Undercut detection method and device for high-speed rail tunnel construction and computer program product

Abstract

The invention belongs to the technical field of tunnel construction detection, and discloses a high-speed rail tunnel construction undercut detection method and device and a computer program product in order to solve the problems of large workload and low data acquisition efficiency in the prior art. The laser can be directly scanned by a laser scanner without point-by-point measurement, so that the workload is reduced, and the detection efficiency is improved; meanwhile, the geometric state of the tunnel can be reflected by the three-dimensional point cloud data, and an accurate construction excavation section can be constructed during detection, so that the detection accuracy can be ensured; on the basis, the invention provides a tunnel under-excavation detection mode which is high in efficiency, low in workload and high in precision, so that the tunnel under-excavation detection method is very suitable for large-scale application and popularization.

Description

Technical Field

[0001] The present invention belongs to the technical field of tunnel construction detection, and in particular relates to a method, device and computer program product for detecting under-break in high-speed railway tunnel construction. Background Art

[0002] In tunnel drilling and blasting construction, over-excavation and under-excavation of tunnels are a great safety hazard, which seriously affects the quality of tunnel excavation and subsequent construction. Among them, tunnel over-excavation and under-excavation are determined based on the designed excavation contour line as the baseline, that is, the part of the section obtained by actual excavation outside the baseline is over-excavation, and the part inside the baseline is under-excavation. Over-excavation will cause waste of materials in advance support during the later mining process, and the support effect is not good, posing a safety hazard. The consequence of under-excavation is that it may cause greater troubles in transportation and support, and sometimes even requires blasting again for correction, which also wastes materials and manpower. Therefore, reducing the occurrence of over-excavation and under-excavation, timely discovering over-excavation and under-excavation and taking corresponding compensatory measures are of great engineering significance.

[0003] At present, tunnel under-break detection mainly uses equipment such as profilers and total stations to measure the excavation section point by point and surface by surface. Although this traditional method can meet the requirements of under-break detection, it adopts a "point-by-point" measurement method (i.e. manually operating the instrument to measure point by point or line by line), which not only has a large workload, but also leads to low data acquisition efficiency, thus affecting the construction progress. Therefore, based on the aforementioned shortcomings, how to provide a tunnel under-break detection method with high efficiency and low workload has become a problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to provide a method, device and computer program product for detecting under-break in high-speed railway tunnel construction, so as to solve the problems of large workload and low data acquisition efficiency in the prior art.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, a method for detecting under-break in high-speed railway tunnel construction is provided, comprising: Acquire a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel; Based on the three-dimensional point cloud data set, a three-dimensional surface model of an actual excavation profile of the target construction tunnel is constructed, wherein the three-dimensional surface model of the actual excavation profile and the three-dimensional surface model of the theoretical excavation profile are both triangular mesh models; For any vertex in the actual excavation contour three-dimensional surface model, determine the triangle patch closest to the any vertex from the theoretical excavation contour three-dimensional surface model as the underbreak detection reference plane corresponding to the any vertex, and use the distance from the any vertex to the underbreak detection reference plane as the construction error distance; Determine whether the construction error distance is 0; If not, the sign of the construction error distance is determined according to the vector of the normal line from any vertex to the underbreak detection reference plane; If the sign of the construction error distance is a negative sign, any of the vertices is determined to be an under-excavation point, and after all vertices in the actual excavation contour three-dimensional surface model are polled, the under-excavation detection result of the target construction tunnel is obtained.

[0006] Based on the above disclosed content, the present invention first obtains a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation contour of the target construction tunnel; then, based on the aforementioned three-dimensional point cloud data set, a three-dimensional surface model of an actual excavation contour of the target construction tunnel is constructed, and the aforementioned two models are both triangular mesh models; based on this, the tunnel under-excavation detection can be performed according to the actual excavation contour three-dimensional surface model and the theoretical excavation contour three-dimensional surface model, that is, the theoretical excavation contour three-dimensional surface model is used as a reference model, and the actual excavation contour three-dimensional surface model is used as a comparison model, and then, the under-excavation detection is converted into calculating the distance between the comparison model and the reference model, that is, by determining in the reference model the triangular facet closest to each vertex in the comparison model, so as to use the nearest triangular facet as the under-excavation detection reference plane (i.e., reference plane) corresponding to each vertex; then, by determining the positive or negative distance between each vertex and the corresponding reference plane, it is determined whether the position of each vertex in the comparison model is an under-excavation point.

[0007] Specifically, the present invention determines the sign of the distance from each vertex to the corresponding reference plane with the help of the vector between the normals from each vertex to the corresponding reference plane, wherein, if the distance from any vertex to the corresponding reference plane is negative, it means that the actual excavation point corresponding to any vertex has not reached the position of the theoretical excavation point, and the position of any vertex is an under-excavation position; conversely, if the distance is positive, it means that it is an over-excavation position, and if the distance is 0, it means that the excavation position of any vertex is on the reference plane; thus, through the aforementioned method, the under-excavation detection result of the target construction tunnel can be quickly obtained.

[0008] Through the above design, the present invention constructs a three-dimensional surface model of the actual excavation contour of the tunnel by collecting a three-dimensional point cloud data set of the target construction tunnel, and then, from the three-dimensional surface model of the theoretical excavation contour of the tunnel, determines the triangular facets closest to each vertex in the actual model, and uses them as the under-excavation detection reference planes of each vertex; finally, by calculating the positive and negative distances from the vertices to the corresponding reference planes, it is determined whether each vertex in the actual model is an under-excavation point, thereby obtaining the final under-excavation detection result; thus, the three-dimensional point cloud data is used for tunnel under-excavation detection, and its acquisition speed is fast, and it can be directly scanned by a laser scanner without point-by-point measurement, therefore, not only reducing the workload, but also improving the detection efficiency; at the same time, the three-dimensional point cloud data can reflect the geometric state of the tunnel, and can construct an accurate construction excavation section during detection, thus, the accuracy of the detection can also be guaranteed; based on this, the present invention provides a tunnel under-excavation detection method with high efficiency, low workload and high precision, which is very suitable for large-scale application and promotion.

[0009] In a possible design, based on the three-dimensional point cloud data set, a three-dimensional surface model of the actual excavation profile of the target construction tunnel is constructed, including: Extracting a plurality of feature point clouds from the three-dimensional point cloud data set to form a feature point cloud set; Calculating the local density of each feature point cloud in the feature point cloud set; Using the feature point cloud set, constructing a three-dimensional surface model of the initial excavation profile of the target construction tunnel; According to the local density of each feature point cloud in the initial excavation contour three-dimensional surface model, the initial excavation contour three-dimensional surface model is subjected to triangular patch optimization processing, so as to obtain the actual excavation contour three-dimensional surface model after the triangular patch optimization processing.

