CN-121767574-B - Method and device for removing surface point cloud of explosion stack

CN121767574BCN 121767574 BCN121767574 BCN 121767574BCN-121767574-B

Abstract

The invention provides a method and a device for removing a surface point cloud of a blasting stack, and relates to the technical field of underground metal mine blasting exploitation engineering. The method comprises the steps of obtaining original three-dimensional point cloud data of a detonation heap, downsampling to obtain three-dimensional point cloud of the detonation heap, determining a segmentation threshold according to characteristic distribution of the three-dimensional point cloud of the detonation heap, segmenting by adopting an area growth algorithm to obtain a plurality of segmentation result point clouds, calculating geometrical characteristic vectors of each segmentation result point cloud, screening, extracting a ground characteristic subset through unsupervised cluster analysis based on a screened segmentation result point cloud characteristic data set, constructing a ground point cloud joint judgment rule, judging each segmentation result point cloud as a ground point Yun Huofei ground point cloud, removing all the identified ground point clouds, and denoising all non-ground point clouds to obtain the detonation heap main body point cloud. The invention aims to improve the quality and usability of the three-dimensional point cloud data of the detonation heap, and ensure the efficient and accurate identification and statistics of the detonation heap blocking degree.

Inventors

Mao Jufeng
AN LONG
JIN RONGCHENG
ZHANG JIAHUA
XU SHIDA
WANG JUN

Assignees

东北大学

Dates

Publication Date: 20260508
Application Date: 20260305

Claims (9)

