Dangerous rock mass discontinuous edge point extraction method and equipment based on improved central axis transformation

Abstract

The invention discloses a dangerous rock mass discontinuous edge point extraction method and device based on improved central axis transformation. The method comprises the following steps: acquiring discontinuous edge line candidate points of a rock mass from three-dimensional point cloud data with a normal in a target area; performing point cloud blocking on the three-dimensional point cloud data to obtain three-dimensional point cloud cluster data; calculating the curvature of discontinuous edge line candidate points in the selected three-dimensional point cloud cluster data, and separating the three-dimensional point cloud cluster data based on a region growing algorithm to obtain a separation region corresponding to each piece of three-dimensional point cloud cluster data; extracting three-dimensional point cloud cluster data and three-dimensional boundary point clouds of the partition areas; and fusing the three-dimensional point cloud cluster data and the three-dimensional boundary point clouds of the separated areas, and carrying out filtering to obtain a final dangerous rock mass discontinuous line. According to the method, the precision and reliability of discontinuous feature extraction of the dangerous rock mass are remarkably improved, the harsh requirement for the uniformity of initial data is reduced, and the robustness and universality under complex geological conditions and different acquisition environments are improved.

Description

Technical Field

[0001] This invention belongs to the field of ground-based lidar point cloud segmentation for extracting discontinuous features in rock mass hazard areas, specifically involving a method and equipment for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation. Background Technology

[0002] The study of discontinuities and discontinuities in rock masses is crucial for understanding and predicting rock failure modes, assessing slope stability, and designing protective measures. Accurate extraction of these features provides key information for numerical simulation and risk assessment of geological hazards, thus playing a vital role in rock engineering and environmental safety. Measurement of discontinuous geometry in rock masses is typically conducted on-site using handheld devices such as calipers and geological compasses to identify rock mass characteristics. However, these methods are time-consuming, dangerous, and the results are heavily influenced by subjective judgment and experience, often failing to meet the needs of rapid engineering construction. Currently, non-contact measurement technologies, such as photogrammetry and LiDAR (laser detection and ranging), provide alternative methods for on-site measurement of discontinuous geometry through high-resolution images and 3D point clouds of the rock mass surface. These methods significantly improve the safety, data collection efficiency, and objectivity of discontinuity trajectory measurement.

[0003] Many studies have focused on discontinuity trajectory detection based on 2D digital images. In these studies, detection is mainly achieved by determining changes in pixel intensity and color, which can be done manually or automatically. For images, automated methods performed using edge detection algorithms or segmentation techniques are faster than manual methods. However, these automated methods are limited by common problems related to the 2D nature of digital images. For example: 1. A single image may be obscured by protrusions, making rock sections invisible. Therefore, results based on a single image may have low integrity. 2. Trajectory detection is based on the color and intensity of data contained in the image and is greatly affected by lighting conditions and shadows. 3. The image range is limited and always relatively small. Therefore, the range of rock mass that can be processed at one time is very small. Therefore, in recent years, many researchers have studied the extraction of discontinuity surface traces from 3D point clouds of rock mass surfaces, mainly divided into four categories: The first category is that discontinuity traces can be obtained through the intersection lines between discontinuous surfaces on the rock mass surface. Therefore, before using these methods, it is necessary to extract discontinuities based on cluster analysis or region growing algorithms. The second category is that traces can be detected based on the principal curvature of vertices on the digital surface model (DSM) of the rock mass. The third type involves obtaining traces through human-computer interaction. The fourth type utilizes deep learning algorithms to identify rock lines within rock masses. However, the results of the first method are highly dependent on the accuracy of the extracted discontinuous planes or fitted planes, while the results of the second method largely depend on the mesh quality and smoothness. The third type of method is not easy to operate and requires professional operators. The fourth type requires manual operation based on the supervision of domain experts to create a certain number of training sets, and collecting raw data is relatively difficult. Although the above techniques can extract discontinuous features of dangerous rock masses to some extent, their accuracy and efficiency still need improvement. Improving the accuracy and robustness of edge extraction is a challenge. Furthermore, the traditional median transformation method has potential in terms of edge line extraction accuracy, but it is difficult to calculate the true median point in complex scenes. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned problems in the existing technology by providing a method and device for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: a method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation, including the following steps: Step 1, extracting the three-dimensional point cloud data of the target area with normals. The 3D point cloud data was obtained by using the median transformation method. Each point in The final central sphere is the point corresponding to the final central sphere selected based on a set radius threshold. Candidate points for discontinuous edge lines of rock mass .

