A coal mine tunnel point cloud grading denoising and surrounding rock structure extraction method and system

Abstract

The application discloses a coal mine roadway point cloud grading denoising and surrounding rock structure extraction method and system, relates to the field of intelligent sensing of coal mines, and comprises the following steps: obtaining original three-dimensional point cloud data of a coal mine roadway, performing voxelization downsampling and statistical outlier removal, eliminating random sparse noise points, and obtaining a pretreated point cloud set; calculating normal vectors of each point in the point cloud set, solving and dividing the point cloud into a floor, a roof and two candidate sets of sides; performing density-based clustering analysis on the candidate sets respectively, retaining main surrounding rock structure clusters, and removing small and medium-sized independent suspended facility clusters, to obtain a purified component point cloud; constructing a robust shell filtering model based on angular sector statistical characteristics, performing axial slicing and polar coordinate transformation on the component point cloud, removing pipelines close to the sides of the roadway and spatial placeholder targets located in the interior of the roadway, and generating a final roadway surrounding rock structure point cloud.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing in coal mines, and in particular to a method and system for hierarchical noise reduction of point clouds and extraction of surrounding rock structure in coal mine roadways. Background Technology

[0002] With the continuous advancement of coal mining technology and the deepening of intelligent construction, building high-precision "transparent mines" has become an urgent need for the industry's development. As the main channels for underground production, transportation, and ventilation in coal mines, the accurate perception and three-dimensional reconstruction of the spatial morphology of roadways are key links in realizing mine safety monitoring and surrounding rock stability analysis.

[0003] Currently, acquiring full-section data of roadways using mobile 3D laser scanning technology is an emerging technique. This technology can quickly acquire massive amounts of 3D point cloud data of the roadway surface, providing rich basic information for mine digitization. However, the underground environment of coal mines is a typical unstructured and complex scenario. In addition to containing effective information about the surrounding rock of the roadway, the original scanning data also contains extremely complex interference noise.

[0004] Current technologies have significant limitations when processing this type of data. On the one hand, underground roadways are typically equipped with numerous auxiliary facilities, such as ventilation ducts, water pipes, gas drainage pipes, and power cables suspended from the roof or walls, as well as belt conveyor frames laid on the floor. These facilities have high point cloud density and are closely connected to the surrounding rock, making it difficult for traditional density-based statistical filtering algorithms to distinguish them from the actual surrounding rock, resulting in severe distortion of the extracted roadway contours. On the other hand, there are often moving or occupying targets such as vehicles, pedestrians, and temporarily stockpiled materials inside the roadways. Existing geometric bounding box clipping or simple plane fitting methods are poorly adapted to roadway undulations, turns, or floor deformations, easily leading to the accidental deletion of real roadway surrounding rock data or the retention of internal interfering targets.

[0005] How to automatically and efficiently remove various types of noise, such as ventilation ducts, pipelines, machine frames, and moving vehicles, from the point cloud of complex coal mine roadways with high noise and interference, and accurately extract pure roadway surrounding rock structure data, is a technical challenge currently faced in the field of coal mine three-dimensional spatial perception and data processing. Summary of the Invention

[0006] This invention aims to at least solve one of the technical problems existing in the prior art. Therefore, one objective of this invention is to propose a method for hierarchical denoising and surrounding rock structure extraction of point cloud data in coal mine roadways. This method can significantly improve the processing accuracy and automation level of point cloud data in roadways under unstructured and complex environments, providing high-fidelity data support for the transparent construction of coal mine roadways and early warning of mine pressure disasters.