1. The method for removing the cloud of the surface points of the explosion stack is characterized by comprising the following steps: Acquiring original three-dimensional point cloud data of the explosion stack, and performing downsampling on the original three-dimensional point cloud data of the explosion stack based on a voxel downsampling algorithm to acquire the three-dimensional point cloud of the explosion stack; Determining a segmentation threshold according to the characteristic distribution of the three-dimensional point cloud of the explosion stack, and segmenting the three-dimensional point cloud of the explosion stack by utilizing a region growing algorithm to obtain a plurality of segmentation result point clouds; respectively calculating geometric feature vectors of each segmentation result point cloud, screening all segmentation result point clouds according to preset screening conditions based on all calculated geometric feature vectors, and generating a screened segmentation result point cloud feature data set; the geometrical characteristic vector comprises distribution concentration, directional height difference and average roughness; Extracting a ground feature subset through unsupervised cluster analysis based on the screened segmentation result point cloud feature dataset; the ground characteristic subset comprises a ground distribution concentration degree subset Directed height difference subset for ground class And a ground-based average roughness subset ; Based on the ground feature subset, constructing a ground point cloud joint judgment rule, and judging each segmentation result point cloud as a ground point Yun Huofei ground point cloud by utilizing the ground point cloud joint judgment rule, wherein the concrete contents are as follows: Computing ground class distribution concentration subsets Mean of (2) And standard deviation And utilize the mean value And standard deviation Calculating a distribution concentration threshold ; Computing a directed height difference subset of a ground class Mean of (2) And standard deviation And utilize the mean value And standard deviation Calculating a directional altitude difference threshold ; Computing a ground-based average roughness subset Mean of (2) And standard deviation And utilize the mean value And standard deviation Calculating the average roughness threshold ; According to the distribution concentration threshold Threshold value of directional altitude difference And average roughness threshold The method comprises the steps of constructing a joint judgment rule of the ground point cloud, namely identifying the point cloud which simultaneously meets the segmentation result that the distribution concentration is larger than a distribution concentration threshold, the directional height difference is smaller than a directional height difference threshold and the average roughness is smaller than an average roughness threshold as the ground point cloud, and otherwise, identifying the point cloud as the non-ground point cloud; identifying each division result point cloud as a surface point Yun Huofei ground point cloud by utilizing a ground point cloud joint judgment rule; and removing all the identified ground point clouds, and removing ground scattered points in all the non-ground point clouds by adopting a denoising algorithm to obtain the main point cloud of the detonation reactor.
2. The method for removing the point cloud of the surface of the explosion stack according to claim 1, wherein the method is characterized in that the segmentation threshold value is determined according to the characteristic distribution of the three-dimensional point cloud of the explosion stack, and the three-dimensional point cloud of the explosion stack is segmented by utilizing a region growing algorithm, so that the specific contents of a plurality of segmentation result point clouds are as follows: obtaining a normal vector of a three-dimensional point cloud of the detonation reactor by a local surface fitting method; adopting a minimum spanning tree method to carry out consistency adjustment on the normal vector direction of the three-dimensional point cloud of the detonation reactor; Based on the normal vector of the adjusted three-dimensional point cloud of the detonation pile, adopting a frequency counting method to adaptively determine a segmentation threshold value of the three-dimensional point cloud of the detonation pile; and according to the segmentation threshold, segmenting the three-dimensional point cloud of the detonation reactor by adopting a region growing algorithm to obtain a plurality of segmentation result point clouds.
3. The method for removing the point cloud of the surface of the explosion stack according to claim 2, wherein the specific content of the segmentation threshold value of the three-dimensional point cloud of the explosion stack is determined in a self-adaptive manner by adopting a frequency statistics method based on the normal vector of the three-dimensional point cloud of the explosion stack after adjustment, and the specific content is as follows: for any point in the three-dimensional point cloud of the detonation reactor Sum point Is a neighborhood of (a) Calculating points according to the normal vector of the three-dimensional point cloud of the adjusted detonation reactor And neighborhood of Normal vector angle mean of points in the interior as points Is included in the normal vector; at the same time, calculate the point Is a curvature of (2); Dividing a plurality of equidistant normal vector angle intervals according to normal vector angles of all points in the three-dimensional point cloud of the explosion stack, respectively counting frequency and frequency in each normal vector angle interval, further constructing a normal vector angle frequency distribution table, and drawing a normal vector angle frequency distribution curve; Dividing a plurality of equidistant curvature intervals according to the curvatures of all points in the three-dimensional point cloud of the detonation reactor, respectively counting the frequency and the frequency in each curvature interval, further constructing a curvature frequency distribution table, and drawing a curvature frequency distribution curve; the normal vector included angle value corresponding to the peak point of the normal vector angular frequency distribution curve is used as a normal vector included angle threshold value, and the curvature value corresponding to the elbow point of the curvature frequency distribution curve is used as a curvature threshold value, wherein the elbow point is a point on the curvature frequency distribution curve, and the perpendicular distance of the point reaches the maximum value relative to the connecting line of the curve starting point and the curve ending point; and the normal vector included angle threshold and the curvature threshold are jointly used as the segmentation threshold of the three-dimensional point cloud of the detonation reactor.
