A method for identifying rock debris pieces in a tunnel excavation process, a terminal device, and a medium

Abstract

The application is suitable for the field of machine vision recognition technology, and provides a rock debris piece identification method in a tunnel excavation process, a terminal device and a medium. Point cloud data of a rock debris piece to be identified is collected, and the point cloud data is voxelized. A quick point feature histogram corresponding to each voxel is constructed, spatial position coordinate values, curvature values and cross kernels of the quick point feature histogram of the voxel are obtained, and the similarity between the voxel and other voxels is calculated. Super voxel segmentation is performed according to the similarity, and a plurality of super voxels are obtained. The plurality of super voxels are coarsely segmented. Local gradient information of each point data in the coarse segmentation result is calculated, and the domain radius thereof is updated based on the local gradient information, and a plurality of point cloud clusters are obtained. The bounding box algorithm is used to obtain a three-dimensional size long axis and a three-dimensional size short axis of each point cloud cluster. The rock debris piece to be identified is identified according to the three-dimensional size long axis and the three-dimensional size short axis. The application can improve the accuracy of rock debris piece identification in the excavation process.

Description

Technical Field

[0001] This application belongs to the field of machine vision recognition technology, and in particular relates to a method, terminal equipment and medium for recognizing rock fragments during tunnel excavation. Background Technology

[0002] During tunnel excavation, hard rock tunnel boring machine operators need to determine the material of the rock mass ahead based on the excavated material, and then adjust the control parameters of the hard rock tunnel boring machine accordingly to ensure construction safety.

[0003] Traditional methods for identifying rock fragments often rely on the operator's construction experience, judging the type of rock fragments by visual observation. However, this method is limited by visual acuity and ambient lighting, resulting in low accuracy. Therefore, there is an urgent need for a method that can accurately identify rock fragments during tunnel excavation. Summary of the Invention

[0004] This application provides a method, terminal equipment, and medium for identifying rock fragments during tunnel excavation, which can solve the problem of low accuracy in identifying rock fragments during tunnel excavation.

[0005] In a first aspect, this application provides a method for identifying rock fragments during tunnel excavation, including:

[0006] Point cloud data of rock fragments to be identified during tunnel excavation is collected, and the point cloud data is voxelized to obtain multiple voxels; wherein, the point cloud data includes multiple point data, and each voxel in the multiple voxels includes at least one point data.

[0007] For each voxel in the multiple voxels, a fast point feature histogram is constructed to obtain the voxel's spatial location coordinates, curvature values, and cross kernel of the fast point feature histogram. Based on the spatial location coordinates, curvature values, and cross kernel, the similarity between the voxel and other voxels is obtained.

[0008] Multiple voxels are segmented into supervoxels based on similarity to obtain multiple supervoxels;

[0009] Clustering based on concavity and convexity of multiple supervoxels yields coarse segmentation results;

[0010] Calculate the local gradient information corresponding to each point data in the coarse segmentation result, update the neighborhood radius of the point data according to the local gradient information, and obtain multiple point cloud clusters based on density-based clustering algorithm;

[0011] For each point cloud cluster in multiple point cloud clusters, the bounding box algorithm is used to obtain the major axis and minor axis of the three-dimensional dimensions of the point cloud cluster respectively; the major axis represents the longest line segment in the point cloud cluster, and the minor axis represents the shortest line segment in the point cloud cluster.

[0012] The rock fragments to be identified are determined based on the major and minor axes of the three-dimensional dimensions.

[0013] Optionally, based on spatial location coordinates, curvature values, and crossover kernels, the similarity between voxels and other voxels is obtained, including:

[0014] Through calculation formula

[0015]

[0016]

[0017]

[0018] Obtain the similarity D i，j ; where D i，j DS represents the similarity between voxels i and j, where i, j = 1, 2, ..., M, i ≠ j, and M represents the total number of voxels. i，j Let represent the Euclidean distance between the i-th voxel and the j-th voxel in three-dimensional space. DC represents the cross kernel of the fast point feature histogram corresponding to the i-th voxel. i ω represents the curvature value of the i-th voxel. dc The weight ω represents the curvature value. ds The weights ω represent the Euclidean distance. HIK R represents the weights of the crossover kernel. seed N represents spatial resolution. l L represents the number of matches between any two adjacent data distributions in the fast point feature histogram, and L represents the total number of data distributions in the fast point feature histogram, l = 1, 2, ..., L, (x i y i , z i (x) represents the spatial coordinates of the i-th voxel. j y j , z j ) represents the spatial coordinates of the j-th voxel, and m represents the maximum curvature value among multiple voxels.

[0019] Optionally, multiple voxels can be segmented into multiple supervoxels based on similarity, including:

[0020] Step 1: Select any one of the multiple voxels as the current seed voxel;

[0021] Step 2: For each voxel except the current seed voxel, if the similarity between the voxel and the current seed voxel is greater than a preset threshold, then add the voxel to the region where the current seed voxel is located; otherwise, use the voxel as the seed voxel to obtain the seed voxel set.