[0010] In a possible design, multiple feature point clouds are extracted from the three-dimensional point cloud data set to form a feature point cloud set, including: Initialize the sampling iteration number s, and obtain the side length of the voxel grid at the s-th sampling, wherein when s is 1, the side length of the voxel grid at the s-th sampling is the initial side length; According to the side length of the voxel grid at the s-th sampling, the three-dimensional point cloud data set is divided into a plurality of voxel grids, wherein any voxel grid contains a plurality of three-dimensional point clouds; Extract the feature point cloud at the sth sampling time from multiple three-dimensional point clouds in each voxel grid; Determine whether a sampling iteration stop condition is met, wherein the sampling iteration stop condition is that the number of feature point clouds at the s-th sampling is equal to a preset number; If not, updating the side length of the voxel grid at the s-th sampling according to the number of feature point clouds at the s-th sampling and the preset number to obtain an updated side length; Increasing s by 1, replacing the side length of the voxel grid at the s-th sampling with the updated side length, and re-dividing the three-dimensional point cloud data set into a plurality of voxel grids according to the side length of the voxel grid at the s-th sampling until the sampling iteration stop condition is satisfied, so as to form an initial feature point cloud set using the feature point cloud sampled when the sampling iteration stop condition is satisfied; The initial feature point cloud set is subjected to denoising processing to obtain the feature point cloud set after the denoising processing.

[0011] In a possible design, a feature point cloud at the sth sampling time is extracted from multiple three-dimensional point clouds in each voxel grid, including: For any three-dimensional point cloud in any voxel grid, determine a neighboring point cloud of the any three-dimensional point cloud from the any voxel grid and a target grid, wherein the target grid is a voxel grid adjacent to the any voxel grid among the plurality of voxel grids; Determine a first normal vector of any three-dimensional point cloud and a second normal vector of each neighboring point cloud; Calculating the angles between the first normal vector and each second normal vector to obtain a plurality of normal vector angles; Find the average value of multiple normal vector angles; Determine whether the average value is greater than or equal to an angle threshold; If so, any one of the three-dimensional point clouds is taken as a feature point cloud in any one of the voxel grids, and after the three-dimensional point clouds in all the voxel grids are polled, the feature point cloud at the s-th sampling is obtained.

[0012] In a possible design, updating the side length of the voxel grid at the s-th sampling according to the number of feature point clouds at the s-th sampling and the preset number to obtain the updated side length includes: Calculating a voxel grid side length adjustment coefficient at the s-th sampling according to the number of feature point clouds at the s-th sampling and the preset number; The updated side length is calculated based on the voxel grid side length adjustment coefficient and the initial side length at the s-th sampling.

[0013] In a possible design, the feature point cloud set is used to construct a three-dimensional surface model of the initial excavation profile of the target construction tunnel, including: Randomly select three feature point clouds from the feature point cloud set to form an initial triangle; Taking any edge of the initial triangle as a growing edge, and selecting the optimal extended point cloud corresponding to the growing edge from the feature point cloud set; Using the optimal extended point cloud and the growing edge, construct an extended triangle; The initial triangle is updated to the extended triangle, and any edge in the initial triangle is used as the growing edge again, and the optimal extended point cloud corresponding to the growing edge is screened out from the feature point cloud set until all feature point clouds in the feature point cloud set are divided, so as to construct the initial excavation contour three-dimensional surface model using the initial triangle and the obtained extended triangle.

[0014] In a possible design, selecting the optimal extended point cloud corresponding to the growing edge from the feature point cloud set includes: Selecting a feature point cloud closest to the growing edge from the feature point cloud set as an initial extended point cloud; Constructing an initial extended triangle based on the initial extended point cloud and the growing edge; Determine whether there is a specified feature point cloud within the circumscribed circle of the initial extended triangle, wherein the specified feature point cloud is all feature point clouds in the feature point cloud set except for the feature point clouds corresponding to the three vertices of the initial extended triangle; If so, one is randomly selected from the specified feature point clouds as the initial extended point cloud, and the initial extended triangle is reconstructed based on the initial extended point cloud and the growth edge until there is no specified feature point cloud within the circumscribed circle of the constructed initial extended triangle, and the initial extended point cloud corresponding to the initial extended triangle in which there is no specified feature point cloud within the circumscribed circle is used as the optimal extended point cloud.

[0015] In a possible design, calculating the local density of each feature point cloud in the feature point cloud set includes: For any feature point cloud in the feature point cloud set, obtaining a k-neighborhood point cloud of the feature point cloud; Calculate the distance between any feature point cloud and each k-neighborhood point cloud, and use the average of all distances as the local density of any feature point cloud; Correspondingly, according to the local density of each feature point cloud in the initial excavation contour three-dimensional surface model, the initial excavation contour three-dimensional surface model is subjected to triangular facet optimization processing, which includes: For any initial triangular facet in the initial excavation contour three-dimensional surface model, calculating the circumscribed circle radius of any initial triangular facet; Calculate a density threshold corresponding to any initial triangular facet according to the local density of the feature point cloud corresponding to the three endpoints of any initial triangular facet; According to the circumscribed circle radius and the density threshold, determining whether any of the initial triangular facets is a segmented triangular facet; If so, a new feature point cloud is inserted into any of the initial triangular facets so as to segment any of the initial triangular facets based on the new feature point cloud, and when all the initial triangular facets in the initial excavation contour three-dimensional surface model are polled, the actual excavation contour three-dimensional surface model is obtained.

[0016] In a second aspect, a high-speed railway tunnel construction under-excavation detection device is provided, comprising: An acquisition unit, used for acquiring a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel; A model building unit, configured to build a three-dimensional surface model of an actual excavation profile of the target construction tunnel based on the three-dimensional point cloud data set, wherein both the three-dimensional surface model of the actual excavation profile and the three-dimensional surface model of the theoretical excavation profile are triangular mesh models; The error detection unit is used to determine, for any vertex in the actual excavation contour three-dimensional surface model, a triangular facet closest to the any vertex from the theoretical excavation contour three-dimensional surface model as an underbreak detection reference plane corresponding to the any vertex, and to use the distance between the any vertex and the underbreak detection reference plane as a construction error distance; An underbreak detection unit, used to determine whether the construction error distance is 0; The underbreak detection unit is used to determine the sign of the construction error distance according to the vector of the normal line from any vertex to the underbreak detection reference plane when it is determined that the construction error distance is not 0; If the sign of the construction error distance is negative, the under-excavation detection unit is further used to determine that any of the vertices is an under-excavation point, and after polling all vertices in the actual excavation contour three-dimensional surface model, obtain the under-excavation detection result of the target construction tunnel.

[0017] According to a third aspect, another device for detecting under-excavation in high-speed railway tunnel construction is provided. Taking the device as an electronic device as an example, the device includes a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the method for detecting under-excavation in high-speed railway tunnel construction as described in the first aspect or any possible design of the first aspect.

[0018] In a fourth aspect, a storage medium is provided, on which instructions are stored. When the instructions are executed on a computer, the high-speed railway tunnel construction under-excavation detection method as described in the first aspect or any possible design of the first aspect is executed.

[0019] In a fifth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, causes the computer to execute the high-speed railway tunnel construction under-excavation detection method as described in the first aspect or any possible design of the first aspect.

[0020] Beneficial effects: (1) The present invention uses three-dimensional point cloud data to detect tunnel under-break. The data can be acquired quickly and can be directly scanned by a laser scanner without the need for point-by-point measurement. Therefore, the workload is reduced and the detection efficiency is improved. At the same time, the three-dimensional point cloud data can reflect the geometric state of the tunnel and can construct an accurate construction excavation section during detection. In this way, the accuracy of the detection can be guaranteed. Based on this, the present invention provides a tunnel under-break detection method with high efficiency, low workload and high precision, which is very suitable for large-scale application and promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of the steps of a method for detecting underbreak in high-speed railway tunnel construction provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a high-speed railway tunnel construction underbreak detection device provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0023] It should be understood that although the terms first, second, etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another unit. For example, a first unit can be referred to as a second unit, and similarly, a second unit can be referred to as a first unit without departing from the scope of the exemplary embodiments of the present invention.