4. The method for removing the point cloud of the surface of the explosion stack according to claim 1, wherein the specific contents of the feature data set of each of the split result point clouds are as follows: For arbitrary segmentation result point clouds The result point cloud will be split The ratio of the number of points of the three-dimensional point cloud of the detonation reactor is used as the distribution concentration of the point cloud of the segmentation result Wherein Index values for the point clouds of the segmentation result; calculating a segmentation result point cloud Is directed to the height difference ; Calculating a segmentation result point cloud Average roughness of (2) ; Splitting result point clouds by combining Distribution concentration of (2) Directional height difference And average roughness Constructing a segmentation result point cloud Is of the geometric feature vector of (1) ; Constructing a feature data set of the segmentation result point cloud according to the geometric feature vectors of all the segmentation result point clouds ; Respectively calculating the point cloud characteristic data sets of the segmentation result Mean and standard deviation of middle distribution concentration, directional height difference and average roughness; based on the calculated mean and standard deviation, if the result point cloud is arbitrarily divided If the preset screening condition is met, reserving the point cloud characteristic data set of the segmentation result Mid-segmentation result point cloud Is of the geometric feature vector of (1) Otherwise, from the characteristic data set of the segmentation result point cloud Removing the segmentation result point cloud Is of the geometric feature vector of (1) Further obtaining the filtered segmentation result point cloud characteristic data set 。
5. The method for removing the cloud of the surface point of the explosion stack according to claim 4, wherein the preset screening conditions are as follows: ; Wherein the method comprises the steps of Representing one geometrical feature of distribution concentration, directed height difference and average roughness, wherein The degree of distribution concentration is indicated and, The directional height difference is indicated as such, Represents average roughness; representing a segmented result point cloud Is the first of (2) Item geometry feature values; representing segmented result point cloud feature datasets Middle (f) A mean value of the term geometry; representing segmented result point cloud feature datasets Middle (f) Standard deviation of term geometry; Is constant.
6. The method for removing a cloud of surface points of a detonation reactor of claim 4, wherein said directional elevation difference The calculation method of (1) is as follows: calculating a segmentation result point cloud And obtaining three eigenvalues by eigenvalue decomposition And (2) and Corresponding three feature vectors ; At the maximum characteristic value Corresponding feature vector Point cloud as segmentation result In the X-axis direction of (2) with the next largest eigenvalue Corresponding feature vector As the Y-axis direction, the rest of the feature vectors As Z-axis direction, constructing a point cloud of a segmentation result Is a directed bounding box; Feature vector in Z-axis direction For direction vectors, the resulting point cloud is split Any point in (2) Calculation points Position vector of (a) And (3) with And takes the dot product as a dot Position vector of (a) Projection distance along Z-axis direction ; Counting the maximum value of projection distances corresponding to all points And minimum value And taking the difference value between the maximum value and the minimum value as a point cloud of the segmentation result Directional height difference along Z-axis direction 。
7. The method of removing a cloud of surface points of a detonation of claim 4, wherein the average roughness The calculation method of (1) is as follows: For the point cloud of the segmentation result Any point in (2) Performing k neighborhood search to obtain points Is a neighborhood point set of (1) ; From a set of neighborhood points Performing plane fitting and calculating points To a fitting plane Is a vertical distance of (2) As a dot Is a roughness of (2); calculating a segmentation result point cloud The average roughness of all points in the graph is taken as the average roughness of the point cloud of the segmentation result 。
8. The method for removing the surface point cloud of the explosion stack according to claim 1, wherein the specific contents of the ground feature subset are extracted by unsupervised cluster analysis based on the screened segmentation result point cloud feature dataset: Extracting and screening a segmentation result point cloud characteristic dataset Distribution concentration characteristic array in (a) Directional height difference characteristic array Average roughness profile array ; Setting the clustering quantity, and respectively setting the distribution concentration characteristic arrays Directional height difference characteristic array Average roughness profile array Performing cluster analysis to obtain a distribution concentration characteristic array Is a result of clustering of (a) Directional height difference characteristic array Is a result of clustering of (a) Average roughness profile array Is a result of clustering of (a) ; Identifying clustered results In a ground class cluster And uses the ground cluster All distribution concentrators in the ground class distribution concentrators are generated ; Identifying clustered results In a ground class cluster And uses the ground cluster All directional height differences in the ground class, and generating a ground class directional height difference subset ; Identifying clustered results In a ground class cluster And uses the ground cluster All roughness averages in (a) to generate a subset of roughness averages of the ground class ; The ground class distribution concentration subset Directed height difference subset for ground class And a ground-based average roughness subset Constituting a ground-class feature subset.
9. A blasting-pile ground point cloud removing device for implementing the blasting-pile ground point cloud removing method according to any one of claims 1 to 8, characterized in that the device comprises: the point cloud downsampling module is used for acquiring original three-dimensional point cloud data of the explosion stack, and downsampling the original three-dimensional point cloud data of the explosion stack based on a voxel downsampling algorithm to acquire the three-dimensional point cloud of the explosion stack; the point cloud segmentation module is used for determining a segmentation threshold according to the characteristic distribution of the three-dimensional point cloud of the explosion stack, and segmenting the three-dimensional point cloud of the explosion stack by utilizing a region growing algorithm to obtain a plurality of segmentation result point clouds; the feature screening module is used for respectively calculating the geometric feature vector of each segmentation result point cloud, screening all segmentation result point clouds according to preset screening conditions based on all calculated geometric feature vectors, and generating a screened segmentation result point cloud feature data set; The ground characteristic clustering module is used for extracting a ground characteristic subset through unsupervised cluster analysis based on the screened segmentation result point cloud characteristic dataset; The ground point cloud identification module is used for constructing a ground point cloud joint judgment rule based on the ground feature subset, and judging each segmentation result point cloud as a ground point Yun Huofei ground point cloud by utilizing the ground point cloud joint judgment rule; The ground point cloud removing module is used for removing all the identified ground point clouds, removing ground scattered points in all the non-ground point clouds by adopting a denoising algorithm, and obtaining the main point cloud of the explosion stack.

