A method and device for removing a blast pile ground point cloud

Abstract

The application provides a kind of muck pile ground point cloud removal method and device, it is related to underground metal mine blasting engineering technical field.The method comprises: obtaining the original three-dimensional point cloud data of muck pile and downsampling, obtaining muck pile three-dimensional point cloud;According to the feature distribution of muck pile three-dimensional point cloud, determine the segmentation threshold, and use region growing algorithm to segment, obtain several segmentation result point clouds;Calculate the geometric feature vector of each segmentation result point cloud and screen;Based on the feature data set of screened segmentation result point cloud, through unsupervised clustering analysis, extract ground class feature subset, and construct ground point cloud joint determination rule, determine each segmentation result point cloud as ground point cloud or non-ground point cloud;Remove all identified ground point clouds, and denoising processing is carried out to all non-ground point clouds, to obtain muck pile main point cloud.The application aims to improve the quality and availability of muck pile three-dimensional point cloud data, to ensure efficient and accurate identification and statistics of muck pile block degree.

Description

Technical Field

[0001] This invention relates to the field of underground metal mine blasting engineering technology, and in particular to a method and apparatus for removing point clouds from the ground after blasting. Background Technology

[0002] In the statistical analysis of blast pile block sizes in underground metal mines, 3D scanning technology is typically used to efficiently acquire blast pile point cloud data, accurately reflecting the spatial morphology of the blast pile. This data is then combined with point cloud segmentation algorithms to accurately segment the blast pile point cloud data, thereby enabling the identification and statistical analysis of blast pile block sizes. However, blast pile point cloud data often contains a large amount of surface point cloud data. Surface point clouds are uniformly distributed and have a high density, which can easily be confused with blast pile rock block point clouds, interfering with the accurate segmentation of rock block point clouds and reducing the accuracy of block size identification. Furthermore, the presence of surface point clouds significantly increases the amount of point cloud data and the computational complexity of processing it, affecting the efficiency and accuracy of subsequent segmentation and block size statistics.

[0003] Currently, common methods for extracting ground points from 3D point clouds mainly include cloth simulation filtering algorithms, progressive morphological filtering, and RANSAC plane fitting. However, since the ground in the blast pile area is usually uneven, plane fitting methods have poor robustness, resulting in incomplete removal of ground points. Cloth simulation filtering algorithms have high computational complexity, limited applicability, and are not ideal when processing complex and irregular blast pile point cloud data. Progressive morphological filtering has poor ability to preserve point cloud boundaries, affecting the accuracy of point cloud segmentation. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, this invention adaptively determines the region growth segmentation threshold by analyzing the statistical characteristics of the normal vector angle and curvature of the 3D point cloud data of the blast pile. Based on the calculation and statistics of the distribution concentration, directional height difference, and average roughness characteristics of the segmented point cloud, a joint judgment rule for the ground point cloud is constructed. This invention proposes a method and device for removing blast pile ground point clouds, aiming to improve the quality and usability of 3D point cloud data of blast piles and ensure efficient and accurate identification and statistics of blast pile block size.

[0005] On the one hand, this invention proposes a method for removing point clouds from a blast pile, which includes the following process:

[0006] The original 3D point cloud data of the burst pile is obtained, and the original 3D point cloud data of the burst pile is downsampled based on the voxel downsampling algorithm to obtain the 3D point cloud of the burst pile.

[0007] The segmentation threshold is determined based on the characteristic distribution of the 3D point cloud of the blast pile, and the 3D point cloud of the blast pile is segmented using the region growing algorithm to obtain several segmentation result point clouds.

[0008] Calculate the geometric feature vector of each segmented point cloud separately, and based on all the calculated geometric feature vectors, filter all the segmented point clouds according to the preset filtering conditions to generate the filtered segmented point cloud feature dataset.

[0009] The geometric feature vectors include: distribution concentration, directional height difference, and average roughness;

[0010] Based on the filtered segmentation result point cloud feature dataset, unsupervised clustering analysis is used to extract the ground class feature subset;

[0011] The ground-type feature subset includes: the ground-type distribution concentration subset. Ground-type directed elevation difference subset and the average roughness subset of the ground class ;

[0012] Based on a subset of ground-type features, a joint determination rule for ground point clouds is constructed, and this rule is used to determine whether each segmented point cloud is a ground point cloud or a non-ground point cloud; the specific content is as follows:

[0013] Calculate the subset of ground class distribution concentration mean and standard deviation and using the mean and standard deviation Calculate the distribution concentration threshold ;

[0014] Calculate the directed elevation difference subset of the ground class mean and standard deviation and using the mean and standard deviation Calculate the directed height difference threshold ;

[0015] Calculate the average roughness subset of the ground class mean and standard deviation and using the mean and standard deviation Calculate the average roughness threshold ;

[0016] Based on the distribution concentration threshold Directed height difference threshold and average roughness threshold The joint determination rule for ground point clouds is as follows: point clouds that simultaneously satisfy the following conditions are identified as ground point clouds: distribution concentration greater than the distribution concentration threshold, directed elevation difference less than the directed elevation difference threshold, and average roughness less than the average roughness threshold; otherwise, they are identified as non-ground point clouds.

[0017] Using the joint determination rules of ground point clouds, each segmentation result point cloud is identified as either a ground point cloud or a non-ground point cloud;

[0018] Remove all identified ground point clouds and use a denoising algorithm to remove ground scattered points in all non-ground point clouds to obtain the main point cloud of the blast pile.

[0019] Furthermore, the specific content of determining the segmentation threshold based on the feature distribution of the 3D point cloud of the blast pile, and segmenting the 3D point cloud of the blast pile using a region growing algorithm to obtain several segmented result point clouds is as follows:

[0020] The normal vector of the three-dimensional point cloud of the burst pile is obtained by fitting the local surface.