[0022] Step 3: After traversing all voxels except the current seed voxel, remove the current seed voxel from the seed voxel set. If the seed voxel set after removing the current seed voxel is not empty, take any seed voxel in the seed voxel set after removing the current seed voxel as the current seed voxel in Step 2 and return to execute Step 2; otherwise, based on the execution result of Step 2, obtain multiple super voxels.

[0023] Optionally, clustering based on convexity and concavity is performed on multiple supervoxels to obtain coarse segmentation results, including:

[0024] Through calculation formula

[0025]

[0026] Obtain the concavity-convexity relationship between multiple supervoxels in, Represents the u-th supervoxel With the vth supervoxel The concave-convex relationship between them Let represent the normal vector of the u-th hypervoxel. Let v represent the normal vector of the v-th hypervoxel. Let represent the centroid vector of the u-th hypervoxel. Let β represent the centroid vector of the v-th hypervoxel. Thresh θ represents the threshold difference between the angle between the normal vector of the u-th supervoxel and the line connecting the centroids of the v-th supervoxel and the line of intersection;

[0027] Based on the concavity-convexity relationship, a region growing algorithm is used to segment multiple supervoxels to obtain coarse segmentation results.

[0028] Optionally, the local gradient information corresponding to each point data in the coarse segmentation result is calculated, and the neighborhood radius of the point data is updated according to the local gradient information. A density-based clustering algorithm is then used to obtain multiple point cloud clusters, including:

[0029] Through calculation formula

[0030]

[0031] Obtain local gradient information LP(P) i ); where LP(P i ) represents the data P of the i-th point. iThe corresponding gradient information, T j This represents the gradient values ​​of the data point Pi in the eight directions at the i-th point. This represents the average gradient value within the neighborhood of the i-th data point Pi;

[0032] Through calculation formula

[0033]

[0034] Obtain the radius R of the domain i Among them, R i Represents the data P of the i-th point. i The radius of the domain, LP max This represents the maximum value in the local gradient information corresponding to all point data;

[0035] For each point data point, the data is segmented according to the neighborhood radius of the point data point to obtain multiple point cloud clusters.

[0036] Optionally, the rock fragments to be identified are determined based on their major and minor axes in three dimensions, including:

[0037] Calculate formula R n =B n / A n The recognition parameter R is obtained. n Among them, R n B represents the identification parameters of the nth point cloud cluster. n Let A represent the minor axis of the three-dimensional dimension of the nth point cloud cluster. n The major axis of the three-dimensional dimension of the nth point cloud cluster is represented by n = 1, 2, ..., Num, where Num represents the total number of point cloud clusters;

[0038] If R n If the value is less than 0.5, the rock fragment to be identified corresponding to the nth point cloud cluster will be identified as a flaky rock fragment; otherwise, the rock fragment to be identified corresponding to the nth point cloud cluster will be identified as a blocky rock fragment.

[0039] Secondly, this application provides a device for identifying rock fragments during tunnel excavation, comprising:

[0040] The data acquisition module is used to acquire point cloud data of the rock fragments to be identified during tunnel excavation, and to perform voxelization processing on the point cloud data to obtain multiple voxels; wherein, the point cloud data includes multiple point data, and each voxel in the multiple voxels includes at least one point data.

[0041] The similarity calculation module is used to construct a fast point feature histogram for each voxel among multiple voxels, obtain the voxel's spatial location coordinates, curvature values ​​and cross kernel of the fast point feature histogram, and obtain the similarity between the voxel and other voxels based on the spatial location coordinates, curvature values ​​and cross kernel.

[0042] The supervoxel segmentation module is used to perform supervoxel segmentation on multiple voxels based on similarity, resulting in multiple supervoxels.

[0043] The coarse segmentation module is used to perform clustering based on concavity and convexity of multiple supervoxels to obtain coarse segmentation results;

[0044] The fine segmentation module is used to calculate the local gradient information corresponding to each point data in the coarse segmentation result, and update the neighborhood radius of the point data according to the local gradient information. Based on the density-based clustering algorithm, multiple point cloud clusters are obtained.

[0045] The 3D size calculation module is used to obtain the major axis and minor axis of the 3D size of each point cloud cluster using the bounding box algorithm. The major axis represents the longest line segment in the point cloud cluster, and the minor axis represents the shortest line segment in the point cloud cluster.

[0046] The identification module is used to identify the rock fragments to be identified based on the major axis and minor axis of the three-dimensional dimensions.

[0047] Thirdly, this application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned method for identifying rock fragments during tunnel excavation.

[0048] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for identifying rock fragments during tunnel excavation.

