A method for identifying rock mass structural surfaces and extracting boundary surface trace length based on fuzzy improvement

Abstract

The present invention relates to the technical field of intelligent identification and extraction of rock mass structures, and in particular to a method for identifying rock mass structural surfaces and extracting boundary surface trace lengths based on fuzziness improvement, comprising: collecting three-dimensional point cloud data of rock mass in a study area, and calculating the normal vector of the three-dimensional point cloud data; clustering analysis of the normal vector based on a C-means algorithm with improved fuzziness value to obtain dominant three-dimensional point cloud data; extracting geometric features of structural surfaces and turning points of boundary surfaces based on the dominant three-dimensional point cloud data, and then calculating the boundary surface trace length, completing the identification of rock mass structural surfaces in the study area and the extraction of boundary surface trace lengths. The present invention can realize accurate identification of structural surfaces, as well as accurate measurement and statistics of structural surface trace lengths.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent identification and extraction of rock mass structures, and in particular to a method for identifying rock mass structure surfaces and extracting boundary surface trace length based on fuzziness improvement. Background Art

[0002] In terms of geological disaster prevention and mitigation, rock mass structural surfaces are key controlling factors in the formation and development of geological disasters such as collapse and landslides. Identifying rock mass structural surfaces and extracting their geometric features are important aspects of slope stability calculation. With the development of low-altitude photogrammetry and 3D laser scanning technologies, it has become possible to quickly and efficiently obtain high-precision 3D geometric data of steep slopes. How to scientifically and accurately identify and extract rock mass structural data from 3D slope geometric data has become an urgent problem for engineering geology researchers and engineers. To this end, predecessors have conducted a lot of exploration and application. For example, Riquelme et al. used the K nearest neighbor algorithm and kernel density estimation method to develop and design a Matlab program for semi-automatic extraction of discontinuous structural surfaces of rock mass, clustering and extracting dominant structural surfaces; Li Shuiqing et al. used drone photogrammetry and the three-point method to semi-automatically extract the occurrence of structural surfaces, and classified and obtained parameters of dominant structural surfaces, providing basic data for analyzing slope stability. However, this process still requires manual selection of structural surface feature points; Xiu Guolin et al. used fuzzy C-means clustering (FCM) and graphics technology to identify and extract the occurrence of structural surfaces; Dong Xiujun et al. used the K-means clustering method based on the points and data of three-dimensional laser scanning to identify and extract information about rock mass structures. Previous studies have found that compared with hard clustering algorithms such as K-means, the FCM algorithm can achieve better clustering effects when clustering and grouping dominant structural surfaces because it introduces membership. However, when applying the FCM algorithm, although everyone knows that the value of its key parameter fuzziness value m has a great influence on the clustering effect, the value of m is still mainly empirically determined, which is difficult to dynamically update and determine according to the iterative process, thus greatly affecting the refinement and accuracy of the clustering results. In addition, after the rock mass structural surface is identified, its geometric data such as trace length needs to be extracted. Previous researchers mainly extracted the geometric data information of the structural surface directly based on the point cloud data of the structural surface identification process. However, due to the unevenness caused by the weathering of the structural surface surface, the interference of vegetation on the slope surface, and the uncertainty of the data acquisition process, many point cloud data outside the structural surface are difficult to be effectively identified and eliminated, and participate in the extraction process of the structural surface geometric data. In this case, the geometric data extracted based on the point cloud data has large uncertainty and error, which directly affects the accuracy of the prediction of rock mass stability.

[0003] Therefore, there is an urgent need for rock mass structural surface identification and boundary surface trace length extraction methods based on fuzzy improvement. Summary of the invention

[0004] The purpose of the present invention is to provide a rock mass structural surface identification and boundary surface trace length extraction method based on fuzziness improvement, aiming at the problem of rock mass structural surface identification deviation caused by fuzziness value in the application of FCM algorithm and inaccuracy in extracting structural surface trace length based on point cloud data. In the FCM algorithm, the fuzziness of the algorithm is adaptively adjusted by introducing new parameters, the robustness of the algorithm is enhanced, and the structural surface clustering analysis effect is optimized; at the same time, a method for automatically extracting structural surface trace length by generating boundary surfaces is constructed to achieve accurate measurement and statistics of structural surface trace length.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The rock mass structural surface recognition and boundary surface trace length extraction method based on fuzzy improvement includes:

[0007] Collecting three-dimensional point cloud data of rock mass in the study area, and calculating normal vectors of the three-dimensional point cloud data;

[0008] Performing cluster analysis on the normal vectors based on the C-means algorithm with improved fuzzy value to obtain dominant three-dimensional point cloud data;

[0009] Based on the advantageous three-dimensional point cloud data, the geometric features of the structural surface and the turning points of the boundary surface are extracted, and then the boundary surface trace length is calculated to complete the identification of the rock mass structural surface in the study area and the extraction of the boundary surface trace length.

