A method for identifying occurrence of a slope outcrop structural plane considering spatial extension

Abstract

The present application relates to a kind of considering spatial extension's slope outcrop structure plane occurrence identification method, method includes: obtaining original data, establishes point cloud model, obtains sparse point cloud, interpolation is then carried out, then to local plane is clustered, and the clustering result is obtained;Filtering first plane set and second plane set;Output all class plane corresponding optimal plane, complete occurrence identification.Compared with prior art, the present application has the advantages of improving the stability of structure plane occurrence identification, etc.

Description

Technical Field

[0001] This invention relates to the technical field of geotechnical engineering, and in particular to a method for identifying the orientation of outcrop structural surfaces of slopes that takes into account spatial extension. Background Technology

[0002] Traditional outcrop surveying relies primarily on tools such as geological compasses and rulers, which are characterized by high risks, low efficiency, and limited scope, making it difficult to meet the requirements of information sufficiency and timeliness. Non-contact methods such as UAV oblique photography, 3D laser scanning, and close-range photography provide new means for outcrop surveying in slope engineering, effectively improving the scope and efficiency of geological exploration.

[0003] While these methods have shown good performance in identifying stratigraphic attitudes, they are not very stable in outcrop identification scenarios with broken or discontinuous structural surfaces, as the exposed planes are discontinuous or indistinct, requiring frequent manual intervention. Summary of the Invention

[0004] The purpose of this invention is to provide a method for identifying the attitude of outcrop structural surfaces of slopes that takes into account spatial extension, in order to improve the stability of the identification of the attitude of outcrop structural surfaces.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A method for identifying the attitude of outcrop structural surfaces on slopes that considers spatial extension, the method comprising:

[0007] S1. Obtain raw data, wherein the raw data is a stratigraphic outcrop image or stratigraphic outcrop point cloud data;

[0008] S2. Generate a point cloud model corresponding to the original data based on the motion recovery structure algorithm;

[0009] S3. Use a uniform downsampling method to downsample the point cloud model to obtain a sparse point cloud. Calculate the normal of each point in the sparse point cloud and interpolate it to obtain the interpolated point cloud.

[0010] S4. Divide the point cloud at the location of the structure surface in the interpolated point cloud into a set composed of local planes, and record the center point position, plane size, and plane direction of all local planes.

[0011] S5. Perform clustering on the local plane to obtain the clustering results;

[0012] S6. Select one type of plane from the clustering results to obtain the initial structure plane set and initialize the number of iterations;

[0013] S7. Select planes that meet the error requirements from the initial structural plane set to form the first plane set;

[0014] S8. Randomly select two planes from the first set of planes, calculate the fitting plane, and select the planes that satisfy the fitting plane error from the first set of planes to form the second set of planes.

[0015] S9. Determine if the number of planes in the second set of planes is greater than the number of optimal matching planes. If not, increment the loop count by 1 and return to S8. If yes, set the number of optimal matching planes to the number of planes in the second set of planes, use the second set of planes as the optimal planes, use the second set of planes as the new first set of planes, increment the loop count by 1, and return to S8. When the number of loops reaches the threshold, output the optimal planes corresponding to all planes of this type, convert the center point position and plane direction of the optimal planes into coordinates expressed in the engineering coordinate system, complete the attitude identification of this type of plane, and then execute S10.

[0016] S10. Select another class plane in the clustering results and repeat S6-S10 until all classes in the clustering results are traversed to complete the identification of the orientation of all class planes.

[0017] Furthermore, the specific steps for generating the point cloud model corresponding to the original data based on the structure-of-motion reconstructive motion algorithm are as follows:

[0018] Features are extracted from the raw data using a multi-view vision algorithm, and then a point cloud model is generated based on the motion reconstruction structure algorithm.

[0019] Furthermore, a uniform downsampling method is used to downsampling the point cloud model to obtain a sparse point cloud. The normal vector of each point in the sparse point cloud is calculated and interpolated to obtain the interpolated point cloud. The specific steps are as follows:

[0020] Iterate through each point in the downsampled point cloud model, and with each point o as the center of a sphere, select all points contained within the sphere according to radius r to form a point set P. Calculate the vector v from point p to point o in point set P. i The cokriging method is used to analyze vector v. i Perform interpolation to obtain the normal of the point, and obtain the interpolated point cloud.