[0021] The minimum spanning tree method is used to consistently adjust the normal vector direction of the burst pile 3D point cloud;

[0022] Based on the normal vector of the adjusted 3D point cloud of the burst pile, the segmentation threshold of the 3D point cloud of the burst pile is adaptively determined by the frequency statistics method.

[0023] Based on the segmentation threshold, the three-dimensional point cloud of the burst pile is segmented using a region growing algorithm to obtain several segmentation result point clouds.

[0024] Furthermore, the specific content of adaptively determining the segmentation threshold of the 3D point cloud of the burst pile based on the adjusted normal vector of the burst pile using a frequency statistics method is as follows:

[0025] For any point in the bursty 3D point cloud and points neighborhood Based on the normal vector of the adjusted 3D point cloud of the burst pile, calculate the points. With neighboring regions The mean angle between the normal vectors of all points inside the point is used as the reference point. The angle between the normal vectors; simultaneously, calculate the point The curvature;

[0026] Based on the angle between the normal vectors of all points in the 3D point cloud of the burst pile, several equidistant angle intervals between normal vectors are divided, and the frequency and relative frequency of each angle interval are counted. Then, a frequency distribution table of the angle between normal vectors is constructed, and a frequency distribution curve of the angle between normal vectors is plotted.

[0027] Based on the curvature of all points in the 3D point cloud of the burst pile, several equidistant curvature intervals are divided, and the frequency and relative frequency within each curvature interval are counted. In this way, a curvature frequency distribution table is constructed, and a curvature frequency distribution curve is plotted.

[0028] The angle value of the normal vector corresponding to the peak point of the normal vector angle frequency distribution curve is used as the normal vector angle threshold, and the curvature value corresponding to the elbow point of the curvature frequency distribution curve is used as the curvature threshold; wherein the elbow point is: the point on the curvature frequency distribution curve where the perpendicular distance relative to the line connecting the start point and the end point of the curve reaches the maximum value;

[0029] The normal vector angle threshold and the curvature threshold are used together as the segmentation threshold for the burst pile 3D point cloud.

[0030] Furthermore, the specific content of calculating the geometric feature vector of each segmented point cloud, and filtering all segmented point clouds according to preset filtering conditions based on all calculated geometric feature vectors to generate the filtered segmented point cloud feature dataset is as follows:

[0031] For any segmentation result point cloud The segmentation result point cloud The ratio of the number of points in the three-dimensional point cloud of the burst pile to the number of points in the segmented point cloud is used as the distribution concentration of the resulting point cloud. ;in The index value of the segmented point cloud;

[0032] Calculate the point cloud segmentation result Directed elevation difference ;

[0033] Calculate the point cloud segmentation result average roughness ;

[0034] The point cloud was segmented by combining the results. Distribution concentration directional elevation difference and average roughness Construct the segmentation result point cloud geometric eigenvectors ;

[0035] Based on the geometric feature vectors of all segmented point clouds, construct a feature dataset of the segmented point clouds. ;

[0036] Calculate the feature dataset of the segmented point cloud separately. The mean and standard deviation of the distribution concentration, directional elevation difference, and average roughness;

[0037] Based on the calculated mean and standard deviation, if the point cloud is arbitrarily segmented... If the preset filtering conditions are met, the segmentation result point cloud feature dataset will be retained. Mid-segmentation result point cloud geometric eigenvectors Conversely, the feature dataset of the segmented point cloud is used. Remove point cloud from segmentation results geometric eigenvectors This leads to the filtered segmentation result point cloud feature dataset. .

[0038] Furthermore, the preset filtering conditions are:

[0039] ;

[0040] in This represents a geometric characteristic among distribution concentration, directional elevation difference, and average roughness, where Indicates the concentration of distribution. Indicates a directional elevation difference. Indicates average roughness; Represents the point cloud of the segmentation result. The Geometric eigenvalues; Point cloud feature dataset representing segmentation results The Middle The mean of the geometric features of the terms; Point cloud feature dataset representing segmentation results The Middle The standard deviation of the geometric characteristics of the terms; It is a constant.

[0041] Furthermore, the directional height difference The calculation method is as follows:

[0042] Calculate the point cloud segmentation result The covariance matrix is ​​obtained, and three eigenvalues ​​are obtained through eigenvalue decomposition. ,and and the corresponding three feature vectors ;

[0043] With the largest eigenvalue Corresponding feature vector Point cloud as segmentation result In the X-axis direction, with the second largest eigenvalue Corresponding feature vector As the Y-axis direction, the remaining feature vectors As the Z-axis direction, construct the segmentation result point cloud. The directed bounding box;

[0044] eigenvectors along the Z-axis The direction vector is used to segment the point cloud. any point in Calculation points position vector and The dot product, and use that dot product as the dot product. position vector Projected distance along the Z-axis ;

[0045] Find the maximum value among the projected distances corresponding to all points. and minimum value The difference between the maximum and minimum values ​​is used as the segmentation result point cloud. Directed elevation difference along the Z-axis .

[0046] Furthermore, the average roughness The calculation method is as follows:

[0047] For the segmentation result point cloud any point in Perform a k-neighborhood search to obtain the point. neighborhood point set ;

[0048] Based on the neighborhood point set Perform plane fitting and calculate the points. to the fitting plane vertical distance , as a point Roughness;

[0049] Calculate the point cloud segmentation result The average roughness of all points is taken as the average roughness of the resulting point cloud. .