[0024] It should be understood that the term "and / or" that may appear in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" that may appear in this article describes another type of association object relationship, indicating that two relationships may exist. For example, A / and B can represent two situations: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0025] Example: See also Figure 1 As shown, the high-speed railway tunnel construction under-break detection method provided in this embodiment collects the three-dimensional point cloud data set of the target construction tunnel to construct a three-dimensional surface model of the actual excavation contour of the tunnel, and then, from the three-dimensional surface model of the theoretical excavation contour of the tunnel, the triangular facet closest to each vertex in the actual model is determined, and it is used as the under-break detection reference plane of each vertex; finally, by calculating the positive and negative distances from the vertex to the corresponding reference plane, it is determined whether each vertex in the actual model is an under-break point, thereby obtaining the final under-break detection result; thus, compared with the traditional technology, the three-dimensional point cloud data can be directly scanned by a laser scanner without the need for point-by-point measurement, thereby not only improving the efficiency but also reducing the workload of measurement; at the same time, a large amount of three-dimensional point cloud data can reflect the actual geometric shape of the tunnel, so that an accurate construction excavation section can be constructed through it, thereby ensuring the accuracy of the detection; therefore, this method is very suitable for large-scale application and promotion; wherein, for example, this method can be but not limited to running on the under-break detection end side, and optionally, the under-break detection end can be but not limited to a personal computer (personal It can be understood that the aforementioned execution subject does not constitute a limitation on the embodiments of the present application. Accordingly, the operation steps of the method can be but are not limited to the following steps S1 to S6.

[0026] S1. Obtain a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel; in this embodiment, the aforementioned three-dimensional point cloud data set is obtained by scanning the target construction tunnel with a laser scanner. Therefore, compared with traditional total stations, profilers and other equipment, this embodiment can obtain a large amount of point cloud data of the target construction tunnel in one scan without the need for repeated measurements at multiple cross-sectional positions of the tunnel, thereby improving the efficiency of detection; further, the example can be but not limited to using the construction design drawings of the target construction tunnel and using software such as 3D Max or Civil 3D to construct a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel; then, it is stored in the under-excavation detection terminal, so as to facilitate the subsequent under-excavation detection of the target construction tunnel in combination with the actual excavation profile three-dimensional surface model of the target construction tunnel.

[0027] Optionally, this embodiment first constructs an actual excavation section model of the tunnel (i.e., the actual excavation contour three-dimensional surface model described below) based on the collected three-dimensional point cloud data set; then, by comparing it with the theoretical excavation contour three-dimensional surface model, the under-excavation detection result of the tunnel is obtained; wherein, the construction process of the actual excavation section model is shown in the following step S2.

[0028] S2. Based on the three-dimensional point cloud data set, a three-dimensional surface model of the actual excavation contour of the target construction tunnel is constructed, wherein the three-dimensional surface model of the actual excavation contour and the three-dimensional surface model of the theoretical excavation contour are both triangular mesh models; in specific implementation, due to the huge number of collected point cloud data sets, in order to reduce the amount of data to be processed and at the same time ensure the accuracy of model construction, this embodiment first extracts the feature point cloud, and then, based on the feature point cloud, constructs the model; at the same time, due to the influence of factors such as equipment, distance, scanning angle, and environment during collection, uneven point cloud density may occur. Therefore, after constructing the model, this embodiment combines the local density of the feature point cloud to optimize the model, thereby reducing the adverse effects caused by uneven point cloud density (such as reducing holes and gaps, improving the smoothness and accuracy of the surface, etc.).

[0029] Optionally, the aforementioned model building process may be but is not limited to the following steps S21 to S24.

[0030] S21. Extract multiple feature point clouds from the three-dimensional point cloud data set to form a feature point cloud set; in this embodiment, in order to reduce the number of point clouds while ensuring the accuracy of constructing the subsequent excavation section model, this embodiment samples the three-dimensional point cloud data set by screening out feature point clouds; optionally, this embodiment provides an adaptive voxel sampling method to screen out feature point clouds from the three-dimensional point cloud set, wherein the specific implementation process can be but is not limited to the following steps S21a to S21g.

[0031] S21a. Initialize the number of sampling iterations s, and obtain the side length of the voxel grid at the s-th sampling, wherein when s is 1, the side length of the s-th voxel grid is the initial side length; in a specific implementation, the initial side length can be set in advance; and after obtaining the side length of the voxel grid at the s-th sampling, each three-dimensional point cloud in the three-dimensional point cloud data set can be divided into different voxel grids; wherein the division process can be but is not limited to as shown in the following step S21b.

[0032] S21b. Divide the three-dimensional point cloud data set into a plurality of voxel grids according to the side length of the voxel grid at the s-th sampling, wherein any voxel grid contains a plurality of three-dimensional point clouds; in specific applications, for example, but not limited to, first obtaining the maximum horizontal coordinate, maximum vertical coordinate, maximum z-axis coordinate, minimum horizontal coordinate, minimum vertical coordinate and minimum z-axis coordinate in the three-dimensional point cloud data set, and then, according to the following formula (3), determining the number of voxel grids divided at the s-th sampling.

[0033] (3) In the above formula (3), Both represent parameters used to calculate the number of voxel grids, In turn, they represent the maximum horizontal coordinate, the maximum vertical coordinate, and the maximum z-axis coordinate. In turn, they represent the minimum horizontal coordinate, the minimum vertical coordinate, and the minimum z-axis coordinate; represents the side length of the voxel grid at the sth sampling time, Represents the ceiling function.

[0034] Therefore, based on the above formula (3), after calculating the parameters for calculating the number of voxel grids, The product of is taken as the number of voxel grids at the sth sampling; after calculating the number of voxel grids, each 3D point cloud in the 3D point cloud data set can be divided into different voxel grids; of course, the division process of 3D point cloud data is a common technology for voxel grid downsampling of point cloud data, and its principle will not be repeated here.

[0035] After completing the division of the voxel grid at the sth sampling, the three-dimensional point cloud in each voxel grid can be sampled to obtain the feature point cloud in each voxel grid; wherein the feature point cloud selection process is shown in the following step S21c.

[0036] S21c. Extract the feature point cloud at the sth sampling from the multiple three-dimensional point clouds in each voxel grid; in the present embodiment, the neighborhood point clouds of each three-dimensional point cloud in the current voxel grid are screened out from the current voxel grid and the voxel grids adjacent to the current voxel grid, and then, by calculating the angle between the normal vector of each three-dimensional point cloud and the corresponding neighborhood point cloud, it is determined whether each three-dimensional point cloud is a feature point cloud, and then the feature point cloud is extracted.

[0037] Optionally, any three-dimensional point cloud in any voxel grid is used as an example for explanation below, and the process may be but is not limited to the following steps S21c1 to S21c6.