[0006] Step 2, use Mean clustering method for 3D point cloud data Perform clustering to obtain 3D point cloud cluster data. .

[0007] Step 3: Calculate the selected 3D point cloud cluster data Candidate points of discontinuous edge lines curvature Based on the region growing algorithm for 3D point cloud cluster data The data is divided to obtain the data of each 3D point cloud cluster. Corresponding dividing area , This is the number of the dividing section.

[0008] Step 4: Extract 3D point cloud cluster data Extract the separator region from the corresponding two-dimensional boundary points of the cluster. The corresponding two-dimensional boundary points of the region are used to obtain the two-dimensional boundary points of the cluster and the three-dimensional boundary points of the region.

[0009] Step 5: Merge the cluster 3D boundary points and the zone 3D boundary points to obtain the final discontinuity line of the unstable rock mass.

[0010] Candidate points of discontinuous edge lines of the rock mass The following steps are used to obtain the data: Step 1.1, calculate the 3D point cloud data. Each point in normal .

[0011] Step 1.2: For each point Establish an initial central sphere. ,point On the surface of the central sphere, where, The center of the initial central sphere, The initial radius of the central sphere.

[0012] Step 1.3, Find the point The iterative correlation of nearest neighbors for the central sphere Perform iterative shrinkage until the point is finally obtained. The final central axis ball and points The final nearest neighbor , The center of the final axis sphere, This is the radius of the final central sphere.

[0013] Step 1.4, Selecting Points Time vector and points Time vector The angle formed is used as a point The final nearest neighbor angle of the central axis.

[0014] Step 1.5: Traverse the 3D point cloud data Each point in Get each point The corresponding final central axis point The final radius of the central sphere and points The final bisector of the nearest neighbor angle of the central axis .

[0015] Step 1.6: Filter out the 3D point cloud data Each point in The final radius of the central sphere The final central sphere that is smaller than a set radius threshold, and the points corresponding to the final central sphere selected. Candidate points for discontinuous edge lines of rock mass .

[0016] The progressive iterative contraction includes the following steps: in the... In the subsequent shrinking process, the 3D point cloud data is found through nearest neighbor queries. midpoint The corresponding number The central sphere of the next iteration Mid-range The center of the central sphere in the next iteration nearest point At point and the center of the ball The line Find the center of the ball. , make point and the center of the ball Length of the connected line segment ,point and the center of the ball Length of the connected line segment and the The radius of the axial sphere in the next iteration Same, with the center of the ball For the first The center of the central sphere in the next iteration For the first The radius of the central sphere in the next iteration will be the same as the initial central sphere. Conduct the first The th iteration of contraction yields the th The central sphere of the next iteration And calculate the first Next iteration shrinkage distance For the center of the ball and the center of the ball The distance between them.

[0017] when , or click and points Overlap, or points and points When they coincide, stop iterating. At this point, the first iteration... The central sphere of the next iteration For point The final central axis ball , No. The point of the next iteration For point The corresponding final central axis point , No. The next iteration For point The final nearest neighbor .

[0018] The use Mean clustering method for 3D point cloud data Clustering includes the following steps: Step 2.1, randomly select 3D point cloud data. In Each point serves as the initial cluster centroid, and each cluster centroid corresponds to a data cluster.

[0019] Step 2.2: Transfer the 3D point cloud data Each point in Assign it to the nearest cluster centroid.

[0020] Step 2.3: Recalculate the cluster centroid of each data cluster obtained in Step 2.2.

[0021] Step 2.4: Repeat steps 2.2 and 2.3 until the cluster centroid no longer changes or the preset number of iterations is reached.

[0022] The dividing area The following steps are used to obtain the following information: Step 3.1, candidate points through discontinuous edge lines. normal Calculate the selected 3D point cloud cluster data Candidate points of discontinuous edge lines curvature And candidate points for discontinuous edge lines By curvature Sort from lowest to highest.

[0023] Step 3.2: Select the first candidate point of the discontinuous edge line after sorting. Seed point as the initial dividing region Add to seed point set In the middle, the initial seed point As the current seed point .

