Automatic extraction method of linear pipeline point cloud in mine tunnels based on region growing

Abstract

The present invention discloses a method for automatically extracting point clouds of linear pipelines in mine shafts and lanes based on regional growth, which belongs to the technical field of point cloud processing and is used for point cloud extraction, including implementing ground point filtering on pre-processed point cloud data, projecting the extracted tunnel ground point cloud onto a two-dimensional fitting plane, constructing a point cloud convex hull and calculating the area of ​​the point cloud convex hull, combining the number of tunnel ground point clouds and the point cloud convex hull, calculating the point cloud density, estimating the maximum clustering point number parameter based on the point cloud density and the surface area of ​​the tunnel wall, performing normal vector smoothing threshold and curvature restriction on the point cloud data based on the regional growth method, introducing an iterative processing mechanism in the regional growth method, and segmenting and extracting a pipeline point cloud set. The present invention combines a ground filtering algorithm to calculate the point cloud density, realizes adaptive adjustment of the maximum clustering point number parameter, effectively eliminates the interference of the tunnel wall, and significantly improves the extraction quality and accuracy of the pipeline point cloud.

Description

Technical Field

[0001] The invention discloses a method for automatically extracting point clouds of linear pipelines in mine shafts and tunnels based on regional growth, and belongs to the technical field of point cloud processing. Background Art

[0002] Accurate pipeline information can not only provide reliable data support for the subsequent intelligent management of underground ventilation systems and water supply and drainage systems in mines, but also improve the quality of overall digital model construction. In view of the above needs, it is particularly urgent to develop an efficient point cloud processing method to quickly extract pipeline data in tunnels. Common methods in the field of point cloud processing include region growing, feature clustering, model fitting, and deep learning-based technologies. Feature clustering algorithms are sensitive to parameter selection, and their computational efficiency in high-dimensional space is often unsatisfactory; model fitting methods require pre-defined geometric models, which limits their scope of application; although deep learning methods are powerful, they usually require a large amount of training data and computing resources to support continuous learning and optimization of the model. In contrast, the region growing method is widely used in point cloud processing because it is easy to implement and has high processing accuracy. Summary of the invention

[0003] The purpose of the present invention is to provide a method for automatically extracting point clouds of linear pipelines in mine shafts and tunnels based on region growing, so as to solve the problem of low accuracy of pipeline point cloud extraction in the prior art.

[0004] The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing includes:

[0005] S1: Use LiDAR to collect point cloud data of underground mines and preprocess the point cloud data, including data registration, downsampling and denoising;

[0006] S2: The pre-processed point cloud data is subjected to ground point filtering. Ground point filtering includes spatial inversion of the point cloud data and covering it from top to bottom with simulated cloth. The final cloth shape is the roadway ground point cloud.

[0007] S3: Project the extracted roadway ground point cloud onto the two-dimensional fitting plane, construct the point cloud convex hull and calculate the area of ​​the point cloud convex hull, and calculate the point cloud density by combining the number of roadway ground point clouds and the point cloud convex hull;

[0008] S4: Construct a triangulated network model of the tunnel point cloud, calculate the surface area of ​​the tunnel wall, and estimate the maximum clustering point number parameter by combining the point cloud density and the surface area of ​​the tunnel wall;

[0009] S5: Based on the region growing method, the normal vector smoothing threshold and curvature limit are applied to the point cloud data. An iterative processing mechanism is introduced into the region growing method to segment and extract the pipeline point cloud set.

[0010] The use of LiDAR to collect point cloud data of underground mines includes using data acquisition equipment to continuously scan and measure the internal space of the mine tunnel, and obtaining a collection of three-dimensional color scattered point data on the inner surface of the tunnel as point cloud data.

[0011] The point cloud data contains the three-dimensional coordinates, reflection intensity and color information of the midpoints in the tunnel.

[0012] In the preprocessing, data registration includes a combination of coarse registration and fine registration, which transfers point cloud data in different coordinate reference systems to the same coordinate reference system through two rigid body transformations of rotation and translation, so as to achieve complementarity between point cloud data and obtain geometric topological information;

[0013] Downsampling includes setting the sampling method of the spatial point spacing to form new roadway point cloud data with less data volume;

[0014] De-noising includes using filtering denoising method to remove discrete points and noise points of varying degrees in point cloud data.