[0010] Optionally, collecting three-dimensional point cloud data of rock mass in the study area includes:

[0011] Collect rock mass images in the study area by using an image acquisition device mounted on a drone;

[0012] A three-dimensional point cloud model is constructed based on the rock mass image, and then the three-dimensional point cloud data is obtained.

[0013] Optionally, a random Hough transform algorithm is used to calculate the normal vector of the three-dimensional point cloud data.

[0014] Optionally, the C-means algorithm for improving fuzzy value selection is: dynamically adjusting fuzzy value selection by introducing parameters into the fuzzy C-means clustering algorithm.

[0015] Optionally, a method for dynamically adjusting the fuzziness value by introducing parameters into the fuzzy C-means clustering algorithm is as follows:

[0016] Calculate the index of membership fuzziness degree;

[0017] The adjustment of the fuzziness value is judged based on the indicator. When the indicator exceeds a preset threshold, the fuzziness value is reduced; when the indicator is lower than or equal to the threshold, the fuzziness value is increased.

[0018] Optionally, the calculation method of the index of the membership fuzziness degree is:

[0019] ex=max(u ij 2 )-min(u ij 2 )

[0020] Among them, ex represents the index of the fuzzy degree of membership; u ij Indicates the membership of the i-th sample in the membership matrix U to the j-th cluster.

[0021] Optionally, the method for judging the adjustment of the ambiguity value based on the indicator is:

[0022]

[0023] Among them, m represents the fuzzy value, and α is the introduced parameter.

[0024] Optionally, extracting geometric features of structural surfaces and extracting turning points of boundary surfaces based on the dominant three-dimensional point cloud data includes:

[0025] The advantageous three-dimensional point cloud data is closed into a surface using a Fast Marching algorithm to obtain a boundary surface;

[0026] The coordinates of the turning points of the boundary surface are extracted based on the boundary surface by using a graphics processing technique.

[0027] Optionally, the method for calculating the boundary surface trace length is:

[0028]

[0029] Among them, D pq represents the boundary surface trace length, (X p ,Y p ,Z p ) represents the coordinates of the turning point p on the boundary surface, p = 1, 2, 3, ..., s-1; (X q ,Y q ,Z q ) represents the coordinates of the turning point q on the boundary surface, q = p+1, p+2,…,s.

[0030] The beneficial effects of the present invention are:

[0031] The present invention proposes an improved FCM algorithm, which introduces new parameters to adaptively adjust the fuzziness of the algorithm, enhance the robustness of the algorithm, and optimize the clustering effect. Based on the Fast-Marching algorithm and combined with ArcGIS software, an automatic extraction method for boundary surface trace length parameters is constructed to realize automatic measurement and statistics of rock structure surface trace length. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 This is a flow chart of a method for rock mass structural surface identification and boundary surface trace length extraction based on fuzzy improvement according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] This embodiment provides a method for identifying rock mass structural surfaces and extracting boundary surface trace length based on fuzziness improvement, such as Figure 1 As shown, including:

[0037] Step 1: Obtain 3D point cloud data;

[0038] The rock mass images in the study area are collected by using an image acquisition device mounted on a drone, and a three-dimensional point cloud model is constructed based on the rock mass images to obtain three-dimensional point cloud data.

[0039] Specifically, the research area is surveyed in detail through UAV aerial photogrammetry, LiDAR 3D scanning, drilling and other methods to obtain high-precision dense point cloud data of the target. The UAV aerial photogrammetry method mainly involves the collection of flight design data, aerial photography and process modeling. In the photogrammetry work, the topographic and geomorphic characteristics of the survey area, as well as the performance and product scale of the digital camera are comprehensively considered to plan the flight path. In order to meet the data resolution requirements, the heading overlap ratio of the aerial photography drone is greater than 80%, and the lateral overlap ratio is greater than 60%. After acquiring the image, the modeling software is used to make and process the three-dimensional model, process the result image, and complete the three-dimensional point cloud model production.