[0049] The above-mentioned solution in this application has the following beneficial effects:

[0050] This application constructs a fast point feature histogram for each voxel, obtaining the voxel's spatial coordinates, curvature values, and the cross kernel of the fast point feature histogram. Based on the spatial coordinates, curvature values, and cross kernel, it calculates the similarity between the voxel and other voxels. It discards color information, which is relatively useless in practical applications, replacing it with curvature values. Furthermore, it uses a cross kernel that better describes spatial features instead of the normal vector angle in traditional methods, considering the local features of the point cloud, thus improving the accuracy of rock fragment identification. A density-based clustering algorithm is used to coarsely segment the supervoxels, then the local gradient information corresponding to each point data in the coarse segmentation result is calculated, and the neighborhood radius of the point data is updated based on the local gradient information. This density-based clustering algorithm yields multiple point cloud clusters, resulting in more accurate point cloud clusters, which is beneficial for improving the accuracy of subsequent rock fragment identification corresponding to the point cloud clusters.

[0051] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating a method for identifying rock fragments during tunnel excavation, as provided in an embodiment of this application;

[0054] Figure 2 A flowchart illustrating a TBM support method decision-making method based on rock fragment identification, provided in an embodiment of this application;

[0055] Figure 3 A schematic diagram of the structure of a rock fragment identification device during tunnel excavation provided in an embodiment of this application;

[0056] Figure 4 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0057] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0058] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0059] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0060] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0061] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0062] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0063] To address the low accuracy of rock fragment identification during tunnel excavation, this application provides a method, terminal device, and medium for rock fragment identification during tunnel excavation. The method constructs a fast point feature histogram for each voxel, obtaining its spatial coordinates, curvature value, and the cross-core of the fast point feature histogram. Based on these coordinates, curvature values, and the cross-core, the similarity between the voxel and other voxels is calculated. Color information, which is less useful in practical applications, is discarded and replaced with curvature values. A cross-core that better describes spatial features replaces the normal vector angle used in traditional methods, considering the local features of the point cloud, thus improving the accuracy of rock fragment identification. A density-based clustering algorithm coarsely segments the super-voxels, then calculates the local gradient information corresponding to each point data point in the coarse segmentation result, and updates the neighborhood radius of the point data point based on the local gradient information. This density-based clustering algorithm yields multiple point cloud clusters, resulting in more accurate clusters and improving the accuracy of subsequent rock fragment identification corresponding to these clusters.

[0064] like Figure 1 As shown, the method for identifying rock fragments during tunnel excavation provided in this application includes the following steps:

[0065] Step 11: Collect point cloud data of the rock fragments to be identified during tunnel excavation, and perform voxelization on the point cloud data to obtain multiple voxels.

[0066] Point cloud data includes multiple point data points, and each voxel in the multiple voxels includes at least one point data point.

[0067] The following provides an exemplary description of the point cloud data acquisition process for rock fragments in some embodiments of this application.

[0068] Step 11.1: Collect point cloud data using sensors.

[0069] The aforementioned sensors can be lidar, cameras, or other sensors. In addition, point cloud data can also be collected using other common point cloud data acquisition methods, without limiting the specific methods and processes involved.

[0070] Step 11.2, point cloud data preprocessing.

[0071] For example, the K-Dimensional Tree model is used to sequentially search the K neighboring points around each point, and the calculation formula is used to...

[0072]

[0073]

[0074]

[0075] Calculate the average Euclidean distance from each point to its K neighboring points. and standard deviation σ; where, Let x represent the average Euclidean distance from the i-th point to its K neighboring points. ij It means that y ij It means that z ij It means that (x) i ,y i ,z i ) represents the three-dimensional spatial coordinates of the i-th point.

[0076] Based on the average Euclidean distance and standard deviation of each point to its K neighboring points, Gaussian filtering is used to filter the point cloud data, thereby removing high-frequency noise points while preserving the original appearance of the data.

[0077] In the embodiments of this application, rock fragments are typically transported via belt conveyors. Therefore, it is also necessary to distinguish between the point cloud on the belt conveyor surface and the point cloud of the rock fragments, as detailed below:

[0078] The belt conveyor under no-load conditions is scanned to obtain its surface profile. The minimum value (min), lower quartile (Q1), median (median), upper quartile (Q3), and maximum value (max) of the same position measurement point are obtained from 100 scans of the belt conveyor under stable operating conditions. The median value of each position measurement point is used as the coordinate of the belt conveyor calibration profile feature point.

[0079] Based on the obtained belt conveyor calibration profile, the collected point cloud data is processed using the box gating method. The point cloud data is divided into belt conveyor profile points and rock fragment point clouds. For point cloud data that is judged as belt conveyor profile points, its position coordinates are converted into the coordinates of belt conveyor calibration profile feature points. For point cloud data that is judged as rock fragment point clouds, no changes are made.