Description

Method and device for removing surface point cloud of explosion stack Technical Field The invention relates to the technical field of underground metal mine blasting exploitation engineering, in particular to a method and a device for removing surface point clouds of a blasting stack. Background In the statistics of the explosion pile block of an underground metal mine, the three-dimensional scanning technology is generally adopted to efficiently acquire the explosion pile point cloud data, accurately reflect the space form of the explosion pile, and accurately segment the explosion pile point cloud data by combining a point cloud segmentation algorithm, so that the identification and statistics of the explosion pile block are realized. However, the data of the point cloud of the explosion stack often contains a large amount of ground point cloud, the ground point cloud is uniformly distributed and has higher density, and is easy to be confused with the point cloud of the rock mass of the explosion stack, the accurate segmentation of the point cloud of the rock mass is interfered, and the accuracy of block identification is reduced. In addition, the existence of the ground point cloud also can obviously increase the data volume of the point cloud data and the calculation complexity of processing, and influence the efficiency and the accuracy of subsequent segmentation and block statistics. The conventional method for extracting the ground point cloud from the three-dimensional point cloud mainly comprises a cloth simulation filtering algorithm, progressive morphological filtering, RANSAC plane fitting and other methods. However, the ground of the detonation heap area is generally concave-convex, so that the robustness of the plane fitting method is poor, the ground point cloud is not completely removed, the cloth simulation filtering algorithm is high in computational complexity and limited in applicability, the effect is not ideal when complex and irregular detonation heap point cloud data are processed, the retention capacity of the point cloud boundary is poor due to progressive morphological filtering, and the precision of point cloud segmentation is affected. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a method and a device for removing the surface point cloud of a detonation heap, which aim to improve the quality and the usability of the three-dimensional point cloud data of the detonation heap and ensure the efficient and accurate identification and statistics of the detonation heap block size by analyzing the normal vector included angle and the curvature statistics characteristics of the three-dimensional point cloud data of the detonation heap to adaptively determine the region growth segmentation threshold value and constructing the joint judgment rule of the surface point cloud based on the distribution concentration degree, the directional height difference and the calculation and statistics of the average roughness characteristics of the point cloud of the segmentation result. In one aspect, the invention provides a method for removing a surface point cloud of a detonation reactor, which comprises the following steps: Acquiring original three-dimensional point cloud data of the explosion stack, and performing downsampling on the original three-dimensional point cloud data of the explosion stack based on a voxel downsampling algorithm to acquire the three-dimensional point cloud of the explosion stack; Determining a segmentation threshold according to the characteristic distribution of the three-dimensional point cloud of the explosion stack, and segmenting the three-dimensional point cloud of the explosion stack by utilizing a region growing algorithm to obtain a plurality of segmentation result point clouds; respectively calculating geometric feature vectors of each segmentation result point cloud, screening all segmentation result point clouds according to preset screening conditions based on all calculated geometric feature vectors, and generating a screened segmentation result point cloud feature data set; the geometrical characteristic vector comprises distribution concentration, directional height difference and average roughness; Extracting a ground feature subset through unsupervised cluster analysis based on the screened segmentation result point cloud feature dataset; the ground characteristic subset comprises a ground distribution concentration degree subset Directed height difference subset for ground classAnd a ground-based average roughness subset; Based on the ground feature subset, constructing a ground point cloud joint judgment rule, and judging each segmentation result point cloud as a ground point Yun Huofei ground point cloud by utilizing the ground point cloud joint judgment rule, wherein the concrete contents are as follows: Computing ground class distribution concentration subsets Mean of (2)And standard deviationAnd utilize th