[0015] S2 includes, S2.1: virtually suspending a piece of cloth above the inverted point cloud data in space, taking the cloth as the starting state of the simulation, projecting the point cloud data and the cloth mesh particles onto the horizontal plane at the same time, and recording the original height value of each point cloud data and cloth mesh particle as the interaction height value, and the interaction height value is recorded as IHV;

[0016] S2.2: Set the iteration parameters to record the number of iterations, and end the iteration when the maximum number of iterations is reached or all cloth mesh particles are immobile;

[0017] Calculate the displacement of each cloth mesh particle under the action of gravity, and compare the height value of the cloth mesh particle after displacement with the IHV value;

[0018] If the height value after displacement is less than or equal to IHV, the cloth mesh particle height is set to IHV, and the cloth mesh particle is marked as fixed, indicating that the cloth mesh particle has contacted the point cloud surface;

[0019] If the height value after displacement is greater than IHV, keep the cloth mesh particles movable and continue to fall in the next iteration;

[0020] The displacement of the cloth mesh particles under gravity is :

[0021] ;

[0022] In the formula, For time, is the time interval, is the gravitational constant, , for The displacement of cloth mesh particles under gravity over time, for The displacement of cloth mesh particles under gravity over time, for Displacement of cloth mesh particles due to gravity over time.

[0023] S2 includes, S2.3: calculating the displacement of cloth mesh particles under the action of internal forces;

[0024] ; ;

[0025] In the formula, is the displacement vector of the cloth mesh particle, is the judgment coefficient, yes The adjacent cloth mesh particles, is the normalized vector in the vertical direction, It is the first cloth mesh particle to perform displacement calculations;

[0026] The internal forces on the cloth mesh particles include: in the first case, both movable cloth mesh particles are acted upon by forces and move equal distances in opposite directions; in the second case, a movable cloth mesh particle and an immovable cloth mesh particle are acted upon by different forces, and only the movable cloth mesh particle is displaced; in the third case, neither of the two immovable cloth mesh particles is acted upon by forces, and there is no relative displacement;

[0027] For movable cloth mesh particles, , for immovable cloth mesh particles, ;

[0028] S2.4: The position of the cloth mesh particles is updated based on the displacement of the cloth mesh particles under the action of gravity and the displacement under the action of internal force, the distance between the preprocessed point cloud and its nearest cloth mesh particle is calculated, and compared with the distance threshold to determine whether the current point cloud belongs to the ground point cloud or the non-ground point cloud, and the tunnel ground point cloud is extracted.

[0029] S3 includes, S3.1: performing plane fitting on the roadway ground point cloud and obtaining coefficients of the plane equation;

[0030] S3.2: Project the preprocessed point cloud data onto the plane fitted in S3.1 according to the plane equation, construct a projection function, map the three-dimensional coordinates of each point of the point cloud data onto the plane, and obtain a projected point cloud;

[0031] For point , the projection on the plane is :

[0032] ;

[0033] In the formula, A, B, C, and D are the four coefficients of the plane equation;

[0034] S3.3: In the coordinate system where the projected point cloud is located, extract the XY coordinates of the projected point cloud, ignore the Z coordinate, form a two-dimensional point set, and use the convex hull to calculate the area of ​​the point cloud on the XY plane. The convex hull area is the area after the point cloud is projected onto the plane;

[0035] Assume that the vertices of the convex hull are arranged in order , the convex hull area The calculation formula is:

[0036] ;

[0037] In the formula, is the number of convex hulls, It is A convex hull, It is The abscissa of the convex hull, It is The ordinate of the convex hull;

[0038] Calculate the point cloud density den:

[0039] ;

[0040] In the formula, is the number of point clouds.

[0041] S4 includes: the tunnel point cloud triangulation model uses a mesh triangulation model, traverses all triangular facets, calculates the area of ​​each triangle and accumulates the surface area of ​​the tunnel wall :

[0042] ;

[0043] In the formula, is the total number of triangle patches, , , It is The coordinates of the three vertices of a triangle;

[0044] according to and Estimate the number of tunnel wall points and introduce a redundancy factor k to ensure coverage:

[0045] ;

[0046] In the formula, is the estimated number of point clouds, As the maximum number of cluster points parameter of the clustering algorithm.