[0038] S21c1. For any three-dimensional point cloud in any voxel grid, determine the neighborhood point cloud of any three-dimensional point cloud from the any voxel grid and the target grid, wherein the target grid is a voxel grid adjacent to the any voxel grid among the several voxel grids; in specific applications, for example but not limited to, a k-neighbor search can be used to search for several neighborhood point clouds corresponding to the any three-dimensional point cloud in the any voxel grid and the target grid.

[0039] After searching for several domain point clouds, the normal vectors of any three-dimensional point cloud and its corresponding neighboring point clouds can be calculated, and the process is shown in the following step S21c2.

[0040] S21c2. Determine the first normal vector of any three-dimensional point cloud and the second normal vector of each neighborhood point cloud; in a specific implementation, the neighborhood point cloud corresponding to any three-dimensional point cloud is used to construct the covariance matrix of any three-dimensional point cloud, and then the normal vector of any three-dimensional point cloud is obtained by solving the eigenvectors in the covariance matrix.

[0041] In this embodiment, the covariance matrix of any three-dimensional point cloud may be constructed by, for example but not limited to, using the following formula (4).

[0042] (4) In the above formula (4), is the covariance matrix of any three-dimensional point cloud, represents the centroid coordinates of all neighboring point clouds corresponding to any three-dimensional point cloud, represents the kth neighborhood point cloud in the neighborhood point cloud of any three-dimensional point cloud, represents the total number of neighborhood point clouds of any three-dimensional point cloud, represents the transpose operation; and the centroid coordinates are the average coordinates of all neighborhood point clouds, that is, the average coordinates.

[0043] After constructing the covariance matrix of any three-dimensional point cloud based on the above formula (4), the eigenvector in the covariance matrix can be solved; then the smallest eigenvector is used as the normal vector of any three-dimensional point cloud; similarly, the process of obtaining the normal vectors of each neighborhood point cloud of any three-dimensional point cloud is the same, which will not be repeated here.

[0044] After obtaining the normal vectors of any three-dimensional point cloud and its corresponding neighborhood point clouds, the angle between the normal vectors can be calculated, and the process is shown in the following step S21c3.

[0045] S21c3. Calculate the angle between the first normal vector and each second normal vector to obtain multiple normal vector angles; in this embodiment, after calculating the angle between the normal vector of any of the aforementioned three-dimensional point clouds and the normal vectors of its corresponding neighboring point clouds, the mean of the angles can be calculated so that the mean can be used to determine whether any of the three-dimensional point clouds is a feature point cloud; the judgment process is shown in the following steps S21c4 to S21c6.

[0046] S21c4. Find the average value of the angles between multiple normal vectors.

[0047] S21c5. Determine whether the average value is greater than or equal to the angle threshold. In this embodiment, the larger the average value of the angle between the normal vectors, the greater the change in the normal vectors within the voxel grid, and the more obvious the change in the feature of the point cloud. Therefore, when the average value of the normal vector angle is greater than or equal to the angle threshold, it can be determined that any three-dimensional point cloud is a feature point cloud in any voxel grid, and the process is shown in the following step S21c6.

[0048] S21c6. If yes, then any one of the three-dimensional point clouds is used as a feature point cloud in any one of the voxel grids, and after the three-dimensional point clouds in all the voxel grids are polled, the feature point cloud at the s-th sampling is obtained.

[0049] In this embodiment, when all three-dimensional point clouds in any voxel grid are polled in the aforementioned manner, the feature point cloud in any voxel grid can be extracted, and then, by traversing the remaining voxel grids, the feature point cloud at the sth sampling time can be obtained; finally, it can be determined whether the number of feature point clouds obtained by sampling reaches a preset number, so as to determine whether it is necessary to adjust the side length of the voxel grid, so as to ensure the accuracy of subsequent model construction while satisfying the reduction of the number of point clouds.

[0050] The iteration stop judgment condition is shown in the following step S21d. S21d. Determine whether the sampling iteration stopping condition is met, wherein the sampling iteration stopping condition is that the number of feature point clouds at the s-th sampling is equal to a preset number; in this embodiment, the preset number can be specifically set according to actual use and is not specifically limited here; wherein, if the number of feature point clouds obtained by the s-th sampling is not equal to the preset number, it is necessary to adjust the side length of the voxel grid in order to adjust the coarse granularity of the sampling; wherein the side length adjustment process is shown in the following step S21e.

[0051] S21e. If not, then based on the number of feature point clouds at the s-th sampling and the preset number, the side length of the voxel grid at the s-th sampling is updated to obtain the updated side length; in specific applications, for example, but not limited to, first calculating the voxel grid side length adjustment coefficient at the s-th sampling based on the number of feature point clouds at the s-th sampling and the preset number; then, based on the voxel grid side length adjustment coefficient at the s-th sampling and the initial side length, the updated side length is calculated.

[0052] Furthermore, for example, but not limited to, the following formula (1) may be used to calculate the aforementioned voxel grid side length adjustment coefficient.

[0053] (1) In the above formula (1), represents the voxel grid side length adjustment coefficient at the s-th sampling, Indicates the number of feature point clouds at the sth sampling time, Indicates the preset quantity.

[0054] After calculating the voxel grid side length adjustment coefficient at the s-th sampling, the updated side length can be calculated. In this embodiment, the following formula (2) is used for calculation.

[0055] (2) In the above formula (2), represents the updated edge length, represents the initial side length, Represents the control factor; in this embodiment, the control factor is a constant and can be specifically set according to actual use and is not specifically limited here.

[0056] After completing the update of the side length of the voxel grid, the number of voxel grids can be recalculated based on this, and the three-dimensional point cloud can be re-divided, that is, the aforementioned steps S21b to S21e are re-executed until the conditions for stopping the iteration are met, and a feature point cloud set can be obtained; wherein, the iterative sampling process is shown in the following step S21f.

[0057] S21f. Add 1 to s, replace the side length of the voxel grid at the sth sampling with the updated side length, and re-divide the three-dimensional point cloud data set into a number of voxel grids according to the side length of the voxel grid at the sth sampling until the sampling iteration stop condition is met, so as to use the feature point cloud sampled when the sampling iteration stop condition is met to form an initial feature point cloud set.

[0058] In this way, through the aforementioned steps S21b to S21f, this embodiment adjusts the side length of the voxel grid by determining whether the number of feature point clouds obtained by sampling reaches a preset number, and resamples accordingly; and the coarse-grained result of the sampling is affected by the side length of the grid. The larger the side length, the more blurred the geometric information of the grid, and the smaller the side length, the more detailed the local geometric information of the grid. Therefore, this embodiment can balance the retention and simplification of features through adaptive side length adjustment, thereby avoiding insufficient features caused by over-simplification, which causes the problem of low accuracy of subsequent models.

[0059] Therefore, after the initial feature point cloud is extracted, in order to further improve the accuracy of the feature point cloud, this embodiment also performs denoising processing on it, and the process is shown in the following step S21g.