[0024] Step 3.3: Using the current seed point Using 3D point cloud data centered at a set multiple, Points within a radius of resolution are the corresponding nearest neighbors. Calculate each neighboring point normal and current seed point normal If the angle between the two points is less than the set growth threshold, then the corresponding neighboring points will be... Added as a growth point to the growth point set middle.

[0025] Step 3.4: Calculate the number of points not added to the growth point set. Each neighboring point in The normal and the initial seed point normal The angle between them; if the angle is less than the set seed threshold, the corresponding neighboring points... As a new seed point Added to the seed point set middle.

[0026] Step 3.5: Set the current seed point From the seed point set Remove from the seed point set Select one seed point as the new current seed point. Repeat steps 3.3 to 3.4 until the seed point set is reached. It is empty and cannot add new seed points. This completes the division of a partition.

[0027] Step 3.6: Transfer the 3D point cloud cluster data Remaining candidate points for discontinuous edge lines By curvature Reorder the data from low to high, repeating steps 3.2 to 3.5 until the 3D point cloud cluster data is obtained. Each point in the array is assigned to a corresponding partition. middle.

[0028] Step 3.7: Select the next 3D point cloud cluster data. Repeat steps 3.1 to 3.6 until all 3D point cloud cluster data have been traversed. Complete the processing of all 3D point cloud cluster data Separate to obtain data for each 3D point cloud cluster. Corresponding dividing area .

[0029] The following steps are taken for the cluster 3D boundary points and region 3D boundary points: Step 4.1, calculate the data of each 3D point cloud cluster. All points in The final bisector of the nearest neighbor angle of the central axis Construct a two-dimensional projection plane using the average direction of the bisector of the normal to the two-dimensional projection plane. The average direction is used to create a projection matrix and each 3D point cloud cluster data is then processed. Projecting the data onto a two-dimensional projection plane yields the data for each three-dimensional point cloud cluster. From a 2D planar point cloud, calculate the 2D boundary points of the 2D planar point cloud as 3D point cloud cluster data. The corresponding two-dimensional boundary point of the cluster.

[0030] Step 4.2: Divide each section The 3D point cloud was fitted to principal component vectors using principal component analysis. The two principal component vectors obtained after the principal component analysis fitting form the dividing region. The corresponding fitting plane will separate the regions. The 3D point cloud is transformed onto the corresponding fitting plane using a projection matrix, and the segmentation region is then processed. The projection onto the fitting plane is calculated to obtain the dividing region. The two-dimensional boundary points are used as the two-dimensional boundary points of the region.

[0031] Step 4.3: Obtain the cluster's two-dimensional boundary points in the three-dimensional point cloud cluster data. The corresponding 3D boundary points of the cluster are obtained from the 2D boundary points of the region in the 3D point cloud cluster data. The corresponding three-dimensional boundary points of the region.

[0032] The final discontinuity line of the unstable rock mass is obtained based on the following steps: the cluster three-dimensional boundary points and the area three-dimensional boundary points are merged and duplicates are removed to obtain the fusion points, and the discrete fusion points are removed to obtain the final discontinuity line of the unstable rock mass.

[0033] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the above-described method for extracting discontinuous edge points of unstable rock masses based on improved central axis transformation.

[0034] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for extracting discontinuous edge points of unstable rock masses based on improved central axis transformation.

[0035] A computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described method for extracting discontinuous edge points of unstable rock masses based on improved central axis transformation.