[0047] S5 includes, S5.1: determining a minimum cluster point number parameter, and discarding results that are less than a threshold value of the minimum cluster point number parameter;

[0048] S5.2: Scan each point in the roadway point cloud data, use the KD tree to search for the K nearest neighboring points of each point, and calculate the local plane in the least squares sense of the neighboring points:

[0049] ;

[0050] In the formula, is a local plane, for The normal vector of for The distance to the origin of the coordinate system, represents the parameter value that makes the function value reach the minimum or minimum, For the Neighborhood points;

[0051] The centroid passing through K neighboring points , solve the covariance matrix :

[0052] ;

[0053] Covariance matrix Perform eigenvalue decomposition and obtain The eigenvalues ​​of The eigenvector corresponding to the minimum eigenvalue of The normal vector .

[0054] S5 includes, S5.3: If The eigenvalues ​​of ,but The surface curvature of a point is:

[0055] ;

[0056] In the formula, yes The characteristic value of is the surface curvature;

[0057] S5.4: Set an empty seed point sequence and an empty clustering array, sort the input point cloud according to the surface curvature of the point cloud, select the point with the smallest surface curvature as the initial seed point, and add the initial seed point to the seed point sequence;

[0058] S5.5: Search for neighboring points of the current seed point, calculate the angle between the normal line of the neighboring points of the seed point and the normal line of the current seed point, and add the neighboring points whose angle is less than the angle threshold to the current region;

[0059] S5.6: Check the surface curvature of each neighboring point, add the neighboring points with a curvature less than the curvature threshold to the seed point sequence, delete the current seed point, and select one point from the seed point sequence as a seed to continue growing;

[0060] S5.7: Repeat S5.4 to S5.6 until the seed point sequence is cleared, then a region is grown and the region is added to the clustering array;

[0061] S5.8: Repeat S5.7 until all point clouds are traversed, the tunnel wall data is removed for the first time, and the initial pipeline point cloud extraction result is obtained;

[0062] S5.9: Increase the normal vector angle threshold, curvature threshold and minimum cluster point number parameters, repeat S5 for iteration, further eliminate the tunnel wall data, and obtain the pipeline point cloud extraction result after iteration until the accuracy requirements are met and stop iteration.

[0063] Compared with the prior art, the present invention has the following beneficial effects: the present invention combines the ground filtering algorithm to calculate the point cloud density, realizes the adaptive adjustment of the maximum clustering point number parameter, effectively eliminates the interference of the tunnel wall, and significantly improves the extraction quality and accuracy of the pipeline point cloud. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a technical flow chart of the method of the present invention;

[0065] Figure 2 The original three-dimensional laser point cloud data of the underground mine tunnel containing the linear pipeline of the present invention;

[0066] Figure 3 It is the classification diagram of the tunnel point cloud and the ground point cloud of the present invention;

[0067] Figure 4 A graph showing the ground point cloud extracted by the present invention projected onto the RANSAC fitting plane;

[0068] Figure 5 This is the linear pipeline point cloud segmented and extracted by the present invention. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing includes:

[0071] S1: Use LiDAR to collect point cloud data of underground mines and preprocess the point cloud data, including data registration, downsampling and denoising;

[0072] S2: The pre-processed point cloud data is subjected to ground point filtering. Ground point filtering includes spatial inversion of the point cloud data and covering it from top to bottom with simulated cloth. The final cloth shape is the roadway ground point cloud.

[0073] S3: Project the extracted roadway ground point cloud onto the two-dimensional fitting plane, construct the point cloud convex hull and calculate the area of ​​the point cloud convex hull, and calculate the point cloud density by combining the number of roadway ground point clouds and the point cloud convex hull;

[0074] S4: Construct a triangulated network model of the tunnel point cloud, calculate the surface area of ​​the tunnel wall, and estimate the maximum clustering point number parameter by combining the point cloud density and the surface area of ​​the tunnel wall;

[0075] S5: Based on the region growing method, the normal vector smoothing threshold and curvature limit are applied to the point cloud data. An iterative processing mechanism is introduced into the region growing method to segment and extract the pipeline point cloud set.

[0076] The use of LiDAR to collect point cloud data of underground mines includes using data acquisition equipment to continuously scan and measure the internal space of the mine tunnel, and obtaining a collection of three-dimensional color scattered point data on the inner surface of the tunnel as point cloud data.

[0077] The point cloud data contains the three-dimensional coordinates, reflection intensity and color information of the midpoints in the tunnel.