[0050] Furthermore, the specific content of the ground class feature subset extracted from the filtered segmentation result point cloud feature dataset through unsupervised clustering analysis is as follows:

[0051] Extract the segmentation result point cloud feature dataset after filtering Distribution concentration feature array Directed elevation difference feature array and average roughness feature array ;

[0052] Set the number of clusters and apply them to the distribution concentration feature array respectively. Directed elevation difference feature array and average roughness feature array Cluster analysis was performed to obtain the distribution concentration feature array. Clustering results Directed elevation difference feature array Clustering results and average roughness feature array Clustering results ;

[0053] Identify clustering results Ground-based clusters And utilize ground-based clustering All distribution concentrations are used to generate a subset of ground-class distribution concentrations. ;

[0054] Identify clustering results Ground-based clusters And utilize ground-based clustering Generate a subset of directed elevation differences for the ground class from all directed elevation differences. ;

[0055] Identify clustering results Ground-based clusters And utilize ground-based clustering All average roughness values ​​are used to generate a subset of the average roughness values ​​for the ground class. ;

[0056] The ground-type distribution concentration subset Ground-type directed elevation difference subset and the average roughness subset of the ground class This constitutes a subset of ground-type features.

[0057] On the other hand, the present invention proposes a device for removing ground point clouds from explosive piles, the device comprising:

[0058] The point cloud downsampling module is used to acquire the original 3D point cloud data of the burst pile and, based on the voxel downsampling algorithm, downsample the original 3D point cloud data of the burst pile to obtain the 3D point cloud of the burst pile.

[0059] The point cloud segmentation module determines the segmentation threshold based on the characteristic distribution of the 3D point cloud of the blast pile, and uses the region growing algorithm to segment the 3D point cloud of the blast pile to obtain several segmented point clouds.

[0060] The feature filtering module is used to calculate the geometric feature vector of each segmented result point cloud, and based on all the calculated geometric feature vectors, to filter all the segmented result point clouds according to preset filtering conditions, and generate a filtered segmented result point cloud feature dataset.

[0061] The ground feature clustering module is used to extract ground class feature subsets based on the filtered segmented point cloud feature dataset through unsupervised clustering analysis.

[0062] The ground point cloud recognition module is used to construct a joint determination rule for ground point clouds based on a subset of ground-type features, and to use the joint determination rule to determine each segmented point cloud as a ground point cloud or a non-ground point cloud.

[0063] The ground point cloud removal module is used to remove all identified ground point clouds and uses a denoising algorithm to remove ground scattered points in all non-ground point clouds to obtain the main point cloud of the blast pile.

[0064] The beneficial effects of adopting the above technical solution are as follows:

[0065] This invention adaptively determines the segmentation threshold based on the statistical characteristics of the normal vector angle and curvature of 3D point cloud data of a burst pile, and employs a region growing algorithm for segmentation. For different burst pile 3D point cloud data, the segmentation threshold can be adaptively determined based on the distribution of the normal vector angle and curvature characteristics of the burst pile 3D point cloud model. This effectively enhances the robustness of the algorithm, making it applicable to different burst pile scenarios. This method can fully utilize the geometric features of the point cloud, dividing point clouds with similar geometric attributes into independent regions, achieving accurate clustering and over-segmentation of burst pile 3D point clouds, thus providing reliable input for subsequent identification and extraction of ground point clouds.

[0066] This invention constructs a multi-feature joint ground point cloud discrimination method based on distribution concentration, directed elevation difference, and average roughness, which can realize the automatic identification and extraction of ground point clouds. Based on the segmentation results of the 3D point cloud region growth from the burst pile, the distribution concentration, directed elevation difference, and average roughness of each segmented point cloud are statistically analyzed. According to the joint ground point cloud judgment rules, accurate identification and automatic extraction of ground point clouds are achieved, exhibiting a high degree of automation and accuracy, and effectively avoiding manual intervention. Attached Figure Description

[0067] Figure 1 This is a flowchart of a method for removing ground point clouds from a blast pile, as described in an embodiment of the present invention.

[0068] Figure 2 This refers to the original three-dimensional point cloud data of the burst pile in this embodiment of the invention;

[0069] Figure 3 This refers to the burst pile three-dimensional point cloud data of an embodiment of the present invention;

[0070] Figure 4 This is a schematic diagram of the point cloud segmentation threshold determination method in an embodiment of the present invention;

[0071] Figure 5This is a three-dimensional point cloud segmentation result of a burst pile according to an embodiment of the present invention;

[0072] Figure 6 This is a schematic diagram of a directed bounding box for a three-dimensional point cloud of a burst pile in an embodiment of the present invention;

[0073] Figure 7 This is a statistical result diagram of the distribution concentration, directional elevation difference and average roughness of the three-dimensional point cloud segmentation result of the explosive pile in an embodiment of the present invention.

[0074] Figure 8 This is a ground point cloud result image extracted using the ground point cloud joint determination rule in this embodiment of the invention;

[0075] Figure 9 This is a point cloud image of the main body of the blast pile after removing the ground point cloud in an embodiment of the present invention;

[0076] Figure 10 This is a structural diagram of a ground point cloud removal device for explosive piles, as described in an embodiment of the present invention. Detailed Implementation

[0077] To facilitate understanding of this application, specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following embodiments are illustrative of the invention but are not intended to limit its scope. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0078] Example 1:

[0079] This embodiment provides a method for removing point clouds from a burst-prone ground surface, such as... Figure 1 As shown, the method includes the following steps:

[0080] The original 3D point cloud data of the burst pile is obtained, and the original 3D point cloud data of the burst pile is downsampled based on the voxel downsampling algorithm to obtain the 3D point cloud of the burst pile.