[0036] Compared with existing technologies, this invention has the following advantages: 1. Significantly improves extraction accuracy and reliability; In the field of extracting discontinuous traces of unstable rock masses, traditional manual extraction methods are not only time-consuming but also easily affected by subjective factors. This invention, however, utilizes median transformation, point cloud segmentation, region growing algorithms, and median transformation features (candidate points for discontinuous edge lines obtained through median transformation). curvature The invention employs auxiliary segmentation and multi-level joint processing of the alpha-shape algorithm to achieve high-precision extraction of discontinuous traces in rock masses. Experimental data shows that for points 1, 2, 3, 4, 5, and 6, the distance deviations of points 1 to 6 obtained using this invention are 0.00631 meters, 0.00474 meters, 0.00349 meters, 0.00719 meters, 0.02366 meters, and 0.0032 meters, respectively. The standard deviations of points 1 to 6 obtained using this invention are 0.00937 meters, 0.00875 meters, 0.00747 meters, 0.01561 meters, 0.13631 meters, and 0.00735 meters, respectively. The average distance deviation is only 0.008 meters, and the average standard deviation is 0.03081 meters. The distance deviations of points 1-6 obtained using the Compass plugin are 0.84703 meters, 1.513 meters, 1.19933 meters, 0.52608 meters, 0.5181 meters, and 0.77768 meters, respectively; the standard deviations of points 1-6 obtained using the Compass plugin are 0.60943 meters, 10.7363 meters, 0.67237 meters, 0.6033 meters, 0.70618 meters, and 0.76893 meters, respectively. These results significantly outperform existing mainstream tools (such as CloudCompare's Compass plugin) and traditional edge detection methods in terms of accuracy and stability. This addresses the industry pain points of traditional image processing methods, such as the need for manual threshold adjustment and high uncertainty in results.

[0037] 2. Achieving a high degree of automation and intelligence: This invention achieves full automation from point cloud data to the extraction of discontinuous edge points of unstable rock masses, effectively avoiding the problems of low efficiency and easy deviation in traditional manual on-site measurement methods. At the same time, it also overcomes the limitations of poor model adaptability and poor accuracy caused by relying solely on one type of fracture morphology for characterization, reduces over-reliance on expert experience, and improves the objectivity and repeatability of the results.

[0038] 3. Enhanced adaptability in complex environments: Through strategies such as point cloud segmentation and regional growth, this invention can better address the actual situation of large differences in point cloud data density among different rock masses; it reduces the stringent requirements for the uniformity of initial data and improves robustness and universality under complex geological conditions and different acquisition environments. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0040] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to embodiments. It should be understood that the embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0041] Example 1

[0042] like Figure 1 As shown, the method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation includes the following steps.

[0043] Step 1: Obtain the 3D point cloud data of the target area with normals. The 3D point cloud data was obtained by using the median transformation method. Each point in The final central sphere is the point corresponding to the final central sphere selected based on a set radius threshold. Candidate points for discontinuous edge lines of rock mass .

[0044] Specifically, candidate points for the discontinuous edge lines of the rock mass The following steps are used to obtain the data: Step 1.1: Calculate the 3D point cloud data using CloudCompare software. Each point in normal .

[0045] 3D point cloud data The data used is a publicly available 3D point cloud dataset, obtained from the internet. This embodiment uses the 3D point cloud data. This is point cloud data of a rock slope near Highway 15 in Ontario, Canada. (The last part, "3D point cloud data," appears to be an unrelated fragment and is omitted from the translation.) This is laser scanning data with a resolution of approximately 1 cm; it is a 3D point cloud data. This 3D point cloud data contains 2,087,187 points. The scan was performed from a distance of approximately 15 meters from the target, and the entire study area measures 13.28 meters × 4.21 meters × 3.71 meters. In this embodiment, the verification results were obtained by cropping the original point cloud data, retaining only a portion as the target area for testing. In other words, the original data was quite large, and only a portion was used.

[0046] Step 1.2: For each point Establish an initial central sphere. ,point On the surface of the central sphere. Among them, Let be the initial radius of the central sphere. The initial radius of the central sphere can be set to any large value, and this radius will be gradually reduced in subsequent iterations. The number of iterations for the radius of the central sphere ( That is, the initial number of iterations. The center of the initial central sphere, Also a point The corresponding initial midline point. Definition point. Coordinates are The initial radius of the central sphere is ,point The unit normal vector on the surface of the central sphere is ; center of the ball coordinates With point coordinate Based on the following relationship: .

[0047] That is: point normal The line passing through the center of the initial central sphere. ,point normal Direction from point The center of the sphere pointing towards the initial central axis ,point to the center of the initial central axis sphere The distance is .

[0048] Step 1.3, Find the point The iterative correlation of nearest neighbors, for the initial central sphere. Perform iterative shrinkage until the point is finally obtained. The final central axis ball and points The final nearest neighbor .

[0049] Find the 3D point cloud data using nearest neighbor query. midpoint The corresponding initial central sphere The center of the ball at the mid-range initial axis nearest point , For point The first iteration relates to the nearest neighbor.