[0078] In the preprocessing, data registration includes a combination of coarse registration and fine registration, which transfers point cloud data in different coordinate reference systems to the same coordinate reference system through two rigid body transformations of rotation and translation, so as to achieve complementarity between point cloud data and obtain geometric topological information;

[0079] Downsampling includes setting the sampling method of the spatial point spacing to form new roadway point cloud data with less data volume;

[0080] De-noising includes using filtering denoising method to remove discrete points and noise points of varying degrees in point cloud data.

[0081] S2 includes, S2.1: virtually suspending a piece of cloth above the inverted point cloud data in space, taking the cloth as the starting state of the simulation, projecting the point cloud data and the cloth mesh particles onto the horizontal plane at the same time, and recording the original height value of each point cloud data and cloth mesh particle as the interaction height value, and the interaction height value is recorded as IHV;

[0082] S2.2: Set the iteration parameters to record the number of iterations, and end the iteration when the maximum number of iterations is reached or all cloth mesh particles are immobile;

[0083] Calculate the displacement of each cloth mesh particle under the action of gravity, and compare the height value of the cloth mesh particle after displacement with the IHV value;

[0084] If the height value after displacement is less than or equal to IHV, the cloth mesh particle height is set to IHV, and the cloth mesh particle is marked as fixed, indicating that the cloth mesh particle has contacted the point cloud surface;

[0085] If the height value after displacement is greater than IHV, keep the cloth mesh particles movable and continue to fall in the next iteration;

[0086] The displacement of the cloth mesh particles under gravity is :

[0087] ;

[0088] In the formula, For time, is the time interval, is the gravitational constant, , for The displacement of cloth mesh particles under gravity over time, for The displacement of cloth mesh particles under gravity over time, for Displacement of cloth mesh particles due to gravity over time.

[0089] S2 includes, S2.3: calculating the displacement of cloth mesh particles under the action of internal forces;

[0090] ; ;

[0091] In the formula, is the displacement vector of the cloth mesh particle, is the judgment coefficient, yes The adjacent cloth mesh particles, is the normalized vector in the vertical direction, It is the first cloth mesh particle to perform displacement calculations;

[0092] The internal forces on the cloth mesh particles include: in the first case, both movable cloth mesh particles are acted upon by forces and move equal distances in opposite directions; in the second case, a movable cloth mesh particle and an immovable cloth mesh particle are acted upon by different forces, and only the movable cloth mesh particle is displaced; in the third case, neither of the two immovable cloth mesh particles is acted upon by forces, and there is no relative displacement;

[0093] For movable cloth mesh particles, , for immovable cloth mesh particles, ;

[0094] S2.4: The position of the cloth mesh particles is updated based on the displacement of the cloth mesh particles under the action of gravity and the displacement under the action of internal force, the distance between the preprocessed point cloud and its nearest cloth mesh particle is calculated, and compared with the distance threshold to determine whether the current point cloud belongs to the ground point cloud or the non-ground point cloud, and the tunnel ground point cloud is extracted.

[0095] S3 includes, S3.1: performing plane fitting on the roadway ground point cloud and obtaining coefficients of the plane equation;

[0096] S3.2: Project the preprocessed point cloud data onto the plane fitted in S3.1 according to the plane equation, construct a projection function, map the three-dimensional coordinates of each point of the point cloud data onto the plane, and obtain a projected point cloud;

[0097] For point , the projection on the plane is :

[0098] ;

[0099] In the formula, A, B, C, and D are the four coefficients of the plane equation;

[0100] S3.3: In the coordinate system where the projected point cloud is located, extract the XY coordinates of the projected point cloud, ignore the Z coordinate, form a two-dimensional point set, and use the convex hull to calculate the area of ​​the point cloud on the XY plane. The convex hull area is the area after the point cloud is projected onto the plane;

[0101] Assume that the vertices of the convex hull are arranged in order , the convex hull area The calculation formula is:

[0102] ;

[0103] In the formula, is the number of convex hulls, It is A convex hull, It is The abscissa of the convex hull, It is The ordinate of the convex hull;

[0104] Calculate the point cloud density den:

[0105] ;

[0106] In the formula, is the number of point clouds.