[0080] For example, through calculation formula

[0081]

[0082] The judgment result ρ is obtained ik ; where ρ ik This represents the i-th point in the k-th scan. This represents the minimum observed value of the feature points of the belt conveyor calibration profile. This represents the maximum observed value of the feature points of the belt conveyor calibration profile.

[0083] {P'1(ρ'1,θ1),P'2(ρ'2,θ2),P'3(ρ'3,θ3)...P' m (ρ'm ,θ m )} represents the polar coordinates of the feature points of the belt conveyor calibration profile; the expressions for the maximum and minimum observed values ​​of the feature points of the belt conveyor calibration profile are as follows:

[0084]

[0085] Step 12: For each voxel among the multiple voxels, construct the fast point feature histogram corresponding to the voxel, obtain the spatial position coordinates, curvature values ​​and cross kernel of the fast point feature histogram of the voxel, and obtain the similarity between the voxel and other voxels based on the spatial position coordinates, curvature values ​​and cross kernel.

[0086] Traditional voxel cloud connectivity segmentation (VCCS) algorithms use spatial similarity, RGB similarity, and normal vectors as the basis for supervoxel segmentation. However, for actual rock fragments, the surface color information differences are not significant, and RGB information has little effect in the segmentation process. Therefore, this application constructs a fast point feature histogram for each voxel, obtains the voxel's spatial coordinates, curvature value, and cross-core of the fast point feature histogram, and calculates the similarity between the voxel and other voxels based on the spatial coordinates, curvature value, and cross-core.

[0087] Specifically, through calculation formula

[0088]

[0089]

[0090]

[0091] Obtain the similarity D i，j ; where D i，j DS represents the similarity between voxels i and j, where i, j = 1, 2, ..., M, i ≠ j, and M represents the total number of voxels. i，j Let represent the Euclidean distance between the i-th voxel and the j-th voxel in three-dimensional space. DC represents the cross kernel of the fast point feature histogram corresponding to the i-th voxel. i ω represents the curvature value of the i-th voxel. dc The weight ω represents the curvature value. ds The weights ω represent the Euclidean distance. HIK R represents the weights of the crossover kernel. seed N represents spatial resolution. lL represents the number of matches between any two adjacent data distributions in the fast point feature histogram, and L represents the total number of data distributions in the fast point feature histogram, l = 1, 2, ..., L, (x i y i , z i (x) represents the spatial coordinates of the i-th voxel. j y j , z j ) represents the spatial coordinates of the j-th voxel, and m represents the maximum curvature value among multiple voxels.

[0092] It is worth mentioning that this application discards the color information, which is relatively useless in actual working conditions, and replaces it with curvature value. It also uses a cross kernel that can better describe spatial features to replace the normal vector angle in the traditional method, taking into account the local features of the point cloud, thereby improving the accuracy of rock fragment identification.

[0093] Step 13: Perform supervoxel segmentation on multiple voxels based on similarity to obtain multiple supervoxels.

[0094] Step 14: Perform clustering based on concavity and convexity on multiple supervoxels to obtain coarse segmentation results.

[0095] Step 15: Calculate the local gradient information corresponding to each point data in the coarse segmentation result, update the neighborhood radius of the point data according to the local gradient information, and obtain multiple point cloud clusters based on the density-based clustering algorithm.

[0096] Step 16: For each point cloud cluster in the multiple point cloud clusters, use the bounding box algorithm to obtain the major axis and minor axis of the three-dimensional dimensions of the point cloud cluster respectively.

[0097] The major axis of the three-dimensional dimensions represents the longest line segment in the point cloud cluster, and the minor axis of the three-dimensional dimensions represents the shortest line segment in the point cloud cluster.

[0098] Step 17: Identify the rock fragments to be identified based on the major axis and minor axis of the three-dimensional dimensions.

[0099] The following is an exemplary description of the specific process of step 13 (segmenting multiple voxels into multiple voxels based on similarity to obtain multiple voxels).

[0100] Step 1: Select any one of the multiple voxels as the current seed voxel.

[0101] Step 2: For each voxel except the current seed voxel, if the similarity between the voxel and the current seed voxel is greater than a preset threshold, then add the voxel to the region where the current seed voxel is located; otherwise, use the voxel as the seed voxel to obtain the seed voxel set.

[0102] Step 3: After traversing all voxels except the current seed voxel, remove the current seed voxel from the seed voxel set. If the seed voxel set after removing the current seed voxel is not empty, take any seed voxel in the seed voxel set after removing the current seed voxel as the current seed voxel in Step 2 and return to execute Step 2; otherwise, based on the execution result of Step 2, obtain multiple super voxels.

[0103] The following is an illustrative description of the specific process of step 14 (clustering multiple supervoxels based on convexity to obtain coarse segmentation results).