[0060] S21g. The initial feature point cloud set is subjected to denoising processing to obtain the feature point cloud set after the denoising processing; in the present embodiment, the feature point cloud set is first used to construct a kd-tree, and then the centroid of the feature point cloud set is calculated (the horizontal coordinate, vertical coordinate and z-axis coordinate of the centroid are the horizontal coordinate mean, vertical coordinate mean and z-axis coordinate mean of all initial feature point clouds in the initial feature point cloud set). Then, the average distance from each initial feature point cloud to the centroid is calculated; then, the standard deviation of the distance from each initial feature point cloud to the centroid is calculated, and based on the standard deviation and the average distance, the search radius is calculated (R=d+α×σ, α is the radius threshold parameter, σ is the standard deviation, and d is the average distance); finally, based on the search radius and the distance between each initial feature point cloud and the centroid in the kd-tree, the kd-tree is searched to obtain the feature point cloud set.

[0061] Specifically, the calculated point cloud centroid is used as the reference point and the search radius R as the search range, which are input into the kd-tree. The kd-tree will retrieve all the initial feature point cloud data within the search radius, and use them as feature point clouds to form a feature point cloud set. Of course, using kd-tree for point cloud search is a common technology of kd-tee filtering, and its specific search process will not be repeated here.

[0062] Therefore, through the aforementioned steps S21a~S21g, while simplifying the point cloud data, the feature point cloud that can retain the characteristics of the three-dimensional point cloud data set can be extracted, and then, based on this, the excavation section of the target construction tunnel can be reconstructed. The process is shown in the following steps S22~S24.

[0063] S22. Calculate the local density of each feature point cloud in the feature point cloud set; in the specific implementation, it has been explained above that the point cloud acquisition may be affected by factors such as equipment, distance, scanning angle, environment, etc., which may cause uneven point cloud density. Although the point cloud simplification can reduce the problem of uneven point cloud density to a certain extent, the feature point cloud set as a whole still has the problem of uneven density. Therefore, in order to avoid the adverse effects of uneven point cloud density on subsequent model construction (such as causing holes and gaps, missing cross-sectional details, etc.), this embodiment calculates the local density of the feature point cloud, thereby optimizing the model after the model is built.

[0064] Optionally, taking any feature point cloud as an example, the calculation process of its local density is explained, that is: for any feature point cloud in the feature point cloud set, obtain the k-neighborhood point cloud of the any feature point cloud; the distance between the any feature point cloud and each k-neighborhood point cloud can be calculated first; then, the average of all distances is used as the local density of the any feature point cloud.

[0065] After the local density of each feature point cloud is calculated, the surface reconstruction of the excavation section can be performed, and the process is shown in the following step S23.

[0066] S23. Using the feature point cloud set, construct a three-dimensional surface model of the initial excavation contour of the target construction tunnel; in specific implementation, this embodiment uses an improved Delaunay triangulation algorithm to reconstruct the three-dimensional surface of the excavation section of the target construction tunnel, and the process can be but is not limited to the following steps S23a to S23d.

[0067] S23a. Randomly select three feature point clouds from the feature point cloud set to form an initial triangle; in this embodiment, after randomly selecting three feature point clouds to construct the initial triangle, the growth edge can be determined from the initial triangle, and the optimal extension point cloud corresponding to the growth edge can be determined, so that the extended triangle can be subsequently constructed based on the optimal extension point cloud, thereby adding the extended triangle to the triangle network corresponding to the initial triangle, and then completing the expansion of the triangle network; wherein, the process of determining the optimal extension point cloud is shown in the following step S23b.

[0068] S23b. Use any edge in the initial triangle as a growing edge, and select the optimal extended point cloud corresponding to the growing edge from the feature point cloud set; in this embodiment, the core of the traditional Delaunay triangulation algorithm is to find the optimal point that meets the conditions. It needs to meet the empty circumscribed circle criterion during triangulation, that is, each time a point is found, a circumscribed circle must be made for the triangle, and then it is determined whether the circumscribed circle contains other points. In this way, this irregular selection method will consume a lot of time when the point cloud data volume is large; therefore, this embodiment reduces the search range by limiting the search range of the extended point cloud, thereby improving the search speed.

[0069] Optionally, for example but not limited to, the following steps S23b1 to S23b4 may be used to select the optimal extended point cloud corresponding to the growing edge.

[0070] S23b1. Select the feature point cloud closest to the growing edge from the feature point cloud set as the initial extended point cloud; after selecting the feature point cloud closest to the growing edge from the feature point cloud set, an initial extended triangle can be constructed based on it, as shown in the following step S23b2.

[0071] S23b2. Based on the initial extended point cloud and the growing edge, an initial extended triangle is constructed; in the present embodiment, the initial extended point cloud is connected to the two ends of the growing edge respectively, so that the initial extended triangle can be obtained, and then, the optimal extended point cloud can be searched within the range of the circumscribed circle of the initial extended triangle, and the process is shown in the following steps S23b3 to S23b4.

[0072] S23b3. Determine whether there is a specified feature point cloud within the circumscribed circle of the initial extended triangle, wherein the specified feature point cloud is all feature point clouds in the feature point cloud set excluding the feature point clouds corresponding to the three vertices of the initial extended triangle; in this embodiment, if there is a specified feature point cloud within the circumscribed circle of the initial extended triangle, randomly select one from the specified feature point clouds as the initial extended point cloud, and then reconnect it to the growing edge to re-initialize the extended triangle; then, re-determine whether there is a specified feature point cloud in the circumscribed circle of the obtained initial extended triangle, and in this way, based on this principle, continue searching until there is no specified feature point cloud in the circumscribed circle. At this time, the initial extended point cloud corresponding to the initial extended triangle is the optimal extended point cloud; wherein the search process is shown in the following step S23b4.

[0073] S23b4. If so, randomly select one from the specified feature point clouds as the initial extended point cloud, and reconstruct the initial extended triangle based on the initial extended point cloud and the growth edge until the specified feature point cloud does not exist in the circumscribed circle of the constructed initial extended triangle, and use the initial extended point cloud corresponding to the initial extended triangle in which the specified feature point cloud does not exist in the circumscribed circle as the optimal extended point cloud.

[0074] In this embodiment, an example is used to illustrate: Assume that the growing edge is AB, and the first initial extended point cloud is the feature point cloud C, among which there are five feature point clouds D, E, F, G and H in the circumscribed circle of the initial triangle ABC. At this time, one is randomly selected. Assuming that the feature point cloud E is selected, there are three feature point clouds F, G and H in the circumscribed circle of its corresponding initial extended triangle ABE. At this time, one is randomly selected from the three feature point clouds F, G and H as the initial extended point cloud. Assuming that G is selected, there are no other feature point clouds in the circumscribed circle of the initial extended triangle ABG constructed by it. Therefore, the feature point cloud G can be used as the optimal extended point cloud of the growing edge AB.

[0075] Therefore, through the above explanation, when selecting the optimal extension point for triangulation, this embodiment limits the selection range to the circumscribed circle of the initial extended triangle constructed by the point cloud closest to the growing edge. In this way, compared with the traditional irregular selection, the search range is greatly reduced, thereby improving the search efficiency.

[0076] After the optimal expansion point cloud corresponding to the growing edge is determined based on the aforementioned step S23b and its sub-steps, the triangle can be expanded based on this, and the steps are shown in the following step S23c.