[0040] Step 2: Calculate the point cloud normal vector;

[0041] Traditionally, the reconstruction of the three-dimensional structural model of the rock mass is done by calculating the point cloud data of a certain triangulation principle to generate a TIN grid, but the overall efficiency of this process is low. In order to solve this problem, this embodiment uses cluster analysis technology to directly extract information from the structural surface point cloud data. If you want to directly use the point cloud to extract the characteristic information of the slope rock structure surface, you must first calculate the normal vector of the point cloud data. The normal vector information helps to study its spatial tilt and inclination characteristics, thereby optimizing the extraction rate of the rock structure surface characteristic information.

[0042] The randomized Hough transform algorithm proposed by Boulch and Marlet is used to calculate the normal vector. This method has a robust normal vector evaluation capability and can effectively handle noise points, outliers, and sharp corners. It also coordinates the balance between accuracy, speed, and robustness in the normal vector information generation process.

[0043] Step 3: Obtain the dominant 3D point cloud data based on the fuzzy C-means algorithm improved by fuzzy value to realize structural surface clustering;

[0044] The fuzzy value is dynamically adjusted by introducing parameters in the fuzzy C-means clustering algorithm. The dynamic adjustment method is as follows: calculate the index of the membership fuzziness degree; judge the adjustment of the fuzziness value based on the index. When the index exceeds the preset threshold, reduce the fuzziness value; when the index is lower than or equal to the threshold, increase the fuzziness value.

[0045] Specifically, after completing the normal vector calculation of each point and obtaining the normal vector information of the structural surface point cloud data, it is necessary to perform cluster analysis on the normal vectors of the structural surface point cloud data to find the dominant three-dimensional point cloud data, that is, the dominant grouping, and complete the clustering. This embodiment adopts the FCM algorithm to improve the value of its fuzziness m value to achieve the optimization of structural surface point cloud data clustering.

[0046] As a common clustering method, the FCM algorithm is different from other clustering algorithms in that it is based on the membership measurement index, assigning weights to each sample and each cluster to express the degree to which the object under study belongs to the cluster, so that it can take continuous values ​​within [0,1], thus improving the accuracy of the clustering results. U represents the membership matrix of the point cloud data, and the membership matrix is ​​expressed as follows:

[0047]

[0048]

[0049]

[0050]

[0051] Where: n is the total number of normal vector data of the point cloud, that is, the number of samples; c is the total number of cluster centers, that is, the number of selected cluster groups; u ij is the i-th sample x in the membership matrix U i The membership degree of the jth cluster is in the range of [0,1]; d ij is the i-th sample x i The Euclidean distance to the jth cluster; d ik is the i-th sample x i The Euclidean distance to the kth cluster; v j is the jth cluster; m is a parameter that controls the classification fuzziness, which is usually a real number greater than 1.

[0052] Use equations (3) and (4) to repeatedly modify the cluster center and data membership for classification. When the algorithm converges, theoretically, the cluster center of each category and the membership of each sample to each category are obtained, thus completing the fuzzy clustering division.

[0053] The FCM algorithm is to find an optimal classification so that the classification can produce the minimum objective function value J b .

[0054]

[0055] in,

[0056] d ij =d(x i -v j ) (6)

[0057] Where: V is the cluster set.

[0058] As a fuzzy parameter, the m value has a great influence on the effect of fuzzy C-means clustering. When the fuzziness value is reasonable, the tolerance of the algorithm to the data can be increased, making the algorithm more tolerant to the existence of noise, thereby improving the robustness to noisy data. For some data distributions that are "fuzzy" or difficult to clearly divide quantitatively, such as edge data or outliers, reasonable fuzziness helps to classify the data more reasonably, and can give more adequate consideration to data that does not obviously belong to a certain category; however, for data with obvious features and easy to divide, unreasonable fuzziness may cause the clustering results to be unclear and the boundaries of the clusters to be unclear. When faced with "fuzzy" data distribution, the algorithm may converge to the minimum value in advance, resulting in inaccurate clustering results. Therefore, choosing the right fuzziness is crucial for the clustering results. However, the current value of the fuzziness m value is mainly determined based on the operator's experience, such as taking m = 2 or 3.