[0081] In this embodiment, a handheld 3D scanner is used. Scanner parameters are set according to the scanner's characteristics, the area of ​​the rock pile, and its fragmentation pattern. A scanning path is planned, and multi-angle, multi-position scanning is performed according to the predetermined scanning path. Automatic stitching is also performed during the scanning process. The obtained original 3D point cloud data of the blast pile is as follows: Figure 2 As shown, the number of points in the 3D point cloud of the burst pile is 1,176,695. Figure 3 As shown, based on the voxel downsampling algorithm, the original 3D point cloud data of the burst pile is downsampled to obtain a 3D point cloud of the burst pile containing 392,752 points.

[0082] The segmentation threshold is determined based on the characteristic distribution of the 3D point cloud of the blast pile, and the 3D point cloud of the blast pile is segmented using the region growing algorithm to obtain several segmentation result point clouds.

[0083] The specific content of the segmentation result point cloud obtained by determining the segmentation threshold based on the feature distribution of the 3D point cloud of the blast pile and segmenting the 3D point cloud of the blast pile using the region growing algorithm is as follows:

[0084] The normal vector of the three-dimensional point cloud of the burst pile is obtained by fitting the local surface.

[0085] In this embodiment, for each point in the 3D point cloud of the burst pile, the points in its k-neighborhood are found, and the normal vector of the point is obtained by fitting the local neighborhood, thereby obtaining the normal vector of the 3D point cloud of the burst pile.

[0086] The minimum spanning tree method is used to consistently adjust the normal vector direction of the burst pile 3D point cloud.

[0087] In this embodiment, a minimum spanning tree is constructed based on the normal vectors of the bursty 3D point cloud. A seed point with a known normal vector direction is selected as the root node, and the minimum spanning tree is traversed starting from this root node. For each traversed child node, it is determined whether its normal vector direction is consistent with the normal vector direction of its parent node. The normal vector directions of inconsistent child nodes are inverted until all nodes in the minimum spanning tree have been traversed, thereby ensuring that the normal vector direction of all points is consistent with that of the seed point.

[0088] Based on the adjusted normal vector of the burst pile 3D point cloud, the segmentation threshold of the burst pile 3D point cloud is adaptively determined by frequency statistics.

[0089] The specific details of adaptively determining the segmentation threshold of the 3D point cloud of the burst pile based on the adjusted normal vector of the burst pile using a frequency statistics method are as follows:

[0090] For any point in the bursty 3D point cloud and points neighborhood Based on the normal vector of the adjusted 3D point cloud of the burst pile, calculate the points. With neighboring regions The mean angle between the normal vectors of all points inside the point is used as the reference point. The angle between the normal vectors; simultaneously, calculate the point The curvature.

[0091] Based on the angle between the normal vectors of all points in the burst pile 3D point cloud, several equidistant angle intervals between normal vectors are divided, and the frequency and relative frequency of each angle interval are counted. Then, a frequency distribution table of angle between normal vectors is constructed, and a frequency distribution curve of angle between normal vectors is plotted.

[0092] Based on the curvature of all points in the 3D point cloud of the burst, several equidistant curvature intervals are divided, and the frequency and relative frequency within each curvature interval are counted. In this way, a curvature frequency distribution table is constructed, and a curvature frequency distribution curve is plotted.

[0093] In this embodiment, frequency analysis is used to perform frequency statistical analysis on the normal vector angle and curvature of the three-dimensional point cloud of the explosion pile. First, the two types of characteristic variables (normal vector angle and curvature) are grouped into equal intervals, and then the sample frequency and frequency within each interval are counted to construct a frequency distribution table and draw the frequency distribution curves of normal vector angle and curvature.

[0094] The angle value of the normal vector corresponding to the peak point of the frequency distribution curve of the normal vector angle is used as the normal vector angle threshold, and the curvature value corresponding to the elbow point of the curvature frequency distribution curve is used as the curvature threshold.

[0095] The elbow point is defined as the point on the curvature frequency distribution curve where the perpendicular distance from the line connecting the start and end points of the curve reaches its maximum value.

[0096] The normal vector angle threshold and the curvature threshold are used together as the segmentation threshold for the burst pile 3D point cloud.

[0097] Based on the segmentation threshold, the three-dimensional point cloud of the burst pile is segmented using a region growing algorithm to obtain several segmentation result point clouds.

[0098] In this embodiment, as Figure 4 As shown, the peak point of the frequency distribution curve of the normal vector angle and the elbow point of the frequency distribution curve of the curvature are selected as the normal vector angle threshold and curvature threshold, respectively. A region growing algorithm is used to segment the 3D point cloud data of the burst pile, and the segmentation results are obtained. The 3D point cloud segmentation results of the burst pile, which are then assigned random colors, are shown below. Figure 5 As shown, a total of 835 segmentation results were generated.

[0099] Calculate the geometric feature vector for each segmented point cloud, and based on all the calculated geometric feature vectors, filter all the segmented point clouds according to the preset filtering conditions to generate a filtered segmented point cloud feature dataset.

[0100] To intuitively verify the effectiveness of the method proposed in this invention, based on the segmentation results of the three-dimensional point cloud of the blast pile used in this embodiment, all segmented point clouds are arranged in descending order of the number of points, and the geometric feature vectors of the top 100 segmented point clouds are calculated and statistically analyzed, including: distribution concentration, directional height difference and average roughness, in order to distinguish the ground point cloud.

[0101] The specific content of calculating the point cloud geometric feature vector for each segmentation result, and filtering all segmentation result point clouds according to preset filtering conditions based on all calculated geometric feature vectors to generate the filtered segmentation result point cloud feature dataset is as follows:

[0102] The geometric feature vector includes: distribution concentration, directional height difference, and average roughness.

[0103] For any segmentation result point cloud The segmentation result point cloud The ratio of the number of points in the three-dimensional point cloud of the burst pile to the number of points in the segmented point cloud is used as the distribution concentration of the resulting point cloud. ;in The index value for the segmented point cloud.