[0007] The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to the present invention includes the following steps: S10: Acquire the original three-dimensional point cloud data of the coal mine roadway, perform voxelization downsampling and statistical outlier removal, eliminate random sparse noise points caused by dust or data acquisition equipment errors, and obtain the preprocessed point cloud set P. S20: Calculate the normal vector of each point in the point cloud set P, and based on the projection component characteristics of the normal vector on the vertical coordinate axis, solve the point cloud and divide it into bottom plate candidate set, top plate candidate set and two side plate candidate set. S30: Perform density-based clustering analysis on the candidate sets of the bottom plate, the candidate sets of the top plate, and the candidate sets of the two side walls respectively. Based on the connectivity and scale characteristics of the point cloud clusters, retain the surrounding rock structure clusters and remove the small and medium-sized independent suspended facility clusters that have spatial gaps with the surrounding rock of the roadway or have a low percentage of points. Complete the denoising of the small and medium-sized independent suspended facility clusters to obtain the purified component point cloud P. refined The cluster of small and medium-sized independent suspended facilities includes a wind tunnel and a belt conveyor frame; S40: Construct a robust shell filtering model based on the statistical characteristics of the corner sector, and apply it to the point cloud P of the component. refined By performing axial slicing and polar coordinate transformation, and calculating the boundary radius quantiles within each corner sector, the outermost surrounding rock points of the roadway are identified and retained. Pipelines close to the roadway walls and spatially occupying targets located inside the roadway are removed, generating the final roadway surrounding rock structure point cloud P. final The space-occupying targets include vehicles, large work equipment in the alley, and pedestrians.

[0008] In some embodiments of the present invention, step S10 specifically includes: S11: Use a voxel grid filter to downsample the original point cloud and set the voxel size; S12: Using a statistical outlier removal algorithm, calculate the average distance from each point to its nearest multiple neighboring points, and calculate the mean and standard deviation of these average distances; S13: Remove points whose average distance is greater than the sum of the mean and twice the standard deviation to obtain a clean and uniform preprocessed point set P.

[0009] In some embodiments of the present invention, step S20 includes: For any point p in the point cloud set P i Calculate its normal vector And set the threshold T for the vertical component of the normal. z T z Greater than 0: If n z >T z Determine p iThese are the base plate points, used for subsequent extraction of pure base plates; If n z <-T z This point is identified as the top plate point and used for subsequent extraction of the pure top plate. If |n z |≤T z These are identified as two support points, used for subsequent extraction of the pure left and right support points.

[0010] In some embodiments of the present invention, step S30 includes: Density-based clustering algorithm was used to perform clustering analysis on the candidate sets of the bottom plate, the top plate, and the two side plates, resulting in multiple independent connected clusters. ; Count the number of points N(c) contained in each of the independent connected clusters. j Determine the maximum number of cluster points N under the current category. max =max(N(c j )); Set a size ratio threshold λ, where 0 < λ < 1, and calculate the dynamic retention threshold N. thresh =N max ×λ; Filter out those that satisfy N(c) j )≥N thresh The aforementioned independent connected clusters, as effective surrounding rock structures, will satisfy N(c j )<N thresh The independent connected clusters are identified as noise points in the clusters of small and medium-sized independent suspended facilities and are removed to obtain the purified component point cloud P. refined .

[0011] In some embodiments of the present invention, step S40 includes: S41: Identify the purified component point cloud P refined The main extension axis A, and along this axis with a predetermined thickness d slice Discretize the point cloud into continuous cross-sectional slices S k ; S42: Calculate slice S k Geometric center O k Project the points inside the slice onto a two-dimensional plane and establish a plane with O k A local polar coordinate system with the origin at the origin; S43: Divide the two-dimensional plane into M uniform angular sectors, calculate the polar angle θ of each point (u,v) within the slice, and determine its sector index idx: S44: Within each sector idx, obtain the set of polar radii for all points. The radius of the surrounding rock boundary of the sector is calculated using the preset high percentile value α. : S45: Set the outer casing thickness δ, polar diameter The points that occupy the internal space are the target points; these are eliminated, and only the points that are retained are left. The point P is used as the final point cloud of the surrounding rock structure of the tunnel. final .

[0012] In some embodiments of the present invention, in step S43, the expression for the sector index idx is: Where θ∈[-π, π], This indicates rounding down to the nearest integer.