[0104] Step 14.1, using the calculation formula

[0105]

[0106] Obtain the concavity-convexity relationship between multiple supervoxels

[0107] in, Represents the u-th supervoxel With the vth supervoxel The concave-convex relationship between them Let represent the normal vector of the u-th hypervoxel. Let v represent the normal vector of the v-th hypervoxel. Let represent the centroid vector of the u-th hypervoxel. Let β represent the centroid vector of the v-th hypervoxel. Thresh θ represents the threshold difference between the angle between the normal vector of the u-th supervoxel and the line connecting the centroids of the v-th supervoxel and the line of intersection.

[0108] Step 14.2: Based on the concavity-convexity relationship, the region growing algorithm is used to segment multiple supervoxels to obtain coarse segmentation results.

[0109] For example, after obtaining the concavity-convexity relationship between each supervoxel, under the constraint that supervoxels are only allowed to grow across convex edges, a region growth algorithm is used to segment each supervoxel to obtain a coarse segmentation result.

[0110] The following is an illustrative description of the specific process of step 15 (calculating the local gradient information corresponding to each point data in the coarse segmentation result, updating the neighborhood radius of the point data according to the local gradient information, and obtaining multiple point cloud clusters based on the density-based clustering algorithm).

[0111] Step 15.1, using the calculation formula

[0112]

[0113] Obtain local gradient information LP(P)i ).

[0114] Among them, LP(P i ) represents the data P of the i-th point. i The corresponding gradient information, T j Represents the data P of the i-th point. i Gradient values ​​in eight directions, Represents the data P of the i-th point. i The average gradient value within the neighborhood.

[0115] Step 15.2, using the calculation formula

[0116]

[0117] Obtain the radius R of the domain i .

[0118] Among them, R i Represents the data P of the i-th point. i The radius of the domain, LP max This represents the maximum value in the local gradient information corresponding to all point data.

[0119] In the embodiments of this application, if the local gradient value of a point data is large, it indicates that the geometric information of the point data changes drastically and is more likely to be an edge point. Therefore, the neighborhood radius of the point data should be reduced to limit the connection of edge points and avoid undersegmentation. If the local gradient value of a point data is small, it indicates that the geometric information of the point data is relatively flat and is more likely to be a non-edge point. Therefore, the neighborhood radius of the point data should be increased to increase the running speed of the algorithm while avoiding oversegmentation.

[0120] Step 15.3: For each point data point, segment all the point data points according to the neighborhood radius of the point data points to obtain multiple point cloud clusters.

[0121] Specifically, a point data is randomly selected. Centered on this point data, all neighboring point data are searched based on the neighborhood radius, and these neighboring point data are added to the current cluster. If the number of other point data within the neighborhood radius of this point data is less than a pre-set threshold, then this point data is considered a core object, and another unvisited point data is selected for processing. If this point data is a core object, then by continuously expanding the neighborhood, all reachable point data are added to the current cluster. The above process is repeated until no new point data can be added to the current cluster, finally resulting in multiple point cloud clusters.

[0122] The following is an exemplary description of the specific process of step 17 (identifying the rock fragments to be identified based on the major axis and minor axis of the three-dimensional dimensions).

[0123] Step 17.1, calculate R using formula n =B n / A n The recognition parameter B is obtained. n .

[0124] Among them, R n B represents the identification parameters of the nth point cloud cluster. n Let A represent the minor axis of the three-dimensional dimension of the nth point cloud cluster. n Let n represent the major axis of the three-dimensional dimension of the nth point cloud cluster, where n = 1, 2, ..., Num, and Num represents the total number of point cloud clusters.

[0125] Step 17.2, if R n If the value is less than 0.5, the rock fragment to be identified corresponding to the nth point cloud cluster will be identified as a flaky rock fragment; otherwise, the rock fragment to be identified corresponding to the nth point cloud cluster will be identified as a blocky rock fragment.

[0126] Example 1: Application of rock fragment identification method in hard rock tunnel boring machine (TBM) support method decision-making during tunnel excavation.

[0127] like Figure 2 As shown, the above application process is as follows:

[0128] Step 2.1: Acquire images of the slag.

[0129] Step 2.1.1: The image sharpness values ​​Lp and Vl are obtained by using the gray-scale gradient function and the Vollath function based on statistical methods, respectively.

[0130] Specifically, through calculation formula

[0131]

[0132] The sharpness value Lp is obtained; where T is the given threshold for edge detection, and G(x,y) represents the convolution of the Laplacian operator at pixel (x,y).

[0133] Through calculation formula

[0134] Vl=∑y∑xf(x,y)*f(x+1,y)-M*N*μ 2

[0135] The sharpness value Vl is obtained; where M and N represent the width and height of the image, respectively, and μ represents the average gray level of the entire image.