[0107] S4 includes: the tunnel point cloud triangulation model uses a mesh triangulation model, traverses all triangular facets, calculates the area of ​​each triangle and accumulates the surface area of ​​the tunnel wall :

[0108] ;

[0109] In the formula, is the total number of triangle patches, , , It is The coordinates of the three vertices of a triangle;

[0110] according to and Estimate the number of tunnel wall points and introduce a redundancy factor k to ensure coverage:

[0111] ;

[0112] In the formula, is the estimated number of point clouds, As the maximum number of cluster points parameter of the clustering algorithm.

[0113] S5 includes, S5.1: determining a minimum cluster point number parameter, and discarding results that are less than a threshold value of the minimum cluster point number parameter;

[0114] S5.2: Scan each point in the roadway point cloud data, use the KD tree to search for the K nearest neighboring points of each point, and calculate the local plane in the least squares sense of the neighboring points:

[0115] ;

[0116] In the formula, is a local plane, for The normal vector of for The distance to the origin of the coordinate system, represents the parameter value that makes the function value reach the minimum or minimum, For the Neighborhood points;

[0117] The centroid passing through K neighboring points , solve the covariance matrix :

[0118] ;

[0119] Covariance matrix Perform eigenvalue decomposition and obtain The eigenvalues ​​of The eigenvector corresponding to the minimum eigenvalue of The normal vector .

[0120] S5 includes, S5.3: If The eigenvalues ​​of ,but The surface curvature of a point is:

[0121] ;

[0122] In the formula, yes The characteristic value of is the surface curvature;

[0123] S5.4: Set an empty seed point sequence and an empty clustering array, sort the input point cloud according to the surface curvature of the point cloud, select the point with the smallest surface curvature as the initial seed point, and add the initial seed point to the seed point sequence;

[0124] S5.5: Search for neighboring points of the current seed point, calculate the angle between the normal line of the neighboring points of the seed point and the normal line of the current seed point, and add the neighboring points whose angle is less than the angle threshold to the current region;

[0125] S5.6: Check the surface curvature of each neighboring point, add the neighboring points with a curvature less than the curvature threshold to the seed point sequence, delete the current seed point, and select one point from the seed point sequence as a seed to continue growing;

[0126] S5.7: Repeat S5.4 to S5.6 until the seed point sequence is cleared, then a region is grown and the region is added to the clustering array;

[0127] S5.8: Repeat S5.7 until all point clouds are traversed, the tunnel wall data is removed for the first time, and the initial pipeline point cloud extraction result is obtained;

[0128] S5.9: Increase the normal vector angle threshold, curvature threshold and minimum cluster point number parameters, repeat S5 for iteration, further eliminate the tunnel wall data, and obtain the pipeline point cloud extraction result after iteration until the accuracy requirements are met and stop iteration.

[0129] The technical process of the present invention is as follows Figure 1 As shown, it includes data preprocessing, ground point projection calculation of point cloud density, and pipeline point cloud segmentation and extraction, wherein data preprocessing includes underground mine point cloud data, data registration, data denoising, and downsampling to obtain the underground mine point cloud data to be processed, and then extract the ground point cloud based on CSF, fit the ground point cloud to the plane by RANSAC, combine the ground point cloud with the fitting plane to obtain two-dimensional ground point cloud data, calculate the point cloud density, and then combine the underground mine point cloud data to be processed to judge the maximum number of cluster points, the minimum number of cluster points, the curvature threshold, and the angle threshold. If there are ungrown points, the seed point selection is performed and the current seed point is updated. If the seed point is not empty, the regional growth algorithm is performed to obtain a new seed point, and the initial segmentation result of the pipeline point cloud is obtained. If there is no under-segmentation, the pipeline point cloud is output. If there is no ungrown point, the initial segmentation result of the pipeline point cloud is directly obtained. If the seed point is empty, return to judge whether there are ungrown points. If there is under-segmentation, return the underground mine point cloud data to be processed.

[0130] The original three-dimensional laser point cloud data of the underground mine tunnel containing linear pipelines in the present invention is as follows: Figure 2 As shown in the figure, the data has been preprocessed. Internal noise data and external outlier data have been removed, and the data has been downsampled to improve the efficiency of subsequent algorithm operation.

