A method for extracting geometric information of structural planes based on three-dimensional laser point clouds

Abstract

The present invention relates to a method for extracting geometric information of structural planes based on three-dimensional laser point clouds, which relates to the field of image processing and includes the following steps: arranging targets according to a scheme; obtaining ranging observation values, lateral scanning angle observation values, and longitudinal scanning angle observation values by using a three-dimensional laser scanner, calculating to obtain point cloud data and performing noise point exclusion; distributing the point cloud data after exclusion into different octree nodes, using a clustering algorithm to perform clustering operations on the point cloud data within each node, and obtaining the structural plane by calculating the normal direction of the clustering; extracting the attitude information of the structural plane, the trace line information of the structural plane, and the spacing information of the structural plane, and simulating and generating a random structural plane network inside the rock mass according to the probability distribution function of the geometric parameters of the structural plane to judge the evolution and influence of the structural plane. The present invention has the advantages of being able to more accurately obtain the structural plane information, and thus being able to more accurately predict the evolution and influence of the structural plane.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a method for extracting geometric information of structural planes based on three-dimensional laser point clouds. Background Art

[0002] The stability of tunnel surrounding rock is mainly controlled by a large number of structural planes developed inside the rock mass. The existence and mutual cutting of the structural planes cause key blocks to form on the free face of the rock mass, posing serious safety hazards to tunnel construction. How to accurately and efficiently obtain the information of the rock mass structural planes and predict and control the key blocks is of great significance for the stability analysis of tunnel surrounding rock.

[0003] Three-dimensional laser scanning technology has the characteristics of high precision, fast speed, and can approximate the original shape of the object, and has become a new remote sensing technology. Since the three-dimensional data of the collected targets by this technology are all actual data, the subsequent data processing results are completely real, thus making up for the defects of traditional collection means. Summary of the Invention

[0004] (I) Technical Problems to be Solved

[0005] The technical problem to be solved by the present invention is how to calculate more accurate geometric information of structural planes based on the accurate point cloud data obtained by three-dimensional laser scanning technology.

[0006] (II) Technical Solutions

[0007] To solve the above technical problems, the present invention provides a method for extracting geometric information of structural planes based on three-dimensional laser point clouds, including the following steps:

[0008] I. Conduct a field survey, determine the scanning range, scanning angle, and the layout scheme of scanning stations and scanning targets according to the topographic features of the measured area, the shape and spatial distribution of the measured object, and arrange the targets according to the scheme;

[0009] II. Use a three-dimensional laser scanner to obtain range measurement observations, lateral scanning angle observations, and longitudinal scanning angle observations, calculate the point cloud data and perform noise point elimination;

[0010] III. Assign the point cloud data after elimination to different octree nodes, use a clustering algorithm to perform clustering operations on the point cloud data within each node, and obtain the structural plane by calculating the normal direction of the clustering;

[0011] IV. Extract the attitude information of the structural plane, the trace line information of the structural plane, and the spacing information of the structural plane, and simulate and generate a random structural plane network inside the rock mass according to the probability distribution function of the geometric parameters of the structural plane to judge the evolution and influence of the structural plane.

[0012] As a further illustration of the present invention, preferably, before distributing the point cloud data, the calculated point cloud data is used to estimate the density or calculate the distance by calculating the number of point clouds within the octree nodes, a density threshold or a distance threshold is set, and the nodes outside the density threshold or the distance threshold and the point cloud data they contain are excluded as noise points or non-structural surface regions.

[0013] As a further illustration of the present invention, preferably, before distributing the point cloud data, for each leaf node of the octree in the calculated point cloud data, the neighborhood of the points within the node is calculated and analyzed, and it is determined whether it is a structural surface according to the characteristics of the points in the neighborhood, and the non-structural surface points are excluded as noise points.

[0014] As a further illustration of the present invention, preferably, before distributing the point cloud data, the calculated point cloud data is used to perform connectivity analysis by calculating the connection relationship between each leaf node of the octree and its adjacent nodes. If the connection relationship of the node does not meet the preset conditions, it is excluded as a noise point.

[0015] As a further illustration of the present invention, preferably, after obtaining the denoised point cloud data, the maximum depth of the octree and the maximum number of points contained in each leaf node are defined; after the clustering algorithm is completed, the clustered leaf nodes are merged into larger clustering units according to the point cloud density to improve the accuracy of structural surface resolution.