[0013] In some embodiments of the present invention, the radius of the surrounding rock boundary of the sector The expression is: In the formula, This indicates the calculation of the αth percentile of a given set of polar radii.

[0014] In some embodiments of the present invention, in step S44, the percentile value α ranges from 95 to 99.

[0015] Another objective of this invention is to provide a system for hierarchical noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways, comprising: The data processing module is configured to acquire the original three-dimensional point cloud data of the coal mine roadway, perform voxelization downsampling and statistical outlier removal, eliminate random sparse noise caused by dust or data acquisition equipment errors, and obtain a preprocessed point cloud set P. The solution and partitioning module is configured to calculate the normal vector of each point in the point cloud set P, and based on the projection component characteristics of the normal vector on the vertical coordinate axis, solve and partition the point cloud into a bottom plate candidate set, a top plate candidate set, and two side plate candidate sets. The clustering analysis module is configured to perform density-based clustering analysis on the candidate sets of the floor, roof, and sidewalls, respectively. Based on the connectivity and scale characteristics of the point cloud clusters, it retains the surrounding rock structure clusters and removes small to medium-sized independent suspended facility clusters that have spatial gaps with the surrounding rock of the roadway or have a low percentage of points. This completes the denoising of the small to medium-sized independent suspended facility clusters, obtaining the purified component point cloud P. refined The cluster of small and medium-sized independent suspended facilities includes a wind tunnel and a belt conveyor frame; The structure point cloud generation module is configured to construct a robust shell filtering model based on the statistical characteristics of corner sectors for the component point cloud P. refinedBy performing axial slicing and polar coordinate transformation, and calculating the boundary radius quantiles within each corner sector, the outermost surrounding rock points of the roadway are identified and retained. Pipelines close to the roadway walls and spatially occupying targets located inside the roadway are removed, generating the final roadway surrounding rock structure point cloud P. final The space-occupying targets include vehicles, large work equipment in the alley, and pedestrians.

[0016] In some embodiments of the present invention, the clustering analysis module includes: The independent connected cluster unit is configured to perform cluster analysis on the candidate sets of the bottom plate, the candidate sets of the top plate, and the candidate sets of the two sides using a density-based clustering algorithm to obtain multiple independent connected clusters. ; The maximum cluster point number determination unit is configured to count the number of points N(c) contained in each of the independent connected clusters. j Determine the maximum number of cluster points N under the current category. max =max(N(c j )); The dynamic retention threshold unit is configured to set a size ratio threshold λ, where 0 < λ < 1, and calculate the dynamic retention threshold N. thresh =N max ×λ; Component point acquisition unit, configured to filter out those satisfying N(c j )≥N thresh The aforementioned independent connected clusters, as effective surrounding rock structures, will satisfy N(c j )<N thresh The independent connected clusters are identified as noise points in the clusters of small and medium-sized independent suspended facilities and are removed to obtain the purified component point cloud P. refined .

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention significantly improves the processing accuracy and automation level of point cloud data for roadways in unstructured and complex environments. Through a hierarchical denoising mechanism of preprocessing, cluster purification, and shell filtering, it effectively solves the technical bottleneck of traditional single-filtering algorithms in handling sparse dust, high-density connected equipment, and internal occupiers. In particular, by utilizing dynamic ratio-based clustering analysis and a robust shell model based on the statistical characteristics of corner sectors, it achieves precise separation of pipeline facilities adhering to the surrounding rock and effectively removes moving targets within the roadway. This overcomes the poor adaptability of regular bounding box clipping to roadway deformation, ensuring the integrity and high fidelity of surrounding rock contour extraction under complex conditions such as roadway turns and floor heaves. This provides a high-quality data foundation for the transparent reconstruction of coal mine roadways and intelligent early warning of mine pressure disasters.