[0077] S23c. Using the optimal extended point cloud and the growing edge, an extended triangle is constructed. In this embodiment, the optimal extended point cloud is connected to the two ends of the growing edge respectively to obtain an extended triangle. For example, based on the above example, the optimal extended point cloud is the feature point cloud G, then the extended triangle is triangle ABG. Thus, after one triangulation, the triangulation network includes the initial extended triangle and the extended triangle ABG. Based on this, a new growing edge can be determined based on the extended triangle, and then the aforementioned steps S23b and S23c are repeated to obtain a new extended triangle. Thus, the aforementioned process is continuously repeated to complete the three-dimensional segmentation of the feature point cloud set, thereby generating a three-dimensional surface model of the initial excavation contour of the target construction tunnel. The aforementioned cyclic segmentation process is shown in the following step S23d.

[0078] S23d. Update the initial triangle to the extended triangle, and use any edge in the initial triangle as the growing edge again, and select the optimal extended point cloud corresponding to the growing edge from the feature point cloud set until all feature point clouds in the feature point cloud set are divided, so as to construct the initial excavation contour three-dimensional surface model using the initial triangle and the obtained extended triangle.

[0079] Thus, through the aforementioned steps S23a to S23d, the surface reconstruction of the excavation section of the target construction tunnel can be completed, and then, the local density of each of the aforementioned feature point clouds can be used to optimize the triangular facets, and the process is shown in the following step S24.

[0080] S24. According to the local density of each feature point cloud in the initial excavation contour three-dimensional surface model, the initial excavation contour three-dimensional surface model is optimized by triangular patch processing, so as to obtain the actual excavation contour three-dimensional surface model after the triangular patch optimization processing; in specific application, the optimization process is shown in the following steps S24a to S24d as an example.

[0081] S24a. For any initial triangular facet in the initial excavation contour three-dimensional surface model, calculate the circumscribed circle radius of any initial triangular facet; in this embodiment, the example can be but not limited to summing up the side lengths of any initial triangular facet, and dividing the sum by 2 to obtain a first intermediate parameter; then, based on the side lengths of the three sides of any initial triangular facet and the first intermediate parameter, calculate the second intermediate parameter; finally, calculate the circumscribed circle radius based on the second intermediate parameter and the side lengths of the three sides of any initial triangular facet.

[0082] Optionally, the second intermediate parameter may be calculated by, for example but not limited to, using the following formula (5).

[0083] (5) In the above formula (5), represents the second intermediate parameter, represents the first intermediate parameter, Represent the lengths of the three sides of any initial triangular facet in sequence.

[0084] Correspondingly, after the second intermediate parameter is calculated, the radius of the circumscribed circle can be calculated based on the following formula (6).

[0085] (6) In the above formula (6), represents the radius of the circumscribed circle.

[0086] After calculating the circumscribed circle radius of any initial triangular facet, the optimization of any initial triangular facet can be performed based on the local density of the feature point cloud corresponding to the three endpoints of the initial triangular facet, and the process is shown in the following steps S24b to S24d.

[0087] S24b. Calculate the density threshold corresponding to any initial triangular patch according to the local density of the feature point cloud corresponding to the three endpoints of any initial triangular patch; in this embodiment, for example, but not limited to, first calculate the density mean of the local density of the feature point cloud corresponding to the three endpoints of any initial triangular patch; then, multiply the density mean by the density coefficient to obtain the density threshold; optionally, the density coefficient is 0.5.

[0088] After calculating the density threshold corresponding to any of the initial triangular facets, it is possible to determine whether any of the initial triangular facets needs to be optimized in combination with the radius of its circumscribed circle, and the process is shown in the following step S24c.

[0089] S24c. According to the circumscribed circle radius and the density threshold, determine whether any of the initial triangular facets is a segmented triangular facet; in this embodiment, whether the circumscribed circle radius is greater than the density threshold is determined to determine whether any of the initial triangular facets needs to be optimized; wherein, the larger the circumscribed circle radius is, the larger the spatial range representing the interior of the triangle is, and the sparser the point cloud distribution is (the reason is: the area of ​​a triangle with a large circumscribed circle radius must be larger, which means that the distance between the three points of the triangle is farther, and therefore, it cannot cover the surrounding point cloud well, that is, the point cloud distribution is relatively sparse, and at the same time, the inner angle of a triangle with a large circumscribed circle radius is usually smaller, which means that the triangle is relatively flat and cannot fit the shape of the surrounding surface well, and is therefore not suitable for reconstructing the surface). Therefore, if it is greater than the density threshold, it means that the density of the point cloud distribution in the area is low, which will affect the reconstruction accuracy; therefore, it is necessary to optimize any of the initial triangular facets, that is, insert new points to increase the density to ensure the reconstruction accuracy; otherwise, it is not necessary to optimize and can be retained; optionally, the optimization process is shown in the following step S24d.

[0090] S24d. If so, a new feature point cloud is inserted into any of the initial triangular facets, so as to segment any of the initial triangular facets based on the new feature point cloud, and when all the initial triangular facets in the initial excavation contour three-dimensional surface model are polled, the actual excavation contour three-dimensional surface model is obtained; in this embodiment, the new feature point cloud is taken as the midpoints of the three edges of any of the initial triangular facets, that is, the midpoints of the three edges are used as the new feature point cloud, and then the original three endpoints are connected to subdivide any of the initial triangular facets into four triangles, so as to achieve the purpose of increasing the density. In this way, according to the aforementioned principle, after all the initial triangular facets in the initial excavation contour three-dimensional surface model are optimized, the actual excavation contour three-dimensional surface model of the target construction tunnel can be obtained.

[0091] Therefore, through the aforementioned steps S21 to S24 and their sub-steps, the collected three-dimensional point cloud data set can be used to construct a three-dimensional surface model of the actual excavation contour of the target construction tunnel; then, the under-excavation detection of the tunnel can be performed in combination with its theoretical excavation contour three-dimensional surface model, and the process is shown in the following steps S3 to S6.

[0092] S3. For any vertex in the actual excavation contour three-dimensional surface model, determine the triangular facet closest to the any vertex from the theoretical excavation contour three-dimensional surface model as the under-excavation detection reference plane corresponding to the any vertex, and use the distance between the any vertex and the under-excavation detection reference plane as the construction error distance; in this embodiment, the distance calculation formula between a point and a plane is used to calculate the distance from any vertex to each triangular facet in the theoretical excavation contour three-dimensional surface model; then, select the triangular facet with the closest distance as the corresponding under-excavation detection reference plane.

[0093] Based on this, it is possible to determine whether each point on the actual construction excavation section is an under-excavated point based on the positive or negative distance from each vertex on the actually constructed three-dimensional surface model of the excavation contour (i.e., the actual construction excavation section) to the corresponding reference plane; wherein the judgment process is shown in the following step S4.

[0094] S4. Determine whether the construction error distance is 0; in this embodiment, if the distance is 0, it means that the point on the actual excavation section is on the reference plane. At this time, it means that there is no problem of under-excavation or over-excavation; on the contrary, if the distance is not 0, it is necessary to determine whether any vertex is an under-excavation point by judging the positive or negative value of the distance; in this embodiment, the sign of the aforementioned distance (i.e., the construction error distance) is determined by the vector of the normal from any vertex to its corresponding reference plane, and the process is shown in the following step S5.