[0050] At point and the center of the ball The line Find the center of the ball. , make point and the center of the ball Length of the connected line segments ,point and the center of the ball The length of the connected line segment And the radius of the axial sphere in the first iteration. Same, that is , with point The center of the central sphere in the first iteration. Let be the radius of the central sphere in the first iteration, for the initial central sphere. The first iteration of shrinkage yields the central sphere of the first iteration. And calculate the shrinkage distance in the first iteration. For the center of the ball and the center of the ball The distance between them.

[0051] Repeat the above process, then the first... The shrinking process in the next iteration is as follows: find the 3D point cloud data through nearest neighbor query. midpoint The corresponding number The central sphere of the next iteration Mid-range The center of the central sphere in the next iteration nearest point At point and the center of the ball The line Find the center of the ball. As the first The center of the central sphere in the next iteration makes the point and the center of the ball Length of the connected line segment ,point and the center of the ball Length of the connected line segment and the The radius of the axial sphere in the next iteration Same, that is With the center of the ball For the first The center of the central sphere in the next iteration For the first The radius of the central sphere in the next iteration will be the same as the initial central sphere. Conduct the first The th iteration of contraction yields the th The central sphere of the next iteration And calculate the first Next iteration shrinkage distance For the center of the ball and the center of the ball The distance between them.

[0052] when , or click and points Overlap, or points and points When they coincide, stop iterating. At this point, the first iteration... The central sphere of the next iteration For point The final central axis ball , No. The center of the ball in the next iteration For point The corresponding final central axis point , The center of the final axis sphere, For the final radius of the central sphere, the first... The next iteration For point The final nearest neighbor .

[0053] Step 1.4, Selecting Points Time vector and points Time vector The angle formed is used as a point The final nearest neighbor angle of the central axis.

[0054] Step 1.5: Traverse the 3D point cloud data Each point in Get each point The corresponding final central axis point The final radius of the central sphere and points The final bisector of the nearest neighbor angle of the central axis .

[0055] Step 1.6: Filter out the 3D point cloud data Each point in The final radius of the central sphere The final central sphere is smaller than a set radius threshold (in this embodiment, the set radius threshold is 0.2m, which can typically be set to the 3D point cloud data). The final points corresponding to the central axis sphere (filtered using twice the resolution) Candidate points for discontinuous edge lines of rock mass .

[0056] Step 2, use Mean clustering method for 3D point cloud data Point cloud segmentation is performed to obtain multiple data clusters, and each data cluster constitutes a corresponding 3D point cloud cluster data. .

[0057] Specifically, use Mean clustering method for 3D point cloud data Dividing the data into multiple clusters includes the following steps: Step 2.1, randomly selecting 3D point cloud data. In Each point serves as the initial cluster centroid, and each cluster centroid corresponds to a data cluster. In this embodiment, The value is 50.

[0058] Step 2.2: Transfer the 3D point cloud data Each point in Assign it to the nearest cluster centroid.

[0059] For each point Calculate the distance from each point to the centroid of each cluster. The data cluster is assigned to the nearest cluster centroid.

[0060] Step 2.3: Recalculate the cluster centroid of each data cluster obtained in Step 2.2. The coordinates of the recalculated cluster centroid are the average of the coordinates of each point in the corresponding data cluster.

[0061] Step 2.4: Repeat steps 2.2 and 2.3 until the cluster centroid no longer changes or the preset number of iterations is reached.

[0062] Through iterative optimization, the 3D point cloud data All points The data clusters that minimize the sum of their distances from the centroids of their respective data clusters constitute a 3D point cloud cluster data. Complete the processing of 3D point cloud data The data is divided into blocks. 3D point cloud cluster data. It includes information such as point cloud coordinates, normals, and the angle bisector of the final nearest neighbor angle along the median. This is the block number.

[0063] Due to discontinuous edge candidate points For special points When 3D point cloud data When completing the block division, candidate points for each discontinuous edge line Different 3D point cloud cluster data were then assigned. middle.

[0064] Step 3: Calculate the selected 3D point cloud cluster data Candidate points of discontinuous edge lines curvature Based on the region growing algorithm for 3D point cloud cluster data The data is divided to obtain the data of each 3D point cloud cluster. Corresponding dividing area .