[0016] As a further illustration of the present invention, preferably, the structural plane attitude information is represented by the plane normal vector N, the dip angle α, and the dip direction β, specifically as follows:

[0017]

[0018] Among them, A, B, and C are the coefficients of x, y, and z in the plane equation respectively;

[0019]

[0020]

[0021] As a further illustration of the present invention, preferably, the structural plane trace information is represented by the trace length L, specifically as follows:

[0022]

[0023] Among them, (x1, y1, z1) and (x2, y2, z2) are the two end points on the structural plane trace.

[0024] As a further illustration of the present invention, preferably, the structural plane spacing information is obtained by arbitrarily selecting two points on two structural planes and obtaining the direction vector starting from the two points, calculating the cosine value of its direction, as well as the dip angle and dip direction of the measuring line, and finally calculating the spacing of the structural plane. Specifically:

[0025] d = L|sinα d cosα L cos(β d -β L )-sinα L cosα d |

[0026] Wherein,

[0027] d is the structural plane spacing;

[0028] α L is the dip angle of the measuring line;

[0029] α d is the dip angle of the structural plane;

[0030] β L is the dip direction of the measuring line;

[0031] β d is the dip direction of the structural plane.

[0032] As a further illustration of the present invention, preferably, the distribution function is obtained from the measured structural plane occurrence, trace length, and spacing information, and the three-dimensional structural plane network simulation is carried out using the Monte-Carlo method to represent the real structural plane network.

[0033] As a further illustration of the present invention, preferably, the Baecher disk model is used to simulate the structural plane, and it is represented by its corresponding dip angle, dip direction, radius, and center coordinates.

[0034] (III) Beneficial Effects

[0035] The above technical solutions of the present invention have the following advantages:

[0036] By introducing the octree algorithm to process the point cloud data, the present invention can not only improve the spatial fineness but also reduce the unnecessary number of nodes, save storage resources, and ensure the accuracy of the obtained structural plane data. Combined with the extraction method of the geometric information of the structural plane, it can accurately obtain the geometric information of the structural plane and simulate a real structural plane network to provide accurate judgment data for the subsequent tunnel support. Description of the Drawings

[0037] Figure 1 is the data acquisition diagram obtained by the laser scanner of the present invention;

[0038] Figure 2 is the tunnel point cloud map of the present invention;

[0039] Figure 3 is the schematic diagram of the attitude of structural planes of the present invention;

[0040] Figure 4 is the model diagram of the tunnel structural plane of the present invention;

[0041] Figure 5 is the marked diagram of 10 structural planes of the whole tunnel of the present invention;

[0042] Figure 6 is the three-dimensional structural plane network model diagram of the present invention. Specific Embodiment

[0043] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0044] A method for extracting geometric information of structural planes based on three-dimensional laser point cloud includes the following steps:

[0045] I. Conduct a field survey, determine the scanning range, scanning angle, and the layout scheme of scanning stations and scanning targets according to the terrain features of the area to be measured, the shape and spatial distribution of the object to be measured, and arrange the targets according to the scheme.

[0046] II. Obtain range measurement observation values, lateral scanning angle observation values, and longitudinal scanning angle observation values using a three-dimensional laser scanner, calculate the point cloud data, and exclude noise points through Gaussian filtering, median filtering, or average filtering, etc.

[0047] In addition, the calculated point cloud data can also be used to estimate the density or calculate the distance by calculating the number of point clouds in the octree nodes, set a density threshold or a distance threshold, and exclude the nodes outside the density threshold or distance threshold and the point cloud data they contain as noise points or non-structural plane areas. It can also be analyzed by calculating the neighborhood of points in each leaf node of the octree, and determine whether it is a structural plane according to the characteristics of the points in the neighborhood, and exclude non-structural plane points as noise points. Or the calculated point cloud data can be used to perform connectivity analysis by calculating the connection relationship between each leaf node of the octree and its adjacent nodes. If the connection relationship of the nodes does not meet the preset conditions, it is excluded as a noise point.

[0048] Ⅲ. After obtaining the denoised point cloud data, define the maximum depth of the octree and the maximum number of points contained in each leaf node; then insert the excluded point cloud data into different octree nodes one by one through a recursive algorithm, traverse the octree, and use a clustering algorithm (such as K-mean or DBSCAN) to perform clustering operations on the point cloud data within each node. Subsequently, merge the clustered leaf nodes into larger clustering units according to the point cloud density to improve the accuracy of structural plane resolution. Finally, the least squares method can be used to fit a plane to obtain the normal direction of the node to obtain the structural plane. Among them, for adjacent leaf nodes, if the normal directions are similar and the angle difference between the normals is less than the designed threshold, they are merged into one structural plane.