[0104] Calculate the point cloud segmentation result Directed elevation difference .

[0105] The directed elevation difference The calculation method is as follows:

[0106] Calculate the point cloud segmentation result The covariance matrix is ​​obtained, and three eigenvalues ​​are obtained through eigenvalue decomposition. ,and and the corresponding three feature vectors .

[0107] With the largest eigenvalue Corresponding feature vector Point cloud as segmentation result In the X-axis direction, with the second largest eigenvalue Corresponding feature vector As the Y-axis direction, the remaining feature vectors As the Z-axis direction, construct the segmentation result point cloud. Directed bounding boxes, such as Figure 6 As shown.

[0108] eigenvectors along the Z-axis The direction vector is used to segment the point cloud. any point in Calculation points position vector and The dot product, and use that dot product as the dot product. position vector Projected distance along the Z-axis .

[0109] ;

[0110] Projection length It is a one-dimensional scalar used to characterize the directed distance of a point relative to the Z-axis reference plane.

[0111] Find the maximum value among the projected distances corresponding to all points. and minimum value The difference between the maximum and minimum values ​​is used as the segmentation result point cloud. Directed elevation difference along the Z-axis .

[0112] Calculate the point cloud segmentation result average roughness .

[0113] The average roughness The calculation method is as follows:

[0114] For the segmentation result point cloud any point in Perform a k-neighborhood search to obtain the point. neighborhood point set .

[0115] Based on the neighborhood point set Perform plane fitting and calculate the points. to the fitting plane vertical distance , as a point The roughness.

[0116] Calculate the point cloud segmentation result The average roughness of all points is taken as the average roughness of the resulting point cloud. .

[0117] The point cloud was segmented by combining the results. Distribution concentration directional elevation difference and average roughness Construct the segmentation result point cloud geometric eigenvectors .

[0118] Based on the geometric feature vectors of all segmented point clouds, construct a feature dataset of the segmented point clouds. ,and ;in The number of point clouds in the segmentation result.

[0119] Calculate the feature dataset of the segmented point cloud separately. Mean values ​​of distribution concentration, directional elevation difference, and average roughness and standard deviation .

[0120] For any segmentation result point cloud If the segmentation result is a point cloud If the preset filtering conditions are met, the segmented point cloud results will be retained. Conversely, the feature dataset of the segmented point cloud is used. Remove point cloud from segmentation results geometric eigenvectors This leads to the filtered segmentation result point cloud feature dataset. .

[0121] The preset filtering conditions are:

[0122] ;

[0123] in This represents a geometric characteristic among distribution concentration, directional elevation difference, and average roughness, where Indicates the concentration of distribution. Indicates a directional elevation difference. Indicates average roughness; Represents the point cloud of the segmentation result. The Geometric eigenvalues; Point cloud feature dataset representing segmentation results The Middle The mean of the geometric features of the terms; Point cloud feature dataset representing segmentation results The Middle The standard deviation of the geometric characteristics of the terms; It is a constant.

[0124] In this embodiment, the following is adopted: The criteria for constructing the preset filtering conditions are as follows:

[0125] ;

[0126] The above filtering criteria can eliminate the interference of extreme values ​​on subsequent threshold calculations.

[0127] Based on the filtered segmentation result point cloud feature dataset, a ground class feature subset is extracted through unsupervised clustering analysis.

[0128] In this embodiment, K-means clustering analysis is used to extract the feature subset of the ground class.

[0129] The specific content of the ground class feature subset extracted from the filtered segmentation result point cloud feature dataset through unsupervised clustering analysis is as follows:

[0130] Extract the segmentation result point cloud feature dataset after filtering Distribution concentration feature array Directed elevation difference feature array and average roughness feature array .

[0131] In this embodiment, the feature dataset of the segmented result points cloud is selected. In the process, the distribution concentration, directional elevation difference, and average roughness of all selected point cloud segmentation results are extracted respectively, and three independent one-dimensional feature arrays are constructed as follows: Distribution concentration feature array: Directed elevation difference feature array: Average roughness feature array: ;in , , The distribution concentration, directional elevation difference, and average roughness of the point cloud segmentation result corresponding to the index value of 2 are respectively.

[0132] Set the number of clusters and apply them to the distribution concentration feature array respectively. Directed elevation difference feature array and average roughness feature array Cluster analysis was performed to obtain the distribution concentration feature array. Clustering results Directed elevation difference feature array Clustering results and average roughness feature array Clustering results .

[0133] In this embodiment, the distribution concentration, directed elevation difference and average roughness feature data are represented as a one-dimensional array. According to the K-means clustering algorithm, the number of clusters is set to 2 (ground class and non-ground class), and the random seed is fixed at 42 to ensure that the clustering results are reproducible.

[0134] In this embodiment, ground clusters are determined based on the physical meaning of the features. Specifically, for the distribution concentration feature, the ground point cloud has stronger spatial clustering and a higher distribution concentration value. Therefore, the cluster with a larger cluster center value is selected as the ground feature subset. For the directional elevation difference and average roughness features, the ground point cloud has smaller topographic undulations and a lower feature value. Therefore, the cluster with a smaller cluster center value is selected as the ground feature subset.

[0135] Identify clustering results Ground-based clusters And utilize ground-based clustering All distribution concentrations are used to generate a subset of ground-class distribution concentrations. .

[0136] Identify clustering results Ground-based clusters And utilize ground-based clustering Generate a subset of directed elevation differences for the ground class from all directed elevation differences. .

[0137] Identify clustering results Ground-based clusters And utilize ground-based clustering All average roughness values ​​are used to generate a subset of the average roughness values ​​for the ground class. .