[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a method for hierarchical denoising and surrounding rock structure extraction of point clouds in coal mine roadways according to an embodiment of the present invention. Figure 2 This is an original point cloud map of a coal mine roadway with noise points such as ventilation ducts, belt conveyor frames, pipelines and equipment, according to an embodiment of the present invention. Figure 3 Point cloud image for voxelization downsampling and statistical outlier removal according to an embodiment of the present invention; Figure 4 This is a candidate set of point clouds of surrounding rock in a roadway based on the vertical component of the normal vector, according to an embodiment of the present invention. Figure 5 This is a point cloud diagram of the components purified by density clustering according to an embodiment of the present invention; Figure 6 This is a point cloud image after further denoising based on corner sector shell filtering according to an embodiment of the present invention; Figure label: 1. Random sparse noise point; 2. Ventilation duct; 3. Belt conveyor frame; 4. Pipeline; 5. Cable; 6. Large working equipment in the tunnel; 7. Floor; 8. Roof; 9. Left side; 10. Right side. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0021] The following is for reference. Figures 1-6 This invention describes a method for hierarchical denoising and surrounding rock structure extraction of point clouds in coal mine roadways according to embodiments of the present invention.

[0022] refer to Figure 1 As shown, the method for hierarchical denoising and surrounding rock structure extraction of point clouds in coal mine roadways according to the first aspect of the present invention includes the following steps: S10: Acquire the original 3D point cloud data of the coal mine roadway, perform voxelization downsampling and statistical outlier removal, eliminate random sparse noise points caused by dust or data acquisition equipment errors, and obtain a preprocessed point cloud set P, such as Figure 3 As shown; S20: Calculate the normal vector of each point in the point cloud set P. Based on the projection component characteristics of the normal vector on the vertical coordinate axis, solve and divide the point cloud into candidate sets for the bottom plate, top plate, and both sides, such as... Figure 4 As shown; S30: Perform density-based clustering analysis on the candidate sets for the floor, roof, and sidewalls respectively. Based on the connectivity and scale characteristics of the point cloud clusters, retain the main surrounding rock structure clusters (the main surrounding rock structure clusters refer to the skeleton of the roadway surrounding rock, including the surrounding rock structures of the roof, floor, and sidewalls), and remove small and medium-sized independent suspended facility clusters that have spatial gaps with the roadway surrounding rock or have a low percentage of points. Complete the denoising of the small and medium-sized independent suspended facility clusters to obtain the purified component point cloud P. refined ,like Figure 5 As shown. The cluster of small and medium-sized independent suspended facilities includes a ventilation duct 2 and a conveyor belt frame 3; S40: Construct a robust shell filtering model based on the statistical characteristics of the corner sector, and apply it to the component point cloud P. refined By performing axial slicing and polar coordinate transformation, and calculating the boundary radius quantiles within each corner sector, the outermost surrounding rock points of the roadway are identified and retained. Pipelines close to the roadway walls and spatially occupying targets located inside the roadway are removed, generating the final roadway surrounding rock structure point cloud P. final ,like Figure 6 As shown. Among them, the pipelines close to the tunnel wall include pipes 4 and cables 5 close to the tunnel wall, and the space occupants include vehicles, large working equipment 6 in the tunnel (e.g., mining transport vehicles, tunneling machines, anchor drilling rigs, crushers, transfer machines, belt conveyors, ventilation ducts, etc.) and pedestrians.

[0023] The proposed method for hierarchical denoising and surrounding rock structure extraction of coal mine roadway point clouds achieves a significant improvement in data utilization through a hierarchical processing mechanism of preprocessing, cluster purification, and shell filtering. It effectively solves the technical bottleneck of the difficulty in separating high-density wall-mounted facilities from the surrounding rock and realizes automated and high-fidelity extraction of roadway surrounding rock structures in complex environments.

[0024] In one example of the present invention, in step S10, S11: Use a voxel grid filter to downsample the original point cloud and set the voxel size; S12: Using a statistical outlier removal algorithm, calculate the average distance from each point to its nearest multiple neighboring points, and calculate the mean and standard deviation of these average distances; S13: Remove points whose average distance is greater than the sum of the mean and twice the standard deviation to obtain a clean and uniform preprocessed point set P.