[0065] Specifically, based on the region growing algorithm, 3D point cloud cluster data The separation process includes the following steps: Step 3.1, selecting candidate points through discontinuous edge lines. normal Calculate the selected 3D point cloud cluster data Candidate points of discontinuous edge lines curvature And candidate points for discontinuous edge lines By curvature Sort from lowest to highest.

[0066] Step 3.2: Select the first candidate point of the discontinuous edge line after sorting. Seed point as the initial dividing region Add to seed point set In the middle, the initial seed point As the current seed point Initial seed point The starting point for dividing this area.

[0067] Step 3.3: For the seed point set Current seed point The current seed point will be used. Centered on a circle, using 10 times the amount of 3D point cloud data The points within a resolution of a radius (in this embodiment, the radius is set to 1m) are the current seed points. Neighboring points Calculate the current seed point Each neighboring point The normal and the current seed point normal The angle between them, if the angle is less than the set growth threshold Then the corresponding neighboring points Added as a growth point to the growth point set In this context, growth points are used to expand the current plane.

[0068] Step 3.4: Then calculate the sets of points not added to the growth point set. Each neighboring point in The normal and the initial seed point normal The angle between them, if the angle is less than the set seed threshold Then the corresponding neighboring points As a new seed point Added to the seed point set In the middle, the new seed point Used to guide new growth directions.

[0069] Step 3.5: After processing the current seed point All neighboring points Then, set the current seed point From the seed point set Remove from the seed point set. Select one seed point as the new current seed point. Repeat steps 3.3 to 3.4 until the seed point set is reached. It is empty and cannot add new seed points. At this point, a partition has been completed, which includes all the regions added to the growth point set. The growth point and all points that have been added to the seed point set Seed point in.

[0070] Step 3.6: Obtain a complete partition. Then, the corresponding 3D point cloud cluster data Remaining candidate points for discontinuous edge lines By curvature Reorder the data from low to high, repeating steps 3.2 to 3.5 until the 3D point cloud cluster data is obtained. Each point in the array is assigned to a corresponding partition. Among them For dividing area The number.

[0071] Step 3.7: Select the next 3D point cloud cluster data. Perform steps 3.1 to 3.6 until all 3D point cloud cluster data have been traversed. Complete the processing of all 3D point cloud cluster data Separate to obtain data for each 3D point cloud cluster. Corresponding dividing area .

[0072] Step 4: Extract 3D point cloud cluster data and dividing area The three-dimensional boundary point cloud.

[0073] Specifically, the Alphashape algorithm is used to process each 3D point cloud cluster data. and each partition The boundary extraction process is as follows: Step 4.1, calculate the data of each 3D point cloud cluster. All points in The final bisector of the nearest neighbor angle of the central axis Construct a two-dimensional projection plane using the average direction of the bisector of the normal to the two-dimensional projection plane. The average direction is used to create a projection matrix and each 3D point cloud cluster data is then processed. Projecting the data onto a two-dimensional projection plane yields the data for each three-dimensional point cloud cluster. The two-dimensional planar point cloud is used to calculate the two-dimensional boundary points of the two-dimensional planar point cloud as three-dimensional point cloud cluster data using the Alphashape algorithm. The corresponding two-dimensional boundary point of the cluster.

[0074] 3D point cloud cluster data All points in The final bisector of the nearest neighbor angle of the central axis The average of the three-dimensional vectors is calculated to obtain the corresponding average direction.

[0075] Step 4.2: Divide each section The 3D point cloud is fitted to principal component vectors using principal component analysis (PCA). The two principal component vectors obtained after PCA fitting form the dividing region. The corresponding fitting plane will separate the regions. The 3D point cloud is transformed onto the corresponding fitting plane using a projection matrix, and the segmentation region is then processed. The Alphashape algorithm is used to calculate the segmentation region by projecting the image onto the fitting plane. The two-dimensional boundary points are used as the two-dimensional boundary points of the region.

[0076] Step 4.3: Obtain the cluster's two-dimensional boundary points in the three-dimensional point cloud cluster data. The corresponding 3D boundary points of the cluster are obtained from the 2D boundary points of the region in the 3D point cloud cluster data. The corresponding three-dimensional boundary points of the region.

[0077] Step 5: Merge the cluster 3D boundary points and the area 3D boundary points and filter them to obtain the final discontinuity line of the unstable rock mass.