[0059] Considering that the fuzziness of the data set and the significance of the category may be different in different situations, the fuzziness needs to be flexibly adjusted according to the actual situation for more reasonable calculation. Therefore, this embodiment proposes a new parameter to dynamically adjust the fuzziness of the algorithm by smooth adjustment, so that the fuzziness will not suddenly jump or be discontinuous, and it is adjusted within a continuous range to make the clustering results more stable and coherent. By judging whether the data category of the data set is obvious, it is determined whether the fuzziness of the algorithm needs to be increased or decreased to obtain the best results, thereby achieving the improvement of the FCM algorithm and enhancing the stability and adaptability of the algorithm.

[0060] Specifically, the fuzziness m is dynamically updated according to the distribution of the membership matrix U. By calculating the square difference between the maximum and minimum values ​​of the membership, an index ex is obtained to characterize the fuzzy nature and category characteristics of the data set.

[0061]

[0062] Where ex represents the fuzzy degree index of the membership calculated according to the membership matrix.

[0063] When the index ex exceeds the preset threshold α, it means that the category is obvious, and the value of the fuzziness m should be reduced to increase the fuzziness of the membership; if the index is lower than or equal to the threshold α, it means that the category is not obvious, and the value of the fuzziness m should be increased accordingly to reduce the fuzziness of the membership.

[0064]

[0065] Among them, α is a newly introduced parameter, that is, a preset threshold, which is used to control the update of the ambiguity value m.

[0066] By dynamically adjusting the fuzziness, the improved FCM algorithm can more finely adjust the fuzziness of the membership, thereby optimizing the algorithm's iterative process to adapt to the characteristics of various data sets and the needs of clustering tasks. The improved FCM algorithm is used to perform cluster analysis on the structural surface point cloud data, and after the dominant grouping is completed, the structural surface can be identified and information extracted from the point cloud data of each group of dominant structural surfaces.

[0067] Step 4, extracting structural surface trace length information;

[0068] The geometric features of the structural surface are extracted from the classified and grouped dominant structural surface point cloud data. Its essence is to divide the point cloud into adjacent point clusters according to a certain coplanarity. The initial space segmentation is achieved by using the FACETS plug-in function of Cloud Compare. In the past, the extraction of structural surface trace length was mainly based on point cloud data, and the distance between each point in the point cloud was calculated pairwise to obtain the trace length.

[0069] Due to the uncertainty characteristics and factors of point cloud data, many point cloud data outside the structural surface are difficult to be effectively identified and eliminated, and are involved in the extraction process of structural surface geometric data, resulting in inaccurate calculation results.

[0070] Therefore, this embodiment first uses the Fast Marching algorithm to close the clustered point cloud data into a surface. The algorithm uses an octree structure to describe the tree-like data elements of the three-dimensional space. The main process is to refine the initial space with a regular grid defined by the octree structure, and then determine whether to merge based on whether the root mean square distance (RMS) of adjacent elements in the octree exceeds the "maximum distance" standard. If not, merge, and iterate. In the recognition process, appropriate parameters should be selected according to the size of the structural surface and noise factors to obtain the best fitting effect. The dominant structural surface group generated in this way better reproduces the geometric shape and spatial interlacing rules of the real structural surface, effectively avoiding the inconvenience of subsequent applications caused by the discrete distribution of point cloud data, and automatically eliminating vegetation noise.

[0071] Secondly, the boundary surface formed by the Fast Marching algorithm is output and imported into ArcGIS software or other graphics processing software to automatically extract the coordinates of the turning points of the boundary surface.

[0072] Finally, (9) is used to directly calculate the distance between turning points, thus obtaining the trace length:

[0073]

[0074] Where D pq represents the boundary surface trace length, (X p ,Y p ,Zp ) represents the coordinates of the turning point p on the boundary surface, p = 1, 2, 3, ..., s-1; (X q ,Y q ,Z q ) represents the coordinates of the turning point q on the boundary surface, q = p+1, p+2,…,s.

[0075] The following takes a rock slope in a quarry as an example to verify the method proposed in this embodiment:

[0076] Unmanned aerial vehicles were used to obtain high-definition images of the rock slope from multiple angles. The acquired images were then used to reconstruct the three-dimensional model of the collected images of the study area using Metashape professional software to obtain a three-dimensional point cloud model of the study area.