[0136] Step 2.1.2: The obtained image sharpness values ​​are normalized to obtain new sharpness values ​​L and V.

[0137] Specifically, through calculation formula

[0138]

[0139] The new resolution values ​​L and V are obtained.

[0140] Step 2.1.3: Assign weights of 0.7 and 0.3 to the sharpness values ​​L and V respectively to obtain the overall sharpness value Q, and then use a sharpness threshold of 0.4 for filtering.

[0141] The threshold of 0.4 was determined based on the average sharpness obtained from previous samples.

[0142] Specifically, the formula for calculating the overall sharpness value is as follows:

[0143] Q = 0.7L + 0.3V

[0144] The overall sharpness value Q is obtained.

[0145] Step 2.1.4: After filtering, the image undergoes preprocessing such as grayscale conversion, binary filtering for noise reduction, and adaptive histogram equalization.

[0146] Step 2.2, Groundwater condition identification.

[0147] Specifically, texture features of images are extracted using texture description operators (LBP, Local Binary Patterns) and gray-level co-occurrence matrices to form training feature vectors. Images are labeled in training samples to form target vectors, which are then trained using a support vector machine (SVM) classifier. The mature model is then encapsulated in an industrial control computer to determine the current groundwater status.

[0148] For example, in step 2.2.1, the image is processed using the LBP operator to obtain an LBP image, and texture information is extracted using the gray-level co-occurrence matrix. Five texture features, namely contrast, correlation, ASM energy (angular second moment), entropy, and inverse difference moment, as well as two non-texture features, mean and standard deviation, are selected to form a feature input vector.

[0149] Step 2.2.2: Label the groundwater features corresponding to the collected slag images, and divide them into four types of labels: dry, low water, high water, and gushing water, to form a target vector.

[0150] Step 2.2.3: Input the corresponding feature input vector and target vector into the SVM classifier, perform multi-class training through the SVM model, and encapsulate the trained model in the industrial control computer to classify groundwater into four levels in the target vector.

[0151] Step 2.3: Obtain the point cloud data of the slag sheet.

[0152] Specifically, a KD-Tree is used to sequentially search for the k neighboring points around each point, obtaining the average Euclidean distance and standard deviation of each point to its k neighbors. Then, each point in the point cloud data is traversed, and a weighted average is taken with the n data points before and after it. Finally, a Gaussian filter is applied to the point cloud data to remove high-frequency noise points, thus preserving the original appearance of the data while filtering.

[0153] Step 2.4: Identify the shape of the rock fragments.

[0154] The point cloud data is voxelized to obtain multiple voxels. Then, for each voxel, a fast point feature histogram is constructed to obtain the voxel's spatial coordinates, curvature values, and cross-core of the fast point feature histogram. Based on the spatial coordinates, curvature values, and cross-core, the similarity between the voxel and other voxels is obtained. Then, the multiple voxels are segmented into multiple supervoxels based on the similarity. Then, the multiple supervoxels are clustered based on concavity and convexity to obtain coarse segmentation results. Then, the local gradient information corresponding to each point data in the coarse segmentation results is calculated, and the neighborhood radius of the point data is updated based on the local gradient information. A density-based clustering algorithm is used to obtain multiple point cloud clusters. Then, for each point cloud cluster, the bounding box algorithm is used to obtain the three-dimensional major axis and three-dimensional minor axis of the point cloud cluster. Finally, the rock fragments to be identified are identified based on the three-dimensional major axis and three-dimensional minor axis.

[0155] Step 2.5, monitoring the flow rate of rock fragments.

[0156] Specifically, by constructing a sector-shaped grid model suitable for laser scanners, the pre-processed point cloud data is divided into multiple sector-shaped grid slices. Then, the area of ​​the rock fragment slice in each slice is calculated using the surface element integration method. The volume is calculated by simplifying the contour line connection between the cross sections of adjacent rock fragment slices. Finally, the volumetric flow rate is monitored by the summation method.

[0157] Step 2.6: Adjust the parameters of the corresponding support level according to the groundwater conditions, the shape of the identified rock fragments, and the flow rate of the rock fragments.

[0158] The following is an exemplary description of the rock fragment identification device provided in this application during tunnel excavation.

[0159] like Figure 3 As shown, the rock fragment identification device 300 during tunnel excavation includes:

[0160] The data acquisition module 301 is used to acquire point cloud data of rock fragments to be identified during tunnel excavation, and to perform voxelization processing on the point cloud data to obtain multiple voxels; wherein, the point cloud data includes multiple point data, and each voxel in the multiple voxels includes at least one point data.

[0161] The similarity calculation module 302 is used to construct a fast point feature histogram for each voxel among multiple voxels, obtain the spatial position coordinates, curvature values ​​and cross kernels of the fast point feature histograms of the voxels, and obtain the similarity between the voxel and other voxels based on the spatial position coordinates, curvature values ​​and cross kernels.