[0138] The ground-type distribution concentration subset Ground-type directed elevation difference subset and the average roughness subset of the ground class This constitutes a subset of ground-type features.

[0139] Based on the ground-type feature subset, a joint determination rule for ground point clouds is constructed, and the joint determination rule for ground point clouds is used to determine whether each segmented point cloud is a ground point cloud or a non-ground point cloud.

[0140] The specific content of constructing a joint determination rule for ground point clouds based on a subset of ground-type features, and using the joint determination rule to determine whether each segmented point cloud is a ground point cloud or a non-ground point cloud, is as follows:

[0141] Calculate the subset of ground class distribution concentration mean and standard deviation and using the mean and standard deviation Calculate the distribution concentration threshold .

[0142] The distribution concentration threshold Represented as:

[0143] ;

[0144] Calculate the directed elevation difference subset of the ground class mean and standard deviation and using the mean and standard deviation Calculate the directed height difference threshold .

[0145] The directed height difference threshold Represented as:

[0146] ;

[0147] Calculate the average roughness subset of the ground class mean and standard deviation and using the mean and standard deviation Calculate the average roughness threshold .

[0148] The average roughness threshold Represented as:

[0149] ;

[0150] Based on the distribution concentration threshold Directed height difference threshold and average roughness threshold The joint determination rule for ground point clouds is as follows: point clouds that simultaneously satisfy the following conditions are identified as ground point clouds: distribution concentration greater than the distribution concentration threshold, directed elevation difference less than the directed elevation difference threshold, and average roughness less than the average roughness threshold; otherwise, they are identified as non-ground point clouds.

[0151] The joint determination rule for ground point clouds is expressed as follows:

[0152] ;

[0153] Specifically, when the distribution concentration of the point cloud is greater than The directional height difference is less than and average roughness less than When the feature is determined to match the features of a ground point cloud, it is determined that the feature matches the features of a ground point cloud.

[0154] In this embodiment, the statistical results of the distribution concentration, directional elevation difference, and average roughness of the point clouds of the first 100 segmentation results are as follows: Figure 7 As shown, the calculated distribution concentration, directional elevation difference, and average roughness threshold are 0.0196, 233.3563, and 0.2537, respectively. Therefore, the joint determination rule for ground point clouds in this embodiment is expressed as follows:

[0155] ;

[0156] By using the joint determination rules of ground point clouds, each segmentation result point cloud is identified as either a ground point cloud or a non-ground point cloud.

[0157] In this embodiment, for any segmented point cloud, if the distribution concentration, directional elevation difference, and average roughness satisfy the above determination rules, the segmented point cloud is determined to be a ground point cloud; otherwise, it is determined to be a non-ground point cloud. The determined and extracted ground point cloud is as follows: Figure 8 As shown.

[0158] Remove all identified ground point clouds and use a denoising algorithm to remove ground scattered points in all non-ground point clouds to obtain the main point cloud of the blast pile.

[0159] In this embodiment, after removing a large amount of ground point cloud data, the ground points in all remaining non-ground point clouds can be directly deleted using CloudCompare software. CloudCompare is a commonly used open-source point cloud processing software; ground points can be selected and deleted using this software. Alternatively, commonly used denoising algorithms in the field, such as statistical filtering algorithms, can be used to remove ground points to obtain the main point cloud of the burst pile. Figure 9 As shown.

[0160] Example 2:

[0161] This embodiment proposes a device for removing point clouds from the ground caused by a blast pile, such as... Figure 10 As shown, the device includes:

[0162] The point cloud downsampling module is used to acquire the original 3D point cloud data of the burst pile and, based on the voxel downsampling algorithm, downsample the original 3D point cloud data of the burst pile to obtain the 3D point cloud of the burst pile.

[0163] The point cloud segmentation module determines the segmentation threshold based on the characteristic distribution of the 3D point cloud of the burst pile, and uses the region growing algorithm to segment the 3D point cloud of the burst pile to obtain several segmented point clouds.

[0164] The feature filtering module is used to calculate the geometric feature vector of each segmented point cloud, and based on all the calculated geometric feature vectors, to filter all the segmented point clouds according to preset filtering conditions, generating a filtered segmented point cloud feature dataset.

[0165] The ground feature clustering module is used to extract ground class feature subsets based on the filtered segmented point cloud feature dataset through unsupervised clustering analysis.

[0166] The ground point cloud recognition module is used to construct a joint determination rule for ground point clouds based on a subset of ground-type features, and to use the joint determination rule to determine whether each segmented point cloud is a ground point cloud or a non-ground point cloud.

[0167] The ground point cloud removal module is used to remove all identified ground point clouds and uses a denoising algorithm to remove ground scattered points in all non-ground point clouds to obtain the main point cloud of the blast pile.

[0168] Example 3:

[0169] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors perform the aforementioned method for removing burst-type ground point clouds.

[0170] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements a method for removing point clouds from a blast-prone ground surface as described in the embodiments. It is understood that the electronic device may also include an input / output (I / O) interface and communication components.

[0171] The processor is used to execute all or part of the steps in the method for removing burst-type ground point clouds as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0172] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the blast-propelled ground point cloud removal method described in the above embodiments.

[0173] Example 4:

[0174] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0175] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method for removing burst-heap ground point clouds as described in the various embodiments of this application.

[0176] The aforementioned storage media include: flash memory, hard disk, multimedia card, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disk, optical disk, server, APP (Application) application store, and other media capable of storing program verification codes, on which computer programs are stored. When the computer programs are executed by the processor, they can implement the various steps of the above-described method for removing dense ground point clouds.

[0177] The storage media include, but are not limited to: mobile storage devices, read-only memory (ROM), flash memory cards, hard disks, optical disks, computer memory, random access memory (RAM), electrical carrier signals, telecommunication signals, and other media capable of storing program code.