[0077] The FCM algorithm with improved fuzzy values ​​was used to perform cluster analysis on the point cloud data of the study area, and a total of 6 groups of dominant structural surface occurrences were calculated. The results obtained by the improved algorithm were compared with the occurrences obtained by traditional empirical values, that is, fuzziness values ​​of 2 and 3. The results are shown in Table 1. It is found that the standard deviation of the structural surface occurrence obtained by the improved FCM algorithm is smaller and the occurrence accuracy is higher, which proves the effectiveness of the algorithm improvement.

[0078] Table 1

[0079]

[0080] As shown in Table 2, the trace length of the structural surface extracted by clustering was automatically extracted. The error values ​​of the extracted spacing results were all at the centimeter level, the maximum error did not exceed 1 cm, and the relative error was less than 2%. The error was small and could meet the accuracy requirements.

[0081] Table 2

[0082]

[0083] In summary, compared with the results extracted by the traditional algorithm, the average standard deviation of the occurrence of each group in this method is reduced by up to 1.17°, and the standard deviation of a single group is reduced by up to 4°, and the extraction results are more accurate. The method of automatically extracting the trace length of the structural surface greatly improves the efficiency of extraction, and the relative error of the result compared with the actual measurement is less than 2%, which is small and can meet the accuracy requirements. In addition, the measured spacing and trace length of the structural surface are consistent with the actual field survey results, which proves the accuracy of this method.

[0084] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for identifying rock mass structural surfaces and extracting boundary surface trace length based on fuzziness improvement, characterized in that: include: Collecting three-dimensional point cloud data of rock mass in the study area, and calculating normal vectors of the three-dimensional point cloud data; Performing cluster analysis on the normal vectors based on the C-means algorithm with improved fuzzy value to obtain dominant three-dimensional point cloud data; Based on the advantageous three-dimensional point cloud data, geometric features of the structural surface and turning points of the boundary surface are extracted, and then the trace length of the boundary surface is calculated to complete the identification of the rock mass structural surface in the study area and the extraction of the trace length of the boundary surface; The C-means algorithm for improving fuzzy value selection is as follows: dynamically adjusting fuzzy value selection by introducing parameters into the fuzzy C-means clustering algorithm; The method of dynamically adjusting the fuzziness value by introducing parameters in the fuzzy C-means clustering algorithm is as follows: Calculate the index of membership fuzziness degree; Based on the indicator, the adjustment of the fuzziness value is judged, when the indicator exceeds a preset threshold, the fuzziness value is reduced; when the indicator is lower than or equal to the threshold, the fuzziness value is increased; The calculation method of the index of the fuzzy degree of membership is: ex=max(u ij 2 )-min(u ij 2 ) Among them, ex represents the index of the fuzzy degree of membership; u ij Indicates the membership of the i-th sample in the membership matrix U to the j-th cluster; The method for judging the adjustment of the ambiguity value based on the indicator is: Among them, m represents the fuzzy value, and α is the introduced parameter.

2. The method for identifying rock mass structural surfaces and extracting boundary surface trace length based on fuzzy improvement according to claim 1 is characterized in that: Collecting the three-dimensional point cloud data of the rock mass in the study area includes: Collect rock mass images in the study area by using an image acquisition device mounted on a drone; A three-dimensional point cloud model is constructed based on the rock mass image, and then the three-dimensional point cloud data is obtained.

3. The rock mass structural surface identification and boundary surface trace length extraction method based on fuzzy improvement according to claim 1 is characterized in that: The normal vector of the three-dimensional point cloud data is calculated using a random Hough transform algorithm.

4. The method for identifying rock mass structural surfaces and extracting boundary surface trace length based on fuzzy improvement according to claim 1, characterized in that: Extracting geometric features of structural surfaces and turning points of boundary surfaces based on the advantageous three-dimensional point cloud data includes: The advantageous three-dimensional point cloud data is closed into a surface using a Fast Marching algorithm to obtain a boundary surface; The coordinates of the turning points of the boundary surface are extracted based on the boundary surface by using a graphics processing technique.

5. The method for identifying rock mass structural surfaces and extracting boundary surface trace length based on fuzzy improvement according to claim 1, characterized in that: The method for calculating the boundary surface trace length is: Among them, D pq represents the boundary surface trace length, (X p ,Y p ,Z p ) represents the coordinates of the turning point p on the boundary surface, p = 1, 2, 3, ..., s-1; (X q ,Y q ,Z q ) represents the coordinates of the turning point q on the boundary surface, q = p+1, p+2,…,s.

Images