[0021] Furthermore, the method used to divide the point cloud at the location of the structural surface in the sparse point cloud into a set composed of local planes is the KD tree method.

[0022] Furthermore, the specific steps for clustering local planes are as follows:

[0023] Clustering of local planes based on planar distance and normal.

[0024] Furthermore, the specific steps for selecting planes that meet the error requirements from the initial structural plane set to form the first plane set are as follows:

[0025] Select any plane p0 from the initial structure plane set as the initial fitting plane sp0. If the plane in the initial structure plane set and the initial fitting plane sp0 meet the requirements of the inter-plane distance within the first threshold range and the angle error within the second threshold range, then add the plane to the first plane set.

[0026] Furthermore, select any two planes p from the first set of planes. i With p j Calculate the fitted plane sp i From the first set of planes, planes that satisfy the fitting plane error are selected, and the selected planes are added to the second set of planes. The fitting plane error is: the error between the fitted plane and the fitted plane sp. i The distance between them is within the first threshold range, and the angle error is within the second threshold range.

[0027] Furthermore, the fitting plane sp i The coordinates of the center point are:

[0028]

[0029] in, For plane p i The coordinates of the center point, For plane p j The coordinates of the center point.

[0030] Furthermore, the fitting plane sp i The normal vector is:

[0031]

[0032] in, For plane p i The normal vector, For plane p j The normal vector.

[0033] Furthermore, the specific steps for downsampling the point cloud model using the uniform downsampling method are as follows:

[0034] Randomly select 1 / n points from the point cloud model according to the downsampling rate n to obtain the downsampled point cloud model.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1) The original data of this invention can be obtained by drones, thereby acquiring large quantities of structural surface occurrence information of outcrops and working faces quickly.

[0037] 2) This invention directly processes the original point cloud and calculates the points directly, rather than analyzing the existing geometric model (triangular mesh) of the point cloud. It uses the Kriging method to avoid smoothing the point normal calculation.

[0038] 3) The method of the present invention first performs clustering, and then performs two screenings on the clustered planes. It takes into account the influence of construction disturbance on the structural surface, that is, it considers the possibility that a structural surface may be divided into several surfaces. Therefore, it can ignore the influence of local structural surfaces. Compared with traditional structural surface orientation identification methods, it has higher accuracy and can effectively reduce the influence of noise such as weathering, construction interference, and local protrusions on the main structural surface, reduce human intervention, and improve the stability of structural surface orientation identification. Attached Figure Description

[0039] Figure 1 This is a flowchart of the present invention;

[0040] Figure 2 A schematic diagram of a sparsed point cloud;

[0041] Figure 3 A schematic diagram of a local plane set;

[0042] Figure 4 This is a schematic diagram of the extracted structural surfaces. Detailed Implementation

[0043] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0044] Definitions:

[0045] 1. Point cloud: A set of discrete points (x, y, z).

[0046] 2. Strata: Layered distribution of similar rock or soil bodies deposited or formed in the same era.

[0047] 3. Attitude: Spatial characteristics of strata, including dip and dip angle.

[0048] 4. Structural planes: A general term for various geological interfaces that cut through rock masses. They are two-dimensional geological interfaces that have a certain direction and are relatively wide and thin.

[0049] 5. Digital photography: Using mobile phones, cameras, video cameras, etc. to obtain digital photos of subjects.

[0050] 6. 3D laser scanning: A sensing method that directly acquires point clouds, requiring the use of a 3D laser scanning instrument.