[0025] To reduce subsequent computational load and remove sparse noise caused by downhole dust, a voxel grid filter was first used to downsample the original point cloud, with the voxel size set to 0.08m. Then, a statistical outlier removal algorithm was used to calculate the average distance from each point to its nearest k=20 neighboring points, calculate the mean and standard deviation of these average distances, and remove points whose average distance is greater than "mean + 2 times standard deviation" to obtain a clean and uniform preprocessed point cloud set P.

[0026] In one example of the present invention, step S20, constructing a normal classification model, specifically includes: The KD-Tree data structure is used to search for neighboring points. To accommodate the differences in point cloud density in different areas of the roadway, a hybrid search parameter is preferred, with a search radius of 0.5m and a maximum number of nearest neighbors of 50. Principal component analysis is used to estimate the value of each point p in the preprocessed point cloud. i normal vector And perform consistency adjustments for the normal direction. Set the threshold T for the vertical component of the normal. z T z Greater than 0.

[0027] Coarse classification based on normal characteristics: If the Z-axis component of the normal at a point is n... z >T z This indicates that the surface of this point is roughly facing upwards, and it is included in the candidate set of substrates for subsequent extraction of pure substrates; if n z <-T z This indicates that the surface of this point is roughly downward, and it is included in the candidate set of the top plate for subsequent extraction of the pure top plate; if its absolute value |n z |≤T z This indicates that the surface at this point is approximately perpendicular, and it is included in the candidate set of two sides for subsequent extraction of the pure left side 9 and right side 10. Optionally, T z It is 0.75 In one example of the present invention, in step S30, for small and medium-sized independent suspended facilities such as ventilation duct 2 and belt conveyor frame 3, which are close to the surrounding rock but usually have independent interconnected structures, a density-based clustering algorithm is used for purification. Specifically: The clustering neighborhood radius Eps = 0.35m and the minimum number of points MinPoints = 50 are set. After clustering each candidate set separately, multiple independent connected clusters are obtained. .

[0028] Count the number of points N(c) contained in each independent connected cluster. j Determine the maximum number of cluster points N under the current category. max =max(N(c j )).

[0029] Set a size ratio threshold λ (0 < λ < 1), and calculate the dynamic retention threshold N. thresh =N max ×λ. Optionally, the size scale threshold λ can be 0.5, and the dynamic retention threshold N is calculated. thresh =N max ×0.5.

[0030] The system automatically retains all points N(c) j )≥N thresh Large-sized independent connected clusters are considered as effective surrounding rock structure clusters, while the number of points N(c) is used as the basis for determining the effective surrounding rock structure clusters. j )<N thresh Small, independent, connected clusters were identified as noise points in small to medium-sized independent suspended facility clusters such as wind duct 2 or belt conveyor frame 3, and were removed to obtain the purified component point cloud P. refined ,like Figure 5 As shown.

[0031] In one example of the present invention, in step S40, in order to further eliminate pipes 4 and cables 5 close to the tunnel wall, as well as vehicles, large equipment 6 and pedestrians inside the tunnel, a robust shell filtering model is constructed, and the specific steps are as follows: S41: Calculate P refined The axis is aligned with the bounding box, its main extension axis X-axis is identified, and along this axis at a predetermined thickness d slice Slice the data to obtain a set of slices. Optionally, d slice It is 0.5m.

[0032] S42: For each slice S k Calculate its geometric center O k The three-dimensional points within the slice are projected onto a two-dimensional plane perpendicular to the X-axis and converted to a polar coordinate system (θ, r).

[0033] S43: Divide the two-dimensional plane into M uniform angular sectors. Optionally, M is 72. For each point within a slice, calculate its sector index idx: Where θ∈[-π, π], Indicates rounding down; S44: Within each sector idx, determine the location of the actual surrounding rock in the roadway using statistical quantile methods. Specifically, obtain the set of extreme radii for all points within that sector. The radius of the surrounding rock boundary of the sector is calculated using the preset high percentile value α. Its expression is: In the formula, This indicates the calculation of the αth percentile of a given set of polar radii.