[0078] Specifically, the process of removing redundant boundary lines is as follows: Step 5.1: Merge the cluster 3D boundary points and the region 3D boundary points, filter out the points that overlap between the cluster 3D boundary points and the region 3D boundary points, and obtain the fusion points.

[0079] Step 5.2: After SOR (Successive Over-Relaxation) filtering to remove discrete fusion points, the final discontinuous edge points are obtained as the final discontinuity line of the unstable rock mass.

[0080] 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 the computer program is executed, it can include the processes of the embodiments of the above methods.

[0081] Example 2

[0082] In this embodiment, a computer device is also provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in Embodiment 1 above.

[0083] Example 3

[0084] In this embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps in Embodiment 1 above.

[0085] Example 4

[0086] In this embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in Embodiment 1 described above.

[0087] It should be noted that the embodiments described in this invention are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains can make various modifications or additions to the described embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for extracting discontinuous edge points of unstable rock masses based on improved median transformation, characterized in that, Includes the following steps: Step 1: Obtain the 3D point cloud data of the target area with normals. The 3D point cloud data was obtained by using the median transformation method. Each point in The final central sphere is the point corresponding to the final central sphere selected based on a set radius threshold. Candidate points for discontinuous edge lines of rock mass ; Step 2, use Mean clustering method for 3D point cloud data Perform clustering to obtain 3D point cloud cluster data. ; Step 3: Calculate the selected 3D point cloud cluster data Candidate points of discontinuous edge lines curvature Based on the region growing algorithm for 3D point cloud cluster data The data is divided to obtain the data of each 3D point cloud cluster. Corresponding dividing area , The number of the dividing section; Step 4: Extract 3D point cloud cluster data Extract the separator region from the corresponding two-dimensional boundary points of the cluster. The corresponding two-dimensional boundary points of the region are used to obtain the two-dimensional boundary points of the cluster and the three-dimensional boundary points of the region; Step 5: Merge the cluster 3D boundary points and the zone 3D boundary points to obtain the final discontinuity line of the unstable rock mass.

2. The method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation according to claim 1, characterized in that, Candidate points of discontinuous edge lines of the rock mass Obtain it based on the following steps: Step 1.1: Calculate the 3D point cloud data Each point in normal ; Step 1.2: For each point Establish an initial central sphere. ,point On the surface of the central sphere, where, The center of the initial central sphere, The initial radius of the central sphere; Step 1.3, Find the point The iterative correlation of nearest neighbors for the central sphere Perform iterative shrinkage until the point is finally obtained. The final central axis ball and points The final nearest neighbor , The center of the final axis sphere, The radius of the final central sphere; Step 1.4, Selecting Points Time vector and points Time vector The angle formed is used as a point The final nearest neighbor angle of the central axis; Step 1.5: Traverse the 3D point cloud data Each point in Get each point The corresponding final central axis point The final radius of the central sphere and points The final bisector of the nearest neighbor angle of the central axis ; Step 1.6: Filter out the 3D point cloud data Each point in The final radius of the central sphere The final central sphere that is smaller than a set radius threshold, and the points corresponding to the final central sphere selected. Candidate points for discontinuous edge lines of rock mass .

3. The method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation according to claim 2, characterized in that, The gradual iterative shrinkage includes the following steps: In the In the subsequent shrinking process, the 3D point cloud data is found through nearest neighbor queries. midpoint The corresponding number The central sphere of the next iteration Mid-range The center of the central sphere in the next iteration nearest point At point and the center of the ball The line Find the center of the ball. , make point and the center of the ball Length of the connected line segment ,point and the center of the ball Length of the connected line segment and the The radius of the axial sphere in the next iteration Same, with the center of the ball For the first The center of the central sphere in the next iteration For the first The radius of the central sphere in the next iteration will be the same as the initial central sphere. Conduct the first The th iteration of contraction yields the th The central sphere of the next iteration And calculate the first Next iteration shrinkage distance For the center of the ball and the center of the ball The distance between them; when , or click and points Overlap, or points and points When they coincide, stop iterating. At this point, the first iteration... The central sphere of the next iteration For point The final central axis ball , No. The point of the next iteration For point The corresponding final central axis point , No. The next iteration For point The final nearest neighbor .