[0051] This invention proposes a method for identifying the attitude of outcrop structural surfaces on slopes that considers spatial extension. The flowchart of the method is as follows: Figure 1 As shown. This invention utilizes UAV oblique photography to register, crop, and denoise the native point cloud. It calculates the point cloud normal using the Kriging method and uses a KD-tree (K-Dimensional Tree) to match local structural surfaces. The method considers the spatial extension of adjacent structural surfaces and investigates an optimal structural surface fast identification method based on minimum vector weighted distance and random sampling cycles. Compared with traditional structural surface orientation identification methods, the method of this invention has higher matching efficiency and accuracy, effectively reducing the impact of factors such as weathering, local protrusions, excavation effects, and construction blasting on the continuity of structural surfaces. The accuracy matches that of manual measurements, and the algorithm exhibits good stability.

[0052] The steps of this invention include:

[0053] S1. Obtain raw data, wherein the raw data is a stratigraphic outcrop image or stratigraphic outcrop point cloud data;

[0054] S2. Generate a point cloud model corresponding to the original data based on the motion recovery structure algorithm;

[0055] S3. Use a uniform downsampling method to downsample the point cloud model to obtain a sparse point cloud. Calculate the normal of each point in the sparse point cloud and interpolate it to obtain the interpolated point cloud.

[0056] S4. Divide the point cloud at the location of the structure surface in the interpolated point cloud into a set composed of local planes, and record the center point position, plane size, and plane direction of all local planes.

[0057] S5. Perform clustering on the local plane to obtain the clustering results;

[0058] S6. Select one type of plane from the clustering results to obtain the initial structure plane set and initialize the number of iterations;

[0059] S7. Select planes that meet the error requirements from the initial structural plane set to form the first plane set;

[0060] S8. Randomly select two planes from the first set of planes, calculate the fitting plane, and select the planes that satisfy the fitting plane error from the first set of planes to form the second set of planes.

[0061] S9. Determine if the number of planes in the second set of planes is greater than the number of optimal matching planes. If not, increment the loop count by 1 and return to S8. If yes, set the number of optimal matching planes to the number of planes in the second set of planes, use the second set of planes as the optimal planes, use the second set of planes as the new first set of planes, increment the loop count by 1, and return to S8. When the number of loops reaches the threshold, output the optimal planes corresponding to all planes of this type, convert the center point position and plane direction of the optimal planes into coordinates expressed in the engineering coordinate system, complete the attitude identification of this type of plane, and then execute S10.

[0062] S10. Select another class plane in the clustering results and repeat S6-S10 until all classes in the clustering results are traversed to complete the identification of the orientation of all class planes. Figure 1 In the process, the judgment of whether the new segmentation is better than the previous segmentation is implemented through S9. If the number of new segmentation planes is greater than the number of optimal matching planes, it means that the new segmentation is better than the previous segmentation; otherwise, it is not better.

[0063] In S1, drones, digital photography, and 3D laser scanners are used to acquire raw photos or point cloud data.

[0064] Drone Method: Due to the irregular shapes of geological outcrops, drone photography generally employs manual flight mode with an overlap rate of 80%-85%. To reduce shading and modeling gaps caused by outcrop protrusions, multi-angle shooting is used. During aerial photography, the drone uses real-time kinematic (RTK) positioning to record the shooting position and pitch angle information of the photos.

[0065] Digital photography method: Using smartphones, digital cameras, etc., take pictures of the structural surface to be modeled from at least six angles, with an overlap rate of more than 80% between different photos.

[0066] The 3D laser scanning method uses a 3D laser scanner to scan the structural surfaces that need to be modeled, directly generating point clouds.

[0067] In S2, point cloud models are obtained using the SFM method for digital and drone photographs. Multi-view vision algorithms are used to extract feature descriptions such as corner points, edges, and textures from each image, including scale-invariant feature transform (SIFT) points. These feature values ​​possess illumination, scaling, and affine invariance. The Structure from Motion (SFM) algorithm is then used to generate the point cloud of the target region. Alternatively, software can be used as an oblique photogrammetry modeling tool to reconstruct the point cloud model.

[0068] In S3, an average sampling method is used to reduce the point cloud density and obtain a sparse point cloud. The inverse distance kriging interpolation method is then used to calculate the normals of the points. A schematic diagram of the sparsed point cloud is shown below. Figure 2 As shown.