[0178] Example 5:

[0179] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned method for removing point clouds from blast-prone ground surfaces.

[0180] The computer program or instructions may include multiple modules, each corresponding to a program segment that implements a different function of the present invention. Modules can exist independently or work together to ultimately complete the task of identifying and extracting the ground point cloud generated during the blasting process. It should be noted that the computer program can be in source code form, object code form, executable file, or other intermediate format. Through the execution of the computer program on the processor, the identification and extraction process of the ground point cloud generated during the blasting process can be completed, thereby effectively identifying and extracting the ground point cloud data generated during the blasting process.

[0181] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0182] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0183] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of this disclosure and its equivalents, then the intent of this disclosure also includes these modifications and variations.

Claims

1. A method for removing point clouds from a blasted surface, characterized in that, This method includes the following steps: The original 3D point cloud data of the burst pile is obtained, and the original 3D point cloud data of the burst pile is downsampled based on the voxel downsampling algorithm to obtain the 3D point cloud of the burst pile. The segmentation threshold is determined based on the characteristic distribution of the three-dimensional point cloud of the blast pile, and the three-dimensional point cloud of the blast pile is segmented using the region growing algorithm to obtain several segmentation result point clouds. Calculate the geometric feature vector of each segmented point cloud separately, and based on all the calculated geometric feature vectors, filter all the segmented point clouds according to the preset filtering conditions to generate the filtered segmented point cloud feature dataset. The geometric feature vectors include: distribution concentration, directional height difference, and average roughness; Based on the filtered segmentation result point cloud feature dataset, unsupervised clustering analysis is used to extract the ground class feature subset; The ground-type feature subset includes: the ground-type distribution concentration subset. Ground-type directed elevation difference subset and the average roughness subset of the ground class ; Based on a subset of ground-type features, a joint determination rule for ground point clouds is constructed, and this rule is used to determine whether each segmented point cloud is a ground point cloud or a non-ground point cloud; the specific content is as follows: Calculate the subset of ground class distribution concentration mean and standard deviation and using the mean and standard deviation Calculate the distribution concentration threshold ; Calculate the directed elevation difference subset of the ground class mean and standard deviation and using the mean and standard deviation Calculate the directed height difference threshold ; Calculate the average roughness subset of the ground class mean and standard deviation and using the mean and standard deviation Calculate the average roughness threshold ; Based on the distribution concentration threshold Directed height difference threshold and average roughness threshold The joint determination rule for ground point clouds is as follows: point clouds that simultaneously satisfy the following conditions are identified as ground point clouds: distribution concentration greater than the distribution concentration threshold, directed elevation difference less than the directed elevation difference threshold, and average roughness less than the average roughness threshold; otherwise, they are identified as non-ground point clouds. Using the joint determination rules of ground point clouds, each segmentation result point cloud is identified as either a ground point cloud or a non-ground point cloud; Remove all identified ground point clouds and use a denoising algorithm to remove ground scattered points in all non-ground point clouds to obtain the main point cloud of the blast pile.

2. The method for removing ground point clouds from explosive piles according to claim 1, characterized in that, The specific content of the segmentation result point cloud obtained by determining the segmentation threshold based on the feature distribution of the 3D point cloud of the blast pile and segmenting the 3D point cloud of the blast pile using the region growing algorithm is as follows: The normal vector of the three-dimensional point cloud of the burst pile is obtained by fitting the local surface. The minimum spanning tree method is used to consistently adjust the normal vector direction of the burst pile 3D point cloud; Based on the normal vector of the adjusted 3D point cloud of the burst pile, the segmentation threshold of the 3D point cloud of the burst pile is adaptively determined by the frequency statistics method. Based on the segmentation threshold, the three-dimensional point cloud of the burst pile is segmented using a region growing algorithm to obtain several segmentation result point clouds.

3. The method for removing ground point clouds from explosive piles according to claim 2, characterized in that, The specific details of adaptively determining the segmentation threshold of the 3D point cloud of the burst pile based on the adjusted normal vector of the burst pile using a frequency statistics method are as follows: For any point in the bursty 3D point cloud and points neighborhood Based on the normal vector of the adjusted 3D point cloud of the burst pile, calculate the points. With neighboring regions The mean angle between the normal vectors of all points inside the point is used as the reference point. The angle between the normal vectors; simultaneously, calculate the point The curvature; Based on the angle between the normal vectors of all points in the 3D point cloud of the burst pile, several equidistant angle intervals between normal vectors are divided, and the frequency and relative frequency of each angle interval are counted. Then, a frequency distribution table of the angle between normal vectors is constructed, and a frequency distribution curve of the angle between normal vectors is plotted. Based on the curvature of all points in the 3D point cloud of the burst pile, several equidistant curvature intervals are divided, and the frequency and relative frequency within each curvature interval are counted. In this way, a curvature frequency distribution table is constructed, and a curvature frequency distribution curve is plotted. The angle value of the normal vector corresponding to the peak point of the normal vector angle frequency distribution curve is used as the normal vector angle threshold, and the curvature value corresponding to the elbow point of the curvature frequency distribution curve is used as the curvature threshold; wherein the elbow point is: the point on the curvature frequency distribution curve where the perpendicular distance relative to the line connecting the start point and the end point of the curve reaches the maximum value; The normal vector angle threshold and the curvature threshold are used together as the segmentation threshold for the burst pile 3D point cloud.