[0162] The supervoxel segmentation module 303 is used to perform supervoxel segmentation on multiple voxels based on similarity to obtain multiple supervoxels;

[0163] The coarse segmentation module 304 is used to perform clustering based on concavity and convexity of multiple supervoxels to obtain coarse segmentation results;

[0164] The fine segmentation module 305 is used to calculate the local gradient information corresponding to each point data in the coarse segmentation result, and update the neighborhood radius of the point data according to the local gradient information. Based on the density-based clustering algorithm, multiple point cloud clusters are obtained.

[0165] The three-dimensional size calculation module 306 is used to obtain the three-dimensional major axis and three-dimensional minor axis of each point cloud cluster using the bounding box algorithm. The three-dimensional major axis represents the longest line segment in the point cloud cluster, and the three-dimensional minor axis represents the shortest line segment in the point cloud cluster.

[0166] The identification module 307 is used to identify the rock fragments to be identified based on the major axis and minor axis of the three-dimensional dimensions.

[0167] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0168] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0169] like Figure 4 As shown, embodiments of this application provide a terminal device, such as... Figure 4 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 4 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 executes the computer program D102 to implement the steps in any of the above method embodiments.

[0170] Specifically, when the processor D100 executes the computer program D102, it collects point cloud data of the rock fragments to be identified during tunnel excavation, and performs voxelization on the point cloud data to obtain multiple voxels. Then, for each voxel, it constructs a fast point feature histogram corresponding to the voxel, obtains the spatial location coordinates, curvature values, and cross kernels of the fast point feature histograms of the voxels, and obtains the similarity between the voxels and other voxels based on the spatial location coordinates, curvature values, and cross kernels. Then, it performs super-voxel segmentation on the multiple voxels based on the similarity to obtain multiple super-voxels, and then performs clustering based on concavity and convexity to obtain coarse segmentation results. Then, it calculates the local gradient information corresponding to each point data in the coarse segmentation results, and updates the neighborhood radius of the point data based on the local gradient information. A density-based clustering algorithm is used to obtain multiple point cloud clusters. Then, the bounding box algorithm is used to obtain the three-dimensional major axis and three-dimensional minor axis of the point cloud clusters respectively. Finally, the rock fragments to be identified are identified based on the three-dimensional major axis and three-dimensional minor axis. Specifically, by constructing a fast point feature histogram for each voxel, the spatial coordinates, curvature values, and cross-cores of the fast point feature histogram for that voxel are obtained. Based on the spatial coordinates, curvature values, and cross-cores, the similarity between the voxel and other voxels is calculated. Color information, which is relatively useless in actual working conditions, is discarded and replaced with curvature values. The cross-core, which better describes spatial features, replaces the normal vector angle in traditional methods, taking into account the local features of the point cloud, thereby improving the accuracy of rock fragment identification. A density-based clustering algorithm is used to coarsely segment the supervoxels, and then the local gradient information corresponding to each point data in the coarse segmentation result is calculated. The neighborhood radius of the point data is updated based on the local gradient information. The density-based clustering algorithm obtains multiple point cloud clusters, which can produce more accurate point cloud clusters, which is beneficial to improving the accuracy of subsequent rock fragment identification corresponding to the point cloud clusters.

[0171] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0172] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.

[0173] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0174] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0175] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a rock fragment identification device / terminal device during tunnel excavation, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0176] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0177] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0178] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0179] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0180] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method of identifying a rock muck slice in a tunnel excavation process, characterized by, The method comprises the following steps: Collecting point cloud data of a rock debris piece to be identified in a tunnel excavation process, and voxelizing the point cloud data to obtain a plurality of voxels; wherein the point cloud data comprises a plurality of point data, and each voxel of the plurality of voxels comprises at least one point data; For each voxel of the plurality of voxels, a quick point feature histogram corresponding to the voxel is constructed to obtain a spatial position coordinate value, a curvature value and a cross kernel of the quick point feature histogram of the voxel, and a similarity between the voxel and other voxels is obtained according to the spatial position coordinate value, the curvature value and the cross kernel, comprising: According to the similarity, the plurality of voxels are super voxel segmented to obtain a plurality of super voxels; ； ； ； The similarity was obtained ;in, Indicates the first Individual elements and the first Similarity between individual elements , , Indicates the total number of voxels. Indicates the first Individual elements and the first Euclidean distance of individual units in three-dimensional space Indicates the first Cross kernel of the fast point feature histogram corresponding to individual elements. Indicates the first The curvature value of the individual element, The weights representing the curvature values, The weights represent the Euclidean distances. This represents the weight of the crossover kernel. Indicates spatial resolution. This represents the number of matches between each pair of adjacent data distributions in the fast point feature histogram. This represents the total number of data rows in the fast point feature histogram. , Indicates the first Spatial coordinates of individual elements Indicates the first Spatial coordinates of individual elements This represents the maximum curvature value among the plurality of voxels; The plurality of super voxels are clustered based on concave-convex relationship to obtain a coarse segmentation result; Local gradient information corresponding to each point data in the coarse segmentation result is calculated, and the domain radius of the point data is updated according to the local gradient information, and a plurality of point cloud clusters are obtained based on a density-based clustering algorithm; For each point cloud cluster of the plurality of point cloud clusters, a bounding box algorithm is used to obtain a three-dimensional size long axis and a three-dimensional size short axis of the point cloud cluster respectively; the three-dimensional size long axis represents the longest line segment in the point cloud cluster, and the three-dimensional size short axis represents the shortest line segment in the point cloud cluster; The rock debris piece to be identified is identified according to the three-dimensional size long axis and the three-dimensional size short axis. The method according to the similarity, the plurality of voxels are super voxel segmented to obtain a plurality of super voxels, comprising:

2. The identification method according to claim 1, characterized in that, Step one, any voxel of the plurality of voxels is taken as a current seed voxel; Step two, for each voxel except the current seed voxel, if the similarity between the voxel and the current seed voxel is greater than a preset threshold, the voxel is added to a region where the current seed voxel is located; otherwise, the voxel is taken as a seed voxel to obtain a seed voxel set; Step three, after all voxels except the current seed voxel are traversed, the current seed voxel is removed from the seed voxel set, if the seed voxel set after removing the current seed voxel is not empty, any seed voxel in the seed voxel set after removing the current seed voxel is taken as the current seed voxel in step two, and step two is executed; otherwise, the plurality of super voxels are obtained according to the execution result of step two. The method of clustering the plurality of super voxels based on concave-convex relationship to obtain a coarse segmentation result, comprising:

3. The identification method according to claim 2, characterized in that, According to the concave-convex relationship, the plurality of super voxels are segmented by using a region growing algorithm to obtain the coarse segmentation result. The method of calculating local gradient information corresponding to each point data in the coarse segmentation result, and updating the domain radius of the point data according to the local gradient information, and obtaining a plurality of point cloud clusters based on a density-based clustering algorithm, comprising: ； ； ； obtaining concave-convex relationships between the plurality of super voxels ; wherein, represents a concave-convex relationship between the th super voxel and the th super voxel , represents a normal vector of the th super voxel represents a normal vector of the th super voxel represents a gravity vector of the th super voxel represents a gravity vector of the th super voxel represents a difference threshold of an angle between the normal vector and the gravity connecting line of the th super voxel and the th super voxel represents an angle between the connecting line and the intersection line between the gravity of the th super voxel and the th super voxel According to the concave-convex relationship, the plurality of super voxels are segmented by using a region growing algorithm to obtain the coarse segmentation result.

4. The identification method according to claim 3, characterized in that, The method of calculating local gradient information corresponding to each point data in the coarse segmentation result, and updating the domain radius of the point data according to the local gradient information, and obtaining a plurality of point cloud clusters based on a density-based clustering algorithm, comprising: According to the concave-convex relationship, the plurality of super voxels are segmented by using a region growing algorithm to obtain the coarse segmentation result. ； obtaining the local gradient information ; wherein the gradient information of the first point data corresponding to the first point data the gradient value of the eight directions the gradient value of the eight directions point data the average value of the gradient values in the neighborhood​ According to the concave-convex relationship, the plurality of super voxels are segmented by using a region growing algorithm to obtain the coarse segmentation result. ； Obtain the radius of the domain ;in, Indicates the first Data points The radius of the domain, This represents the maximum value in the local gradient information corresponding to all point data; ​ 5. The identification method according to claim 4, characterized in that, The identifying the rock debris piece to be identified according to the long axis of the three-dimensional size and the short axis of the three-dimensional size comprises: Through calculation formula To obtain the recognition parameters ;in, Indicates the first The identification parameters of a point cloud cluster Indicates the first The minor axis of the three-dimensional dimensions of a point cloud cluster, Indicates the first The major axis of the three-dimensional dimensions of a point cloud cluster, , This indicates the total number of point cloud clusters; If , the to-be-identified rock debris piece corresponding to the first point cloud cluster is identified as a flaky rock debris piece; otherwise, the to-be-identified rock debris piece corresponding to the first point cloud cluster is identified as a massive rock debris piece.

6. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the method for identifying the rock debris piece in the tunnel excavation process according to any one of claims 1 to 5.

7. A computer-readable storage medium storing a computer program, wherein the computer program comprises the following steps of: receiving a request for a resource from a client; determining whether the client is authorized to access the resource; and if the client is authorized to access the resource, providing the resource to the client. The computer program is executed by the processor to realize the method for identifying the rock debris piece in the tunnel excavation process according to any one of claims 1 to 5.

Images