[0069] The specific method is as follows:

[0070] For the original point cloud C, number all points sequentially, and randomly select 1 / n of the points according to the downsampling rate n (i.e., the number of points after sampling is 1 / n of the original) to generate a new point cloud.

[0071] For each downsampled point, iterate through all points, and with each point o as the center of a sphere, select all points contained within the sphere according to radius r to form a point set P. Calculate the vector v from point p to point o in point set P. i The Kriging method is used (Equations 1 and 2, where var represents the covariance and λ). i Representative vector v i The corresponding normal weights are obtained through the constraints in Formula 2. To obtain the final estimated normal, we use formulas 1 and 2 to minimize the covariance, thereby obtaining v0, and then perform interpolation based on the weights.

[0072]

[0073]

[0074] v0 represents the normal vector to be obtained, λ i Representative vector v i For calculating the weights of v0, the cokriging method is used to ensure that v0 is related to v... i Solving for λ under the constraint of minimizing covariance. i Thus utilizing λ i For v i Interpolation is performed to obtain the final v0.

[0075] In S4, the point cloud at the location of the structural plane is extracted, and the KD-tree method is used to divide the point cloud into a set composed of local planes. The center point position, plane size, and plane orientation of each plane are recorded. A schematic diagram of the local plane set is shown below. Figure 3 As shown.

[0076] In step S5, spatially adjacent planes (with a planar interval of no more than 1m) with a local planar normal error of less than 10° are clustered. During this step, the same plane may appear in different clusters, but this does not affect the subsequent identification of the orientation of structural surfaces.

[0077] In S6, for each plane cluster, the maximum number of matching planes is set to 0, and the number of iterations is set.

[0078] In S7, a matching error is selected, an initial plane is randomly selected, and a set of first planes that satisfy the error is obtained based on the initial plane.

[0079] Based on the clustering results, an initial set of structural planes, SPF0, is obtained. The number of optimal matching planes is set to 0, and the number of iterations and matching error are set. The number of iterations is determined based on the number of planes contained in SPF0, typically 500-10000. The matching error is set according to the actual distribution of structural planes: the distance between planes is within a first threshold range (0.1-0.5m), and the angle error is within a second threshold range (3°-10°). A plane p0 is randomly selected as the initial fitting plane sp0, and its equation is... Where (x0, y0, z0) are the coordinates of the center point of the plane. For the plane normal, set the included angle threshold to error. angle The planar distance threshold is error. dis Select structural surfaces within the sp0 error range to form the set SPF. i (i is the iteration number), and calculate SPF. i The number of inner surfaces.

[0080] In S8, two planes are randomly selected from the local planes, a fitting plane is calculated, and based on the fitted plane, all local planes that satisfy the error are calculated, and the number of matching planes is counted.

[0081] According to SPF i Randomly select two new planes p i With p j According to formulas 3 and 4, calculate the new fitting plane sp. i Select sp i SPF (Surface Formation Array) within the Error Range i+1 .

[0082]

[0083]

[0084] In S9, if the number of iterations is less than the set number of iterations, repeat S7; otherwise, output the center point position and orientation of the optimal plane. (SPF is then checked.) i+1 If the number of planes is greater than the number of optimal matching planes, discard them; otherwise, set the number of optimal matching planes to SPF. i+1 The number of planes within.

[0085] In S10, the center point position and plane direction of the optimal plane are transformed into coordinates expressed in the engineering coordinate system, completing the extraction of the attitude corresponding to one cluster. Repeating S6 to S10 completes the attitude identification of all clusters (structural planes). A schematic diagram of the extracted structural planes is shown below. Figure 4 As shown.

[0086] The advantages of this invention are:

[0087] 1) It can quickly acquire structural surface attitude information of outcrops and working faces in large quantities.

[0088] 2) Process the original point cloud directly to avoid the reduction in measurement accuracy that may be caused by meshing and smoothing.

[0089] 3) It can effectively reduce the impact of noise such as weathering and flaking, construction interference, and local protrusions on the main structural surfaces, and reduce human intervention.