4. The method for removing ground point clouds from explosive piles according to claim 1, characterized in that, The specific content of calculating the geometric feature vector of each segmented point cloud, and filtering all segmented point clouds according to preset filtering conditions based on all calculated geometric feature vectors to generate the filtered segmented point cloud feature dataset is as follows: For any segmentation result point cloud The segmentation result point cloud The ratio of the number of points in the three-dimensional point cloud of the burst pile to the number of points in the segmented point cloud is used as the distribution concentration of the resulting point cloud. ;in The index value of the segmented point cloud; Calculate the point cloud segmentation result Directed elevation difference ; Calculate the point cloud segmentation result average roughness ; The point cloud was segmented by combining the results. Distribution concentration directional elevation difference and average roughness Construct the segmentation result point cloud geometric eigenvectors ; Based on the geometric feature vectors of all segmented point clouds, construct a feature dataset of the segmented point clouds. ; Calculate the feature dataset of the segmented point cloud separately. The mean and standard deviation of the distribution concentration, directional elevation difference, and average roughness; Based on the calculated mean and standard deviation, if the point cloud is arbitrarily segmented... If the preset filtering conditions are met, the segmentation result point cloud feature dataset will be retained. Mid-segmentation result point cloud geometric eigenvectors Conversely, the feature dataset of the segmented point cloud is used. Remove point cloud from segmentation results geometric eigenvectors This leads to the filtered segmentation result point cloud feature dataset. .

5. The method for removing ground point clouds from explosive piles according to claim 4, characterized in that, The preset filtering conditions are: ； in This represents a geometric characteristic among distribution concentration, directional elevation difference, and average roughness, where Indicates the concentration of distribution. Indicates a directional elevation difference. Indicates average roughness; Represents the point cloud of the segmentation result. The Geometric eigenvalues; Point cloud feature dataset representing segmentation results The Middle The mean of the geometric features of the terms; Point cloud feature dataset representing segmentation results The Middle The standard deviation of the geometric characteristics of the terms; It is a constant.

6. The method for removing ground point clouds from explosive piles according to claim 4, characterized in that, The directed elevation difference The calculation method is as follows: Calculate the point cloud segmentation result The covariance matrix is ​​obtained, and three eigenvalues ​​are obtained through eigenvalue decomposition. ,and and the corresponding three feature vectors ; With the largest eigenvalue Corresponding feature vector Point cloud as segmentation result In the X-axis direction, with the second largest eigenvalue Corresponding feature vector As the Y-axis direction, the remaining feature vectors As the Z-axis direction, construct the segmentation result point cloud. The directed bounding box; Eigenvectors along the Z-axis The direction vector is used to segment the point cloud. any point in Calculation points position vector and The dot product, and use that dot product as the dot product. position vector Projected distance along the Z-axis ; Find the maximum value among the projected distances corresponding to all points. and minimum value The difference between the maximum and minimum values ​​is used as the segmentation result point cloud. Directed elevation difference along the Z-axis .

7. The method for removing ground point clouds from explosive piles according to claim 4, characterized in that, The average roughness The calculation method is as follows: For the segmentation result point cloud any point in Perform a k-neighborhood search to obtain the point. neighborhood point set ; Based on the neighborhood point set Perform plane fitting and calculate the points. to the fitting plane vertical distance , as a point Roughness; Calculate the point cloud segmentation result The average roughness of all points is taken as the average roughness of the resulting point cloud. .

8. The method for removing ground point clouds from explosive piles according to claim 1, characterized in that, The specific content of the ground class feature subset extracted from the filtered segmentation result point cloud feature dataset through unsupervised clustering analysis is as follows: Extract the segmentation result point cloud feature dataset after filtering Distribution concentration feature array Directed elevation difference feature array and average roughness feature array ; Set the number of clusters and apply them to the distribution concentration feature array respectively. Directed elevation difference feature array and average roughness feature array Cluster analysis was performed to obtain the distribution concentration feature array. Clustering results Directed elevation difference feature array Clustering results and average roughness feature array Clustering results ; Identify clustering results Ground-based clusters And utilize ground-based clustering All distribution concentrations are used to generate a subset of ground-class distribution concentrations. ; Identify clustering results Ground-based clusters And utilize ground-based clustering Generate a subset of directed elevation differences for the ground class from all directed elevation differences. ; Identify clustering results Ground-based clusters And utilize ground-based clustering All average roughness values ​​are used to generate a subset of the average roughness values ​​for the ground class. ; The ground-type distribution concentration subset Ground-type directed elevation difference subset and the average roughness subset of the ground class This constitutes a subset of ground-type features.

9. A device for removing point clouds from the ground of a blasted pile, used to implement the method for removing point clouds from the ground of a blasted pile as described in any one of claims 1-8, characterized in that, The device includes: The point cloud downsampling module is used to acquire the original 3D point cloud data of the burst pile and, based on the voxel downsampling algorithm, downsample the original 3D point cloud data of the burst pile to obtain the 3D point cloud of the burst pile. The point cloud segmentation module determines the segmentation threshold based on the characteristic distribution of the 3D point cloud of the blast pile, and uses the region growing algorithm to segment the 3D point cloud of the blast pile to obtain several segmented point clouds. The feature filtering module is used to calculate the geometric feature vector of each segmented result point cloud, and based on all the calculated geometric feature vectors, to filter all the segmented result point clouds according to preset filtering conditions, and generate a filtered segmented result point cloud feature dataset. The ground feature clustering module is used to extract ground class feature subsets based on the filtered segmented point cloud feature dataset through unsupervised clustering analysis. The ground point cloud recognition module is used to construct a joint determination rule for ground point clouds based on a subset of ground-type features, and to use the joint determination rule to determine each segmented point cloud as a ground point cloud or a non-ground point cloud. The ground point cloud removal module is used to remove all identified ground point clouds and uses a denoising algorithm to remove ground scattered points in all non-ground point clouds to obtain the main point cloud of the blast pile.

Images