[0049] Ⅳ. Extract the attitude information, trace line information, and spacing information of the structural plane. Specifically:

[0050] 1. The attitude information of the structural plane refers to the distribution status of the structural plane in space, usually represented by the plane normal vector N, dip angle α, and dip direction β. As Figure 3 shown, the plane normal vector N refers to the direction of the structural plane on the horizontal plane. The dip direction β of the rock stratum refers to the direction indicated by the projection on the horizontal plane of the straight line perpendicular to the strike line and extending downward along the inclined plane; the dip angle α refers to the angle between the structural plane and the horizontal plane. Specifically:

[0051]

[0052] Among them, A, B, and C are the coefficients of x, y, and z in the plane equation respectively;

[0053]

[0054]

[0055] 2. The trace line information of the structural plane is represented by the trace length L. Specifically:

[0056]

[0057] where (x1, y1, z1) and (x2, y2, z2) are two end points on the trace line of the structural plane.

[0058] 3. The spacing information of the structural plane is obtained by randomly selecting two points on two structural planes and obtaining the direction vector starting from the two points, calculating the cosine value of its direction, as well as the dip angle and dip direction of the measuring line. Finally, calculate the spacing of the structural plane. Specifically:

[0059] d = L|sinα d cosα L cos(β d -β L ) - sinα L cosαd |

[0060] Wherein,

[0061] d is the spacing of structural planes;

[0062] (x1, y1, z1) and (x2, y2, z2) are the point coordinates on two structural planes;

[0063] α L is the dip angle of the survey line;

[0064] α d is the dip angle of the structural plane;

[0065] β L is the trend of the survey line;

[0066] β d is the trend of the structural plane.

[0067] The point cloud data is partitioned according to spatial positions by an octree, making the query and processing of the point cloud data more efficient, enabling rapid location and extraction of regions of interest. Moreover, the depth of traversing the octree can be freely selected as needed, enabling rapid retrieval and query of the point cloud data within a specific region, reducing unnecessary calculations and processing. While achieving automatic identification of structural planes, it can also adaptively adjust the granularity of spatial partitioning according to the data distribution during structural plane identification, continue to partition nodes when the point cloud data is less to improve spatial fineness, and stop partitioning when the number of point clouds is large to reduce unnecessary node numbers, combining with the extraction algorithm of structural plane geometric information to reduce the difficulty of the extraction calculation amount of structural plane geometric information. In addition, by only storing non-empty leaf nodes and the point cloud data therein, the storage requirements can be greatly reduced and storage resources can be saved.

[0068] Ⅴ. Simulate and generate a random structural plane network inside the rock mass according to the probability distribution function of structural plane geometric parameters to judge the evolution and influence of the structural planes. Obtain its distribution function through the measured structural plane attitude, trace length, and spacing information, and use the Monte-Carlo method for three-dimensional structural plane network simulation to represent the real structural plane network, as Figure 6 shown, and use the Baecher disk model to simulate the structural planes, and represent them with their corresponding dip angle, trend, radius, and center coordinates to help understand the spatial distribution and mutual relationship of the structural planes. In addition, by overlaying with other geological attributes or models subsequently, more comprehensive geological information can be obtained.

[0069] Apply this solution to the No. 7 tunnel of Xulong Hydropower Station for measurement, as Figure 1As shown in the figure, at the beginning of the measurement, the target is set up, and mainly more than three common targets are used as landmark points for station splicing. The method of sectional splicing is adopted. Taking the first reference station as the standard, all station data are spliced pairwise and successively transmitted in the same coordinate system to achieve the purpose of massive data station splicing.

[0070] The point cloud data obtained by scanning is denoised and data-reduced through an octree to obtain the point cloud data as Figure 2 shown. The structural plane is extracted according to the octree algorithm to obtain the effect diagram as Figure 4 shown. In order to more comprehensively analyze the recognition effect of the octree, as Figure 5 shown, 10 structural planes are selected and marked within the global range. Finally, the calculation results are compared with the manual calculation results of the classic open-source software CloudCompare. The comparison results are shown in Table 1:

[0071] Table 1 Comparison results of tunnel point cloud extraction of structural planes

[0072]