[0095] S5. If not, the sign of the construction error distance is determined according to the vector of the normal from any vertex to the under-excavation detection reference plane; in the present embodiment, it is assumed that the three endpoints of the under-excavation detection reference plane are FHK, wherein the coordinates of points F and H can be used to obtain vector FH, and the coordinates of points F and K can be used to obtain vector FK, and then, the cross product of vector FH and vector FK is obtained to obtain the normal (i.e., normal vector) of the under-excavation detection reference plane; then, the vector from any vertex to the normal can be calculated; finally, the dot product of the vector and the normal can be calculated, if the dot product is greater than 0, it means that the two vectors are in the same direction, the any vertex is substantially located on the inner side of the reference plane, and the distance is positive, therefore, it means that the any vertex is an over-excavation point, otherwise, if the dot product result is less than 0, it means that the any vertex is substantially located on the outer side of the reference plane, and the distance is negative, indicating that the any vertex is an under-excavation point.

[0096] In this embodiment, the calculation process of the vector from any vertex to the normal line is: , where represents the vector from any vertex to the normal, the position vector of any vertex (i.e. the position vector from any vertex to the origin), and the unit vector of the normal of the undercut detection reference plane, respectively. Indicates the distance from any vertex to the reference plane.

[0097] Further, the above judgment process is shown in the following step S6 S6. If the sign of the construction error distance is negative, any of the vertices is determined to be an under-excavation point, and after all vertices in the actual excavation contour three-dimensional surface model are polled, the under-excavation detection result of the target construction tunnel is obtained; in this way, the under-excavation point in the actual excavation contour three-dimensional surface model can be determined through the sign of the aforementioned construction error distance, thereby obtaining the under-excavation detection result of the target construction tunnel; based on this, it is visualized and displayed, which can help construction personnel adjust the construction plan.

[0098] Thus, through the high-speed railway tunnel construction under-break detection method described in detail in the aforementioned steps S1 to S6, the present invention constructs a three-dimensional surface model of the actual excavation contour of the tunnel by collecting the three-dimensional point cloud data set of the target construction tunnel, and then, from the three-dimensional surface model of the theoretical excavation contour of the tunnel, the triangular facets closest to each vertex in the actual model are determined, and used as the under-break detection reference plane of each vertex; finally, by calculating the positive and negative distances from the vertices to the corresponding reference planes, it is determined whether each vertex in the actual model is an under-break point, thereby obtaining the final under-break detection result; thus, compared with traditional technologies, the three-dimensional point cloud data can be directly obtained by scanning with a laser scanner without the need for point-by-point measurement, thereby not only improving efficiency but also reducing the workload of measurement; at the same time, a large amount of three-dimensional point cloud data can reflect the actual geometric shape of the tunnel, so that an accurate construction excavation section can be constructed therethrough, thereby ensuring the accuracy of detection; thus, the present invention is very suitable for large-scale application and promotion.

[0099] like Figure 2 As shown, the second aspect of this embodiment provides a hardware device for implementing the high-speed railway tunnel construction under-break detection method described in the first aspect of the embodiment, including: The acquisition unit is used to acquire a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel.

[0100] A model building unit is used to build a three-dimensional surface model of the actual excavation contour of the target construction tunnel based on the three-dimensional point cloud data set, wherein the actual excavation contour three-dimensional surface model and the theoretical excavation contour three-dimensional surface model are both triangular mesh models.

[0101] The error detection unit is used to determine, for any vertex in the actual excavation contour three-dimensional surface model, the triangular facet closest to any vertex in the theoretical excavation contour three-dimensional surface model, as the under-excavation detection reference plane corresponding to any vertex, and the distance between any vertex and the under-excavation detection reference plane is used as the construction error distance.

[0102] The underbreak detection unit is used to determine whether the construction error distance is 0.

[0103] The underbreak detection unit is used to determine the sign of the construction error distance according to the vector of the normal line from any vertex to the underbreak detection reference plane when it is judged that the construction error distance is not 0.

[0104] If the sign of the construction error distance is negative, the under-excavation detection unit is further used to determine that any of the vertices is an under-excavation point, and after polling all vertices in the actual excavation contour three-dimensional surface model, obtain the under-excavation detection result of the target construction tunnel.

[0105] The working process, working details and technical effects of the device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0106] like Figure 3 As shown, the third aspect of this embodiment provides another device for detecting under-excavation in high-speed railway tunnel construction. Taking the device as an electronic device as an example, it includes: a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the method for detecting under-excavation in high-speed railway tunnel construction as described in the first aspect of the embodiment.

[0107] For example, the memory may include, but is not limited to, random access memory (RAM), read only memory (ROM), flash memory, first input first output (FIFO) and / or first in last out (FILO), etc. Specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). At the same time, the processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state.

[0108] In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. For example, the processor may not be limited to a microprocessor of the STM32F105 series, a reduced instruction set computer (RISC) microprocessor, an X86 architecture processor, or a processor with an integrated embedded neural network processing unit (NPU); the transceiver may be, but is not limited to, a wireless fidelity (WIFI) wireless transceiver, a Bluetooth wireless transceiver, a General Packet Radio Service (GPRS) wireless transceiver, a ZigBee protocol (a low-power LAN protocol based on the IEEE802.15.4 standard, ZigBee) wireless transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver. In addition, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0109] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0110] The fourth aspect of this embodiment provides a storage medium storing instructions for the high-speed railway tunnel construction under-excavation detection method as described in the first aspect of the embodiment, that is, the storage medium stores instructions, and when the instructions are run on a computer, the high-speed railway tunnel construction under-excavation detection method as described in the first aspect of the embodiment is executed.

[0111] The storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, a CD, a hard disk, a flash memory, a USB flash drive and / or a memory stick, and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0112] The working process, working details and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0113] The fifth aspect of this embodiment provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the method for detecting under-excavation in high-speed railway tunnel construction as described in the first aspect of the embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0114] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting under-break in high-speed railway tunnel construction, characterized in that: include: Acquire a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel; Based on the three-dimensional point cloud data set, a three-dimensional surface model of an actual excavation profile of the target construction tunnel is constructed, wherein the three-dimensional surface model of the actual excavation profile and the three-dimensional surface model of the theoretical excavation profile are both triangular mesh models; For any vertex in the actual excavation contour three-dimensional surface model, determine the triangle patch closest to the any vertex from the theoretical excavation contour three-dimensional surface model as the underbreak detection reference plane corresponding to the any vertex, and use the distance from the any vertex to the underbreak detection reference plane as the construction error distance; Determine whether the construction error distance is 0; If not, the sign of the construction error distance is determined according to the vector of the normal line from any vertex to the underbreak detection reference plane; If the sign of the construction error distance is a negative sign, any of the vertices is determined to be an under-excavation point, and after all vertices in the actual excavation contour three-dimensional surface model are polled, the under-excavation detection result of the target construction tunnel is obtained.

2. The method according to claim 1, characterized in that Based on the three-dimensional point cloud data set, a three-dimensional surface model of the actual excavation contour of the target construction tunnel is constructed, including: Extracting a plurality of feature point clouds from the three-dimensional point cloud data set to form a feature point cloud set; Calculating the local density of each feature point cloud in the feature point cloud set; Using the feature point cloud set, constructing a three-dimensional surface model of the initial excavation profile of the target construction tunnel; According to the local density of each feature point cloud in the initial excavation contour three-dimensional surface model, the initial excavation contour three-dimensional surface model is subjected to triangular patch optimization processing, so as to obtain the actual excavation contour three-dimensional surface model after the triangular patch optimization processing.