[0131] The classification diagram of the tunnel point cloud and the ground point cloud of the present invention is as follows Figure 3 As shown in FIG. 1 , the red part is the extracted ground point cloud, and the green part is the ground data. The ground point cloud extracted by the present invention is projected onto the RANSAC fitting plane. Figure 4 As shown in FIG. 1 , the original point cloud data is fitted by RANSAC plane to obtain a best fitting plane, the red points are the original point cloud, and the blue points are the projection points. The linear pipeline point cloud extracted by segmentation of the present invention is shown in FIG. Figure 5 As shown, the method of the present invention is used to extract continuous linear pipeline data from underground mine shaft point cloud data.

[0132] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features may be replaced by equivalents, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing is characterized by: include: S1: Use LiDAR to collect point cloud data of underground mines and preprocess the point cloud data, including data registration, downsampling and denoising; S2: The pre-processed point cloud data is subjected to ground point filtering. Ground point filtering includes spatial inversion of the point cloud data and covering it from top to bottom with simulated cloth. The final cloth shape is the roadway ground point cloud. S3: Project the extracted roadway ground point cloud onto the two-dimensional fitting plane, construct the point cloud convex hull and calculate the area of ​​the point cloud convex hull, and calculate the point cloud density by combining the number of roadway ground point clouds and the point cloud convex hull; S4: Construct a triangulated network model of the tunnel point cloud, calculate the surface area of ​​the tunnel wall, and estimate the maximum clustering point number parameter by combining the point cloud density and the surface area of ​​the tunnel wall; S4 includes: the tunnel point cloud triangulation model uses a mesh triangulation model, traverses all triangular facets, calculates the area of ​​each triangle and accumulates the surface area of ​​the tunnel wall : ； In the formula, is the total number of triangle patches, , , It is The coordinates of the three vertices of a triangle; according to and Estimate the number of tunnel wall points and introduce a redundancy factor k to ensure coverage: ； In the formula, is the estimated number of point clouds, As the maximum cluster point number parameter of the clustering algorithm; S5: Based on the region growing method, the normal vector smoothing threshold and curvature limit are applied to the point cloud data, and an iterative processing mechanism is introduced into the region growing method to segment and extract the pipeline point cloud set; S5 includes, S5.1: determining a minimum cluster point number parameter, and discarding results that are less than a threshold value of the minimum cluster point number parameter; S5.2: Scan each point in the roadway point cloud data, use the KD tree to search for the K nearest neighboring points of each point, and calculate the local plane in the least squares sense of the neighboring points: ； In the formula, is a local plane, for The normal vector of for The distance to the origin of the coordinate system, represents the parameter value that makes the function value reach the minimum or minimum, For the Neighborhood points; The centroid passing through K neighboring points , solve the covariance matrix : ； Covariance matrix Perform eigenvalue decomposition and obtain The eigenvalues ​​of The eigenvector corresponding to the minimum eigenvalue of The normal vector ; S5 includes, S5.3: If The eigenvalues ​​of ,but The surface curvature of a point is: ； In the formula, yes The characteristic value of is the surface curvature; S5.4: Set an empty seed point sequence and an empty clustering array, sort the input point cloud according to the surface curvature of the point cloud, select the point with the smallest surface curvature as the initial seed point, and add the initial seed point to the seed point sequence; S5.5: Search for neighboring points of the current seed point, calculate the angle between the normal line of the neighboring points of the seed point and the normal line of the current seed point, and add the neighboring points whose angle is less than the angle threshold to the current region; S5.6: Check the surface curvature of each neighboring point, add the neighboring points with a curvature less than the curvature threshold to the seed point sequence, delete the current seed point, and select one point from the seed point sequence as a seed to continue growing; S5.7: Repeat S5.4 to S5.6 until the seed point sequence is cleared, then a region is grown and the region is added to the clustering array; S5.8: Repeat S5.7 until all point clouds are traversed, the tunnel wall data is removed for the first time, and the initial pipeline point cloud extraction result is obtained; S5.9: Increase the normal vector angle threshold, curvature threshold and minimum cluster point number parameters, repeat S5 for iteration, further eliminate the tunnel wall data, and obtain the pipeline point cloud extraction result after iteration until the accuracy requirements are met and stop iteration.

2. The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing according to claim 1 is characterized in that: The use of LiDAR to collect point cloud data of underground mines includes using data acquisition equipment to continuously scan and measure the internal space of the mine tunnel, and obtaining a collection of three-dimensional color scattered point data on the inner surface of the tunnel as point cloud data.