4. The method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation according to claim 3, characterized in that, The use Mean clustering method for 3D point cloud data Clustering includes the following steps: Step 2.1: Randomly select 3D point cloud data In Each point serves as the initial cluster centroid, and each cluster centroid corresponds to a data cluster; Step 2.2: Transfer the 3D point cloud data Each point in Assigned to the nearest cluster centroid; Step 2.3: Recalculate the cluster centroid of each data cluster obtained in Step 2.2; Step 2.4: Repeat steps 2.2 and 2.3 until the cluster centroid no longer changes or the preset number of iterations is reached.

5. The method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation according to claim 4, characterized in that, The dividing area Obtain it based on the following steps: Step 3.1: Select candidate points through discontinuous edge lines. normal Calculate the selected 3D point cloud cluster data Candidate points of discontinuous edge lines curvature And candidate points for discontinuous edge lines By curvature Sort from lowest to highest; Step 3.2: Select the first candidate point of the discontinuous edge line after sorting. As the initial seed point for the dividing region Add to seed point set In the middle, the initial seed point As the current seed point ; Step 3.3: Using the current seed point Using 3D point cloud data centered at a set multiple, Points within a radius of resolution are the corresponding nearest neighbors. Calculate each neighboring point normal and current seed point normal If the angle between the two points is less than the set growth threshold, then the corresponding neighboring points will be... Added as a growth point to the growth point set middle; Step 3.4: Calculate the number of points not added to the growth point set. Each neighboring point in The normal and the initial seed point normal The angle between them; if the angle is less than the set seed threshold, the corresponding neighboring points... As a new seed point Added to the seed point set middle; Step 3.5: Set the current seed point From the seed point set Remove from the seed point set Select one seed point as the new current seed point. Repeat steps 3.3 to 3.4 until the seed point set is reached. It is empty and cannot add new seed points. This completes the division of a partition; Step 3.6: Transfer the 3D point cloud cluster data Remaining candidate points for discontinuous edge lines By curvature Reorder the data from low to high, repeating steps 3.2 to 3.5 until the 3D point cloud cluster data is obtained. Each point in the array is assigned to a corresponding partition. middle, Step 3.7: Select the next 3D point cloud cluster data Repeat steps 3.1 to 3.6 until all 3D point cloud cluster data have been traversed. Complete the processing of all 3D point cloud cluster data Separate to obtain data for each 3D point cloud cluster. Corresponding dividing area .

6. The method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation according to claim 5, characterized in that, The following steps are taken for the cluster 3D boundary points and region 3D boundary points: Step 4.1: Calculate the data for each 3D point cloud cluster. All points in The final bisector of the nearest neighbor angle of the central axis Construct a two-dimensional projection plane using the average direction of the bisector of the normal to the two-dimensional projection plane. The average direction is used to create a projection matrix and each 3D point cloud cluster data is then processed. Projecting the data onto a two-dimensional projection plane yields the data for each three-dimensional point cloud cluster. From a 2D planar point cloud, calculate the 2D boundary points of the 2D planar point cloud as 3D point cloud cluster data. The corresponding two-dimensional boundary point of the cluster; Step 4.2: Divide each section The 3D point cloud was fitted to principal component vectors using principal component analysis. The two principal component vectors obtained after the principal component analysis fitting form the dividing region. The corresponding fitting plane will separate the regions. The 3D point cloud is transformed onto the corresponding fitting plane using a projection matrix, and the segmentation region is then processed. The projection onto the fitting plane is calculated to obtain the dividing region. The two-dimensional boundary points are used as the two-dimensional boundary points of the region; Step 4.3: Obtain the cluster's two-dimensional boundary points in the three-dimensional point cloud cluster data. The corresponding 3D boundary points of the cluster are obtained, and the 2D boundary points of the region are represented in the 3D point cloud cluster data. The corresponding three-dimensional boundary points of the region.

7. The method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation according to claim 6, characterized in that, The final discontinuity line of the unstable rock mass is obtained based on the following steps: The cluster's 3D boundary points and the area's 3D boundary points are merged and duplicates are removed to obtain the fusion points. Discrete fusion points are then removed to obtain the final discontinuity line of the unstable rock mass.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method for extracting discontinuous edge points of dangerous rock masses based on improved central axis transformation as described in any one of claims 1 to 7.

Images