[0090] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for identifying the attitude of outcrop structural surfaces on slopes considering spatial extension, characterized in that the method... include: S1. Obtain raw data, wherein the raw data is a stratigraphic outcrop image or stratigraphic outcrop point cloud data; S2. Generate the point cloud model corresponding to the original data based on the motion recovery structure algorithm; S3. Use a uniform downsampling method to downsample the point cloud model to obtain a sparse point cloud. Calculate the normal of each point in the sparse point cloud and interpolate it to obtain the interpolated point cloud. S4. Divide the point cloud at the location of the structure surface in the interpolated point cloud into a set composed of local planes, and record the center point position, plane size, and plane direction of all local planes. S5. Perform clustering on the local plane to obtain the clustering results; S6. Select one type of plane from the clustering results to obtain the initial structure plane set and initialize the number of iterations; S7. Select planes that meet the error requirements from the initial structural plane set to form the first plane set; S8. Randomly select two planes from the first set of planes, calculate the fitting plane, and select the planes that satisfy the fitting plane error from the first set of planes to form the second set of planes. S9. Determine if the number of planes in the second set of planes is greater than the number of optimal matching planes. If not, increment the loop count by 1 and return to S8. If yes, set the number of optimal matching planes to the number of planes in the second set of planes, use the second set of planes as the optimal planes, use the second set of planes as the new first set of planes, increment the loop count by 1, and return to S8. When the number of loops reaches the threshold, output the optimal planes corresponding to all planes of this type, convert the center point position and plane direction of the optimal planes into coordinates expressed in the engineering coordinate system, complete the attitude identification of this type of plane, and then execute S10. S10. Select another class plane in the clustering results and repeat S6-S10 until all classes in the clustering results are traversed to complete the identification of the orientation of all class planes.

2. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 1, characterized in that, The specific steps for generating the point cloud model corresponding to the original data based on the structure-of-motion (SOG) algorithm are as follows: Features are extracted from the raw data using a multi-view vision algorithm, and then a point cloud model is generated based on the motion reconstruction structure algorithm.

3. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 2, characterized in that, The specific steps for downsampling the point cloud model using a uniform downsampling method to obtain a sparse point cloud, calculating the normal of each point in the sparse point cloud for interpolation, and obtaining the interpolated point cloud are as follows: Iterate through each point in the downsampled point cloud model, and with each point o as the center of a sphere, select all points contained within the sphere according to radius r to form a point set P. Calculate the vector v from point p to point o in point set P. i The cokriging method is used to analyze vector v. i Perform interpolation to obtain the normal of the point, and obtain the interpolated point cloud.

4. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 1, characterized in that, The method used to divide the point cloud at the location of the structure plane in the sparse point cloud into a set composed of local planes is the KD tree method.

5. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 1, characterized in that, The specific steps for clustering local planes are as follows: Clustering of local planes based on planar distance and normal.

6. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 1, characterized in that, The specific steps for selecting planes that meet the error requirements from the initial structural plane set to form the first plane set are as follows: Select any plane p0 from the initial structure plane set as the initial fitting plane sp0. If the plane in the initial structure plane set and the initial fitting plane sp0 meet the requirements of the inter-plane distance within the first threshold range and the angle error within the second threshold range, then add the plane to the first plane set.

7. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 6, characterized in that, Choose any two planes p from the first set of planes i With p j Calculate the fitted plane sp i From the first set of planes, planes that satisfy the fitting plane error are selected, and the selected planes are added to the second set of planes. The fitting plane error is: the error between the fitted plane and the fitted plane sp. i The distance between them is within the first threshold range, and the angle error is within the second threshold range.

8. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 7, characterized in that, Fitting plane sp i The coordinates of the center point are: in, For plane p i The coordinates of the center point, For plane p j The coordinates of the center point.

9. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 7, characterized in that, Fitting plane sp i The normal vector is: in, For plane p i The normal vector, For plane p j The normal vector.

10. The method for identifying the attitude of outcrop structural surfaces of slopes considering spatial extension according to claim 1, characterized in that, The specific steps for downsampling the point cloud model using the uniform downsampling method are as follows: Randomly select 1 / n points from the point cloud model according to the downsampling rate n to obtain the downsampled point cloud model.

Images