3. The method for automatically extracting point clouds of linear pipelines in mine shafts and tunnels based on region growing according to claim 2 is characterized in that: The point cloud data contains the three-dimensional coordinates, reflection intensity and color information of the midpoints in the tunnel.

4. The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing according to claim 3 is characterized in that: In the preprocessing, data registration includes a combination of coarse registration and fine registration, which transfers point cloud data in different coordinate reference systems to the same coordinate reference system through two rigid body transformations of rotation and translation, so as to achieve complementarity between point cloud data and obtain geometric topological information; Downsampling includes setting the sampling method of the spatial point spacing to form new roadway point cloud data with less data volume; De-noising includes using filtering denoising method to remove discrete points and noise points of varying degrees in point cloud data.

5. The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing according to claim 4 is characterized in that: S2 includes, S2.1: virtually suspending a piece of cloth above the inverted point cloud data in space, taking the cloth as the starting state of the simulation, projecting the point cloud data and the cloth mesh particles onto the horizontal plane at the same time, and recording the original height value of each point cloud data and cloth mesh particle as the interaction height value, and the interaction height value is recorded as IHV; S2.2: Set the iteration parameters to record the number of iterations, and end the iteration when the maximum number of iterations is reached or all cloth mesh particles are immobile; Calculate the displacement of each cloth mesh particle under the action of gravity, and compare the height value of the cloth mesh particle after displacement with the IHV value; If the height value after displacement is less than or equal to IHV, the cloth mesh particle height is set to IHV, and the cloth mesh particle is marked as fixed, indicating that the cloth mesh particle has contacted the point cloud surface; If the height value after displacement is greater than IHV, keep the cloth mesh particles movable and continue to fall in the next iteration; The displacement of the cloth mesh particles under gravity is : ； In the formula, For time, is the time interval, is the gravitational constant, , for The displacement of cloth mesh particles under gravity over time, for The displacement of cloth mesh particles under gravity over time, for Displacement of cloth mesh particles due to gravity over time.

6. The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing according to claim 5 is characterized in that: S2 includes, S2.3: calculating the displacement of cloth mesh particles under the action of internal forces; ； ； In the formula, is the displacement vector of the cloth mesh particle, is the judgment coefficient, yes The adjacent cloth mesh particles, is the normalized vector in the vertical direction, It is the first cloth mesh particle to perform displacement calculations; The internal forces on the cloth mesh particles include: in the first case, two movable cloth mesh particles are both acted upon by the force and move equal distances in opposite directions; In the second case, a movable cloth mesh particle and an immovable cloth mesh particle are subjected to different forces, and only the movable cloth mesh particle is displaced; in the third case, there is no force between the two immovable cloth mesh particles, and there is no relative displacement; For movable cloth mesh particles, , for immovable cloth mesh particles, ; S2.4: The position of the cloth mesh particles is updated based on the displacement of the cloth mesh particles under the action of gravity and the displacement under the action of internal force, the distance between the preprocessed point cloud and its nearest cloth mesh particle is calculated, and compared with the distance threshold to determine whether the current point cloud belongs to the ground point cloud or the non-ground point cloud, and the tunnel ground point cloud is extracted.

7. The method for automatically extracting point clouds of linear pipelines in mine tunnels based on region growing according to claim 6 is characterized in that: S3 includes, S3.1: performing plane fitting on the roadway ground point cloud and obtaining coefficients of the plane equation; S3.2: Project the preprocessed point cloud data onto the plane fitted in S3.1 according to the plane equation, construct a projection function, map the three-dimensional coordinates of each point of the point cloud data onto the plane, and obtain a projected point cloud; For point , the projection on the plane is : ； In the formula, A, B, C, and D are the four coefficients of the plane equation; S3.3: In the coordinate system where the projected point cloud is located, extract the XY coordinates of the projected point cloud, ignore the Z coordinate, form a two-dimensional point set, and use the convex hull to calculate the area of ​​the point cloud on the XY plane. The convex hull area is the area after the point cloud is projected onto the plane; Assume that the vertices of the convex hull are arranged in order , the convex hull area The calculation formula is: ； In the formula, is the number of convex hulls, It is A convex hull, It is The abscissa of the convex hull, It is The ordinate of the convex hull; Calculate the point cloud density den: ； In the formula, is the number of point clouds.

Images