[0073] The results show that the calculation results using the octree are basically consistent with the manual calculation results of CloudCompare. The minimum error is 0.1°, and the maximum error is only 4.69°, both within the allowable error range of the project.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting geometric information of structural planes based on 3D laser point clouds, characterized in that: It includes the following steps: Ⅰ. Conduct a field survey, determine the scanning range, scanning angle, and the layout plan of scanning stations and scanning targets according to the terrain features of the measured area, the shape and spatial distribution of the measured objects, and arrange the targets according to the plan; Ⅱ. Use a three-dimensional laser scanner to obtain ranging observation values, horizontal scanning angle observation values, and vertical scanning angle observation values, calculate the point cloud data, and perform noise removal; Ⅲ. After obtaining the denoised point cloud data, define the maximum depth of the octree and the maximum number of points contained in each leaf node; and adaptively adjust the granularity of spatial division according to the data distribution. Continue to divide the nodes to improve the spatial fineness when the point cloud data is small, and stop dividing when the point cloud quantity is large to reduce the unnecessary number of nodes; Insert the point cloud data after exclusion into different octree nodes one by one through a recursive algorithm, traverse the octree, perform clustering operations on the point cloud data within each node using a clustering algorithm. After the clustering algorithm is completed, merge the clustered leaf nodes into larger clustering units according to the point cloud density to improve the accuracy of structural plane resolution. Finally, use the least squares method to fit the plane to obtain the normal direction of the node to obtain the structural plane; among them, leaf nodes that are adjacent and have similar normal directions and the angle difference between the normals is less than the design threshold are merged into one structural plane; Ⅳ. Extract the structural plane attitude information, structural plane trace information, and structural plane spacing information; among them, the structural plane spacing information is obtained by randomly selecting two points on two structural planes and obtaining the direction vectors starting from the two points, calculating the cosine value of its direction, the dip angle and dip direction of the measuring line, and finally calculating the spacing of the structural plane. Specifically: d = L|sinα d cosα L cos(β d -β L ) - sinα L cosα d | Wherein, d is the structural plane spacing; α L is the dip angle of the survey line; α d is the dip angle of the structural plane; β L is the dip direction of the survey line; β d is the dip direction of the structural plane; a random structural plane network inside the rock mass is simulated and generated according to the probability distribution function of the geometric parameters of the structural plane to judge the evolution and influence of the structural plane.

2. The method for extracting geometric information of structural planes based on 3D laser point clouds according to claim 1, wherein: Before allocating the point cloud data, estimate the density or calculate the distance of the calculated point cloud data by calculating the number of point clouds in the octree node, set the density threshold or distance threshold, and exclude the nodes and the point cloud data they contain outside the density threshold or distance threshold as noise or non-structural plane areas.

3. A method for extracting geometric information of structural planes based on three-dimensional laser point clouds according to claim 1, characterized in that: Before allocating the point cloud data, for each leaf node of the octree in the calculated point cloud data, calculate the neighborhood of the points in the node and conduct an analysis, judge whether it is a structural plane according to the characteristics of the points in the neighborhood, and exclude the non-structural plane points as noise.

4. A method for extracting geometric information of structural planes based on three-dimensional laser point clouds according to claim 1, characterized in that: Before allocating the point cloud data, conduct a connectivity analysis on the calculated point cloud data by calculating the connection relationship between each leaf node of the octree and its adjacent nodes. If the connection relationship of the node does not meet the preset conditions, exclude it as noise.

5. A method for extracting geometric information of structural planes based on three-dimensional laser point clouds according to claim 1, characterized in that: The structural plane attitude information is represented by the plane normal vector N, dip angle α, and dip direction β. Specifically: Wherein, A, B, and C are the coefficients of x, y, and z in the plane equation respectively; 6. A method for extracting geometric information of structural planes based on three-dimensional laser point clouds according to claim 5, characterized in that: The structural plane trace information is represented by the trace length L. Specifically: Where (x1, y1, z1) and (x2, y2, z2) are the two end points on the structural plane trace.

7. A method for extracting geometric information of discontinuity surfaces based on 3D laser point clouds according to claim 6, characterized in that: Obtain its distribution function through the measured structural plane attitude, trace length, and spacing information, and use the Monte-Carlo method to conduct a three-dimensional structural plane network simulation to represent the real structural plane network.

8. A method for extracting geometric information of structural planes based on three-dimensional laser point clouds according to claim 7, characterized in that: The Baecher disc model is used to simulate the structural plane, and it is represented by its corresponding dip angle, dip direction, radius, and center coordinates.

Images