combining the current point with the local plane formed by the obtained K neighboring points, a fitted plane is obtained; ; in, Let the normal vector and the distance to the origin be... Defined plane, The unit normal vector of the plane. The sign distance from the plane to the origin of the coordinate system. For the first The coordinate vectors of adjacent points, where k takes 8 to 10 values; S202. Obtain the covariance matrix T of the point set formed by the point and its neighboring points based on principal component analysis, and obtain the normal vector of the plane P based on the covariance matrix T of the point set formed by the point and its neighboring points.

[0032] Furthermore, the method of obtaining the normal vector of plane P based on the covariance matrix T of the point set composed of the point and its neighboring points further includes: The covariance matrix T can be represented as follows: ; ; Eigenvalue decomposition of the coequation matrix T yields the three eigenvalues ​​of M. , , and the corresponding feature vectors , and ; By comparing the obtained eigenvalues, we obtain the eigenvector corresponding to the smallest eigenvalue, which is the normal vector of the current point.

[0033] Furthermore, the step S3, which identifies group discontinuities based on rock point cloud normal vectors using DBSCAN clustering, also includes: Based on the distribution of normal vectors of all points in the acquired point cloud in 3D space, DBSCAN clustering is performed on the normal vector space to obtain group discontinuities.

[0034] Specifically, the flowchart for obtaining group discontinuous surfaces is as follows: Figure 2 As shown.

[0035] After obtaining the normal vectors of all points, the distribution of the normal vectors in 3D space can be obtained by taking the three components of the normal vector as the three coordinates of the spatial points. The point cloud of the group discontinuity surface has a high degree of consistency in the normal direction, so its distribution in space is relatively concentrated. Based on this characteristic, we perform DBSCAN clustering on the normal vector space, so that the normal vectors are nearly consistent and form a group discontinuity surface. The specific clustering process is as follows: DBSCAN is a typical density clustering algorithm that uses neighborhood radius and number of points to describe the density of samples. Before clustering, the following concepts need to be clarified: A core point is any point p in space whose neighborhood contains only k samples; density reachability means that if point Q is located in the neighborhood of p and p is a core point, then Q is density reachable from p; density accessibility means that in the sample sequence p1, p2, p3, p2 is density reachable from p1, p3 is density reachable from p2, and p3 is density reachable from p1; density connectivity means that for spatial points P and Q, if there exists a core point T such that both P and Q are density reachable from T, then P and Q are density connected.

[0036] (1) Select any point from the dataset and determine whether it is a core point.

[0037] (2) If it is a core point, start a new cluster, add the point and all the points in its neighborhood to the cluster, and continuously expand the neighborhood of the new core point.

[0038] (3) If a point is neither in any cluster nor meets the conditions to be a core point, then it is marked as a noise point.

[0039] (4) Continue to check all unvisited points until all points have been visited.

[0040] Figure 3 The results of identifying discontinuous surfaces are shown in the figure, compared with the results without feature point extraction. As can be seen from the figure, even without feature point filtering, there is still a large amount of noise interference.

[0041] Furthermore, S4, which involves using DBSCAN clustering based on group discontinuities to obtain all discontinuities in the point cloud, also includes: Based on DBSCAN clustering, individual discontinuities in the group discontinuities are obtained. All group discontinuities are clustered separately to obtain all discontinuities in the point cloud.

[0042] Specifically, after obtaining the group discontinuities, since group discontinuities with the same orientation differ in spatial location, individual discontinuities within the group discontinuities are obtained through DBSCAN clustering in S2. The clustering results are as follows: Figure 4As shown, all discontinuous surfaces in the group are clustered separately, and all discontinuous surfaces in the point cloud are identified. Concave and convex points, boundary points, and noise points after DBSCAN clustering are not included in the final result, thus filtering out all corner points and noise points.

[0043] To evaluate the accuracy of the results, this invention extracts discontinuities from rock point clouds that are commonly used as benchmark data and have been measured and commented on by experts. Figure 5 As shown in Table 1, the orientation angles of the extracted discontinuous surfaces were calculated and compared with the results from the commercial software CloudCompare.

[0044] Table 1 shows the comparison results with CloudCompare (tendency / inclination angle). ))

[0045] Based on DBSCAN clustering, individual discontinuities in the group discontinuities are obtained. All group discontinuities are clustered separately to obtain all discontinuities in the point cloud.

[0046] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0047] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0048] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0049] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0050] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0051] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0052] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering, characterized in that, The method includes: S1. Identify concave, convex, and boundary feature points in rock point clouds; S2. Calculate the normal vector of the rock point cloud based on local fitting; S3. Identify group discontinuities using DBSCAN clustering based on rock point cloud normal vectors; S4. Perform DBSCAN clustering based on group discontinuities to obtain all discontinuities in the point cloud.

2. The rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering according to claim 1, characterized in that, The process of identifying concave, convex, and boundary feature points in the rock point cloud in S1 also includes: S101. Obtain the local centroid of the current point and its neighborhood; S102. Calculate the distance weights of the current point and its neighborhood. ; S103. Calculate the shape change weights of the current point and its neighborhood. ; S104. Based on distance weights and shape change weight Optimize the feature vector of the current point's neighborhood to obtain the second-level feature vector of the current point's neighborhood; S105. Based on the secondary feature vectors of all points in the point cloud, the points in the rock point cloud are classified to obtain multiple types of feature points.

3. The rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering according to claim 1, characterized in that, The method for calculating the rock point cloud normal vector based on local fitting also includes: S201. Obtain the K neighboring points of the current point in the point cloud; By combining the current point with the local plane formed by the obtained K neighboring points, a fitted plane is obtained; ； in, Let be a plane defined by the normal vector and the distance d to the origin. Let be the plane unit normal vector, d be the signed distance from the plane to the origin, and pi be the coordinate vector of the i-th adjacent point; S202. Obtain the covariance matrix T of the point set formed by the point and its neighboring points based on principal component analysis, and obtain the normal vector of the plane P based on the covariance matrix T of the point set formed by the point and its neighboring points.

4. The rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering according to claim 3, characterized in that, The method of obtaining the normal vector of plane P based on the covariance matrix T of the point set composed of the point and its neighboring points also includes: The covariance matrix T can be represented as follows: ； ； Eigenvalue decomposition of the coequation matrix T yields the three eigenvalues ​​of M. , , and the corresponding feature vectors , and ; By comparing the obtained eigenvalues, we obtain the eigenvector corresponding to the smallest eigenvalue, which is the normal vector of the current point.

5. The rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering according to claim 4, characterized in that, The S3 method for identifying group discontinuities based on rock point cloud normal vectors using DBSCAN clustering also includes: Based on the distribution of normal vectors of all points in the acquired point cloud in 3D space, DBSCAN clustering is performed on the normal vector space to obtain group discontinuities.

6. The rock discontinuity identification method based on NRLC-enhanced two-stage DBSCAN clustering according to claim 5, characterized in that, The step S4, which involves using DBSCAN clustering based on group discontinuities to obtain all discontinuities in the point cloud, also includes: Based on DBSCAN clustering, individual discontinuities in the group discontinuities are obtained. All group discontinuities are clustered separately to obtain all discontinuities in the point cloud.

Images