[0034] Optionally, the percentile value α can range from 95 to 99. In a specific example, the percentile α could be 98. Using statistical quantiles instead of maximum values ​​to determine the boundary radius effectively prevents the boundary radius from being calculated too large due to a few extremely distant outliers. This eliminates the interference of individual outliers on boundary determination, ensuring the robustness of the shell extraction.

[0035] S45: Set the outer casing thickness to δ=0.2m, retaining only the extreme diameter. The point P is used as the final point cloud of the surrounding rock structure of the tunnel. final Polar diameter The points are the target points for occupancy in the internal space, while those considered "non-shell points" are precisely eliminated, ultimately generating a pure point cloud P that preserves the surrounding rock structure of the tunnel floor 7, roof 8, left side 9, and right side 10. final ,like Figure 6 As shown.

[0036] A coal mine roadway point cloud hierarchical denoising and surrounding rock structure extraction system according to a second aspect of the present invention includes: The data processing module is configured to acquire the original three-dimensional point cloud data of the coal mine roadway, perform voxelization downsampling and statistical outlier removal, eliminate random sparse noise caused by dust or data acquisition equipment errors, and obtain a preprocessed point cloud set P. The solution and partitioning module is configured to calculate the normal vector of each point in the point cloud set P, and based on the projection component characteristics of the normal vector on the vertical coordinate axis, solve and partition the point cloud into a bottom plate candidate set, a top plate candidate set, and two side plate candidate sets. The clustering analysis module is configured to perform density-based clustering analysis on the candidate sets of the floor, roof, and sidewalls, respectively. Based on the connectivity and scale characteristics of the point cloud clusters, it retains the main surrounding rock structure clusters and removes small and medium-sized independent suspended facility clusters that have spatial gaps with the surrounding rock of the roadway or have a low percentage of points. This completes the denoising of the small and medium-sized independent suspended facility clusters, obtaining the purified component point cloud P. refined The cluster of small and medium-sized independent suspended facilities includes a wind tunnel and a belt conveyor frame; The structure point cloud generation module is configured to construct a robust shell filtering model based on the statistical characteristics of corner sectors for the component point cloud P. refined By performing axial slicing and polar coordinate transformation, and calculating the boundary radius quantiles within each corner sector, the outermost surrounding rock points of the roadway are identified and retained. Pipelines close to the roadway walls and spatially occupying targets located inside the roadway are removed, generating the final roadway surrounding rock structure point cloud P. final The space-occupying targets include vehicles, large work equipment in the alley, and pedestrians.

[0037] The proposed method for hierarchical denoising and surrounding rock structure extraction of coal mine roadway point clouds, through a hierarchical processing mechanism of preprocessing, cluster purification, and shell filtering, significantly improves data utilization and effectively solves the technical bottleneck of the difficulty in separating high-density wall-mounted facilities from the surrounding rock. It also enables automated and high-fidelity extraction of roadway surrounding rock structures in complex environments.

[0038] In some embodiments of the present invention, the clustering analysis module includes: The independent connected cluster unit is configured to perform cluster analysis on the candidate sets of the bottom plate, the candidate sets of the top plate, and the candidate sets of the two sides using a density-based clustering algorithm to obtain multiple independent connected clusters. ; The maximum cluster point number determination unit is configured to count the number of points N(c) contained in each of the independent connected clusters. j Determine the maximum number of cluster points N under the current category. max =max(N(c j )); The dynamic retention threshold unit is configured to set a size ratio threshold λ, where 0 < λ < 1, and calculate the dynamic retention threshold N. thresh =N max ×λ; Component point acquisition unit, configured to filter out those satisfying N(c j )≥N thresh The aforementioned independent connected clusters, as effective surrounding rock structures, will satisfy N(c j )<N thresh The independent connected clusters are identified as noise points in the clusters of small and medium-sized independent suspended facilities and are removed to obtain the purified component point cloud P. refined .