3. The method according to claim 2, characterized in that Extracting multiple feature point clouds from the three-dimensional point cloud data set to form a feature point cloud set, including: Initialize the sampling iteration number s, and obtain the side length of the voxel grid at the s-th sampling, wherein when s is 1, the side length of the voxel grid at the s-th sampling is the initial side length; According to the side length of the voxel grid at the s-th sampling, the three-dimensional point cloud data set is divided into a plurality of voxel grids, wherein any voxel grid contains a plurality of three-dimensional point clouds; Extract the feature point cloud at the sth sampling time from multiple three-dimensional point clouds in each voxel grid; Determine whether a sampling iteration stop condition is met, wherein the sampling iteration stop condition is that the number of feature point clouds at the s-th sampling is equal to a preset number; If not, updating the side length of the voxel grid at the s-th sampling according to the number of feature point clouds at the s-th sampling and the preset number to obtain an updated side length; Increasing s by 1, replacing the side length of the voxel grid at the s-th sampling with the updated side length, and re-dividing the three-dimensional point cloud data set into a plurality of voxel grids according to the side length of the voxel grid at the s-th sampling until the sampling iteration stop condition is satisfied, so as to form an initial feature point cloud set using the feature point cloud sampled when the sampling iteration stop condition is satisfied; The initial feature point cloud set is subjected to denoising processing to obtain the feature point cloud set after the denoising processing.

4. The method according to claim 3, characterized in that From the multiple 3D point clouds in each voxel grid, extract the feature point cloud at the sth sampling, including: For any three-dimensional point cloud in any voxel grid, determine a neighboring point cloud of the any three-dimensional point cloud from the any voxel grid and a target grid, wherein the target grid is a voxel grid adjacent to the any voxel grid among the plurality of voxel grids; Determine a first normal vector of any three-dimensional point cloud and a second normal vector of each neighboring point cloud; Calculating the angles between the first normal vector and each second normal vector to obtain a plurality of normal vector angles; Find the average value of multiple normal vector angles; Determine whether the average value is greater than or equal to an angle threshold; If so, any one of the three-dimensional point clouds is taken as a feature point cloud in any one of the voxel grids, and after the three-dimensional point clouds in all the voxel grids are polled, the feature point cloud at the s-th sampling is obtained.

5. The method according to claim 3, characterized in that: According to the number of feature point clouds at the s-th sampling and the preset number, the side length of the voxel grid at the s-th sampling is updated to obtain the updated side length, including: Calculating a voxel grid side length adjustment coefficient at the s-th sampling according to the number of feature point clouds at the s-th sampling and the preset number; The updated side length is calculated based on the voxel grid side length adjustment coefficient and the initial side length at the s-th sampling.

6. The method according to claim 2, characterized in that Using the feature point cloud set, a three-dimensional surface model of the initial excavation profile of the target construction tunnel is constructed, including: Randomly select three feature point clouds from the feature point cloud set to form an initial triangle; Taking any edge of the initial triangle as a growing edge, and selecting the optimal extended point cloud corresponding to the growing edge from the feature point cloud set; Using the optimal extended point cloud and the growing edge, construct an extended triangle; The initial triangle is updated to the extended triangle, and any edge in the initial triangle is used as the growing edge again, and the optimal extended point cloud corresponding to the growing edge is screened out from the feature point cloud set until all feature point clouds in the feature point cloud set are divided, so as to construct the initial excavation contour three-dimensional surface model using the initial triangle and the obtained extended triangle.

7. The method according to claim 6, characterized in that Filtering out the optimal extended point cloud corresponding to the growing edge from the feature point cloud set includes: Selecting a feature point cloud closest to the growing edge from the feature point cloud set as an initial extended point cloud; Constructing an initial extended triangle based on the initial extended point cloud and the growing edge; Determine whether there is a specified feature point cloud within the circumscribed circle of the initial extended triangle, wherein the specified feature point cloud is all feature point clouds in the feature point cloud set except for the feature point clouds corresponding to the three vertices of the initial extended triangle; If so, one is randomly selected from the specified feature point clouds as the initial extended point cloud, and the initial extended triangle is reconstructed based on the initial extended point cloud and the growth edge until there is no specified feature point cloud within the circumscribed circle of the constructed initial extended triangle, and the initial extended point cloud corresponding to the initial extended triangle in which there is no specified feature point cloud within the circumscribed circle is used as the optimal extended point cloud.

8. The method according to claim 2, characterized in that: Calculating the local density of each feature point cloud in the feature point cloud set includes: For any feature point cloud in the feature point cloud set, obtaining a k-neighborhood point cloud of the feature point cloud; Calculate the distance between any feature point cloud and each k-neighborhood point cloud, and use the average of all distances as the local density of any feature point cloud; Correspondingly, according to the local density of each feature point cloud in the initial excavation contour three-dimensional surface model, the initial excavation contour three-dimensional surface model is subjected to triangular facet optimization processing, which includes: For any initial triangular facet in the initial excavation contour three-dimensional surface model, calculating the circumscribed circle radius of any initial triangular facet; Calculate a density threshold corresponding to any initial triangular facet according to the local density of the feature point cloud corresponding to the three endpoints of any initial triangular facet; According to the circumscribed circle radius and the density threshold, determining whether any of the initial triangular facets is a segmented triangular facet; If so, a new feature point cloud is inserted into any of the initial triangular facets so as to segment any of the initial triangular facets based on the new feature point cloud, and when all the initial triangular facets in the initial excavation contour three-dimensional surface model are polled, the actual excavation contour three-dimensional surface model is obtained.

9. A high-speed railway tunnel construction under-break detection device, characterized in that: include: An acquisition unit, used for acquiring a three-dimensional point cloud data set of a target construction tunnel and a three-dimensional surface model of a theoretical excavation profile of the target construction tunnel; A model building unit, configured to build a three-dimensional surface model of an actual excavation profile of the target construction tunnel based on the three-dimensional point cloud data set, wherein both the three-dimensional surface model of the actual excavation profile and the three-dimensional surface model of the theoretical excavation profile are triangular mesh models; The error detection unit is used to determine, for any vertex in the actual excavation contour three-dimensional surface model, a triangular facet closest to the any vertex from the theoretical excavation contour three-dimensional surface model as an underbreak detection reference plane corresponding to the any vertex, and to use the distance between the any vertex and the underbreak detection reference plane as a construction error distance; An underbreak detection unit, used to determine whether the construction error distance is 0; The underbreak detection unit is used to determine the sign of the construction error distance according to the vector of the normal line from any vertex to the underbreak detection reference plane when it is determined that the construction error distance is not 0; If the sign of the construction error distance is negative, the under-excavation detection unit is further used to determine that any of the vertices is an under-excavation point, and after polling all vertices in the actual excavation contour three-dimensional surface model, obtain the under-excavation detection result of the target construction tunnel.

10. A computer program product comprising instructions, characterized in that When the instruction is executed on a computer, the computer is caused to execute the method for detecting under-excavation in high-speed railway tunnel construction as claimed in any one of claims 1 to 8.

Images