[0039] It should be noted that the parameter values ​​in the above embodiments are preferred values ​​set for the specific resolution and noise level of the roadway point cloud data in this embodiment. In practical applications, those skilled in the art can adaptively adjust the above parameters according to the point cloud acquisition density, roadway cross-sectional dimensions, and noise distribution, which does not deviate from the core algorithm logic of this invention.

[0040] In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0041] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0042] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for hierarchical noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways, characterized in that, Includes the following steps: S10: Acquire the original three-dimensional point cloud data of the coal mine roadway, perform voxelization downsampling and statistical outlier removal, eliminate random sparse noise points caused by dust or data acquisition equipment errors, and obtain the preprocessed point cloud set P. S20: Calculate the normal vector of each point in the point cloud set P, and based on the projection component characteristics of the normal vector on the vertical coordinate axis, solve the point cloud and divide it into bottom plate candidate set, top plate candidate set and two side plate candidate set. S30: Perform density-based clustering analysis on the candidate sets of the bottom plate, the candidate sets of the top plate, and the candidate sets of the two sidewalls respectively. Based on the connectivity and scale characteristics of the point cloud clusters, retain the surrounding rock structure clusters and remove the small and medium-sized independent suspended facility clusters that have spatial gaps with the surrounding rock of the roadway or have a low percentage of points. Complete the denoising of the small and medium-sized independent suspended facility clusters to obtain the purified component point cloud P. refined The cluster of small and medium-sized independent suspended facilities includes a wind tunnel and a belt conveyor frame; S40: Construct a robust shell filtering model based on the statistical characteristics of the corner sector, and apply it to the point cloud P of the component. refined By performing axial slicing and polar coordinate transformation, and calculating the boundary radius quantiles within each corner sector, the outermost surrounding rock points of the roadway are identified and retained. Pipelines close to the roadway walls and spatially occupying targets located inside the roadway are removed, generating the final roadway surrounding rock structure point cloud P. final The space-occupying targets include vehicles, large work equipment in the alley, and pedestrians.

2. The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 1, characterized in that, Step S10 specifically includes: S11: Use a voxel grid filter to downsample the original point cloud and set the voxel size; S12: Using a statistical outlier removal algorithm, calculate the average distance from each point to its nearest multiple neighboring points, and calculate the mean and standard deviation of these average distances; S13: Remove points whose average distance is greater than the sum of the mean and twice the standard deviation to obtain a clean and uniform preprocessed point set P.

3. The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 1, characterized in that, Step S20 specifically includes: For any point p in the point cloud set P i Calculate its normal vector And set the threshold T for the vertical component of the normal. z T z Greater than 0: If n z >T z Determine p i These are the base plate points, used for subsequent extraction of pure base plates; If n z <-T z This point is identified as the top plate point and used for subsequent extraction of the pure top plate. If |n z |≤T z These are identified as two support points, used for subsequent extraction of the pure left and right support points.

4. The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 1, characterized in that, Step S30 includes: S31: A density-based clustering algorithm is used to perform clustering analysis on the candidate sets of the bottom plate, the top plate, and the two side plates to obtain multiple independent connected clusters. ; S32: Count the number of points N(c) contained in each of the independent connected clusters. j Determine the maximum number of cluster points N under the current category. max =max(N(c j )); S33: Set the size ratio threshold λ, where 0 < λ < 1, and calculate the dynamic retention threshold N. thresh =N max ×λ; S34: Filter out those that satisfy N(c) j )≥N thresh The aforementioned independent connected clusters, as effective surrounding rock structures, will satisfy N(c j ) < N thresh The independent connected clusters are identified as noise points in the clusters of small and medium-sized independent suspended facilities and are removed to obtain the purified component point cloud P. refined .

5. The method for hierarchical noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 1, characterized in that, Step S40 includes: S41: Identify the purified component point cloud P refined The main extension axis A, and along this axis with a predetermined thickness d slice Discretize the point cloud into continuous cross-sectional slices S k ; S42: Calculate slice S k Geometric center O k Project the points inside the slice onto a two-dimensional plane and establish a plane with O k A local polar coordinate system with the origin at the origin; S43: Divide the two-dimensional plane into M uniform angular sectors, calculate the polar angle θ of each point (u,v) within the slice, and determine its sector index idx: S44: Within each sector idx, obtain the set of polar radii for all points. The radius of the surrounding rock boundary of the sector is calculated using the preset high percentile value α. : S45: Set the outer casing thickness δ, polar diameter The points that occupy the internal space are the target points; these are eliminated, and only the points that are retained are left. The point P is used as the final point cloud of the surrounding rock structure of the tunnel. final .

6. The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 5, characterized in that, In step S43, the expression for the sector index idx is: Where θ∈[-π, π], This indicates rounding down to the nearest integer.

7. The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 5, characterized in that, radius of the surrounding rock boundary of the sector The expression is: In the formula, This indicates the calculation of the αth percentile of a given set of polar radii.

8. The method for graded noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways according to claim 5, characterized in that, In step S44, the percentile value α ranges from 95 to 99.

9. A system for hierarchical noise reduction and surrounding rock structure extraction of point clouds in coal mine roadways, characterized in that, include: The data processing module is configured to acquire the original three-dimensional point cloud data of the coal mine roadway, perform voxelization downsampling and statistical outlier removal, eliminate random sparse noise caused by dust or data acquisition equipment errors, and obtain a preprocessed point cloud set P. The solution and partitioning module is configured to calculate the normal vector of each point in the point cloud set P, and based on the projection component characteristics of the normal vector on the vertical coordinate axis, solve and partition the point cloud into a bottom plate candidate set, a top plate candidate set, and two side plate candidate sets. The clustering analysis module is configured to perform density-based clustering analysis on the candidate sets of the floor, roof, and sidewalls, respectively. Based on the connectivity and scale characteristics of the point cloud clusters, it retains the surrounding rock structure clusters and removes small to medium-sized independent suspended facility clusters that have spatial gaps with the surrounding rock of the roadway or have a low percentage of points. This completes the denoising of the small to medium-sized independent suspended facility clusters, obtaining the purified component point cloud P. refined The cluster of small and medium-sized independent suspended facilities includes a wind tunnel and a belt conveyor frame; The structure point cloud generation module is configured to construct a robust shell filtering model based on the statistical characteristics of corner sectors for the component point cloud P. refined By performing axial slicing and polar coordinate transformation, and calculating the boundary radius quantiles within each corner sector, the outermost surrounding rock points of the roadway are identified and retained. Pipelines close to the roadway walls and spatially occupying targets located inside the roadway are removed, generating the final roadway surrounding rock structure point cloud P. final The space-occupying targets include vehicles, large work equipment in the alley, and pedestrians.

10. The coal mine roadway point cloud hierarchical noise reduction and surrounding rock structure extraction system according to claim 9, characterized in that, The cluster analysis module includes: The independent connected cluster unit is configured to perform cluster analysis on the candidate sets of the bottom plate, the candidate sets of the top plate, and the candidate sets of the two sides using a density-based clustering algorithm to obtain multiple independent connected clusters. ; The maximum cluster point number determination unit is configured to count the number of points N(c) contained in each of the independent connected clusters. j Determine the maximum number of cluster points N under the current category. max =max(N(c j )); The dynamic retention threshold unit is configured to set a size ratio threshold λ, where 0 < λ < 1, and calculate the dynamic retention threshold N. thresh =N max ×λ; Component point acquisition unit, configured to filter out those satisfying N(c j )≥N thresh The aforementioned independent connected clusters, as effective surrounding rock structures, will satisfy N(c j ) < N thresh The independent connected clusters are identified as noise points in the clusters of small and medium-sized independent suspended facilities and are removed to obtain the purified component point cloud P. refined .

Images