Verification method of tunnel surrounding rock space structure recognition algorithm based on 3D printing

Abstract

The invention relates to a verification method for a tunnel surrounding rock space structure recognition algorithm based on 3D printing, and the method comprises the steps: employing three-dimensional modeling software to make a 3D virtual tunnel face model based on a semicircle, and carrying out the printing through a 3D printing technology, and obtaining a 3D printing test piece; carrying out photogrammetry on the 3D printing test piece by adopting a multi-view digital photographing method to obtain a plurality of photos of the 3D printing test piece; performing three-dimensional reconstruction according to the plurality of photos of the 3D printing test piece to obtain a preliminary point cloud model, and cutting and editing the preliminary point cloud model to obtain point cloud information of the 3D printing test piece; and identifying the point cloud information of the 3D printing test piece by adopting a to-be-verified algorithm to obtain discontinuous surface groups, independent structural surface information, trace identification results and base vector distribution conditions of the 3D printing test piece, and comparing the base vector distribution conditions with original feature data of the 3D printing test piece to complete verification of the to-be-verified algorithm. According to the invention, the rock mass structural surface automatic identification algorithm can be verified.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel engineering, and particularly to a verification method for an algorithm for identifying the spatial structure of tunnel surrounding rock based on 3D printing. Background Art

[0002] In tunnel engineering, a large number of randomly distributed and geometrically complex structural planes are contained in rock masses. Conducting investigations on the geometric distribution and attitude analysis of rock mass structural planes is one of the basic tasks for carrying out engineering rock mass quality classification and stability research. The tunnel face is an important part of the spatial structure of tunnel surrounding rock. Conducting investigations on the geometric structural planes and attitude analysis of the tunnel face is of great significance for the classification and stability research of engineering rock masses. Information such as the attitude, roughness, and trace lines of the structural planes on the rock mass tunnel face has a great influence on the mechanical behavior of the rock mass and is an important basis in the process of tunnel design and construction.

[0003] The non-contact measurement method is a three-dimensional refined acquisition technology for tunnel surrounding rock information. It can be divided into photogrammetry and three-dimensional laser scanning method according to different equipment and principles used. For automatic spatial structure recognition, many scholars at home and abroad have been conducting algorithm research. Many scholars use the normal vector distribution of point clouds to conduct clustering and extraction of discontinuous directions. However, there are certain problems with this method. The method of extracting the attitude information of structural planes through the point cloud clustering algorithm cannot distinguish between tectonic joints and non-tectonic joints generated by blasting, resulting in non-tectonic joints also being included in the calculation of tunnel surrounding rock quality evaluation. Moreover, this method is rarely verified by standardized indoor tests, and field verification can only be based on manual operation tools such as compasses and tape measures, and the accuracy is difficult to meet. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a verification method for an algorithm for identifying the spatial structure of tunnel surrounding rock based on 3D printing, which can verify the automatic recognition algorithm of rock mass structural planes.

[0005] The technical solution adopted by the present invention to solve its technical problems is: to provide a verification method for an algorithm for identifying the spatial structure of tunnel surrounding rock based on 3D printing, including the following steps:

[0006] Using three-dimensional modeling software to produce a 3D virtual tunnel face model based on a semi-circle and printing it through 3D printing technology to obtain a 3D printed specimen;

[0007] Using multi-view digital photography to conduct photogrammetry on the 3D printed specimen to obtain multiple photos of the 3D printed specimen;

[0008] Performing three-dimensional reconstruction based on the multiple photos of the 3D printed specimen to obtain a preliminary point cloud model, and trimming and editing the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen;

[0009] The point cloud information of the 3D printed specimen is identified using the algorithm to be verified, obtaining the discontinuous surface grouping, independent structural surface information, trace identification results, and base vector distribution of the 3D printed specimen, and comparing the base vector distribution with the original feature data of the 3D printed specimen to complete the verification of the algorithm to be verified.

[0010] The 3D virtual tunnel face model based on a semi - circle is made using 3D modeling software, specifically including:

[0011] Using 3D modeling software to generate several triangular pyramid modules within a semi - circular area, recording the characteristic parameters of the triangular pyramid modules, and combining several triangular pyramid modules into a set;

[0012] Using 3D modeling software to generate a semi - circular cylinder;

[0013] Align the semi - circular part of the semi - circular cylinder with the set and have partial overlap;

[0014] Perform a Boolean difference operation on the set and the semi - circular cylinder. After the operation is completed, move the set away to obtain a 3D virtual tunnel face model based on a semi - circle; the Boolean difference operation targets the set and is applied to the semi - circular cylinder.

[0015] When using the multi - camera digital photography method to perform photogrammetry on the 3D printed specimen, the 3D printed specimen is photographed in sequence from left to right. Among them, in the multiple photos of the 3D printed specimen obtained, at least 50% or more of the adjacent photos have the same part.

[0016] Performing three - dimensional reconstruction based on the multiple photos of the 3D printed specimen to obtain a preliminary point cloud model, specifically including:

[0017] Import the multiple photos of the 3D printed specimen into 3D reconstruction software, perform data calculation of feature points and camera positions to obtain a preliminary three - dimensional point cloud distribution;

[0018] Add details and color to the preliminary three - dimensional point cloud distribution to obtain a preliminary point cloud model.

[0019] Cropping and editing the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen. Specifically, import the preliminary point cloud model into point cloud editing software, adjust the position and crop the non - test area of the three - dimensional point cloud of the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen.

[0020] Beneficial effects

[0021] Due to the adoption of the above technical solution, compared with the prior art, the present invention has the following advantages and positive effects: The present invention designs a unique three-dimensional tunnel surrounding rock space structure heading face model, and manufactures it through 3D printing technology. Based on the photogrammetry method of multi-camera digital photography, the above 3D heading face model is photographed, and then the photographed result is re-modeled. The 3D printed heading face model is identified through the automatic identification algorithm of rock mass structural planes, and the identification result is analyzed, completing the verification of the identification algorithm, thereby moving the verification research and analysis process of the automatic identification algorithm of rock mass structural planes indoors. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flowchart of a method for verifying an algorithm for identifying the space structure of tunnel surrounding rock based on 3D printing according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The following further elaborates the present invention in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

[0024] An embodiment of the present invention relates to a method for verifying an algorithm for identifying the space structure of tunnel surrounding rock based on 3D printing, as Figure 1 shown, including the following steps:

[0025] Step 1, use 3D modeling software to make a 3D virtual heading face model based on a semi-circle, and print it through 3D printing technology to obtain a 3D printed specimen.

[0026] The 3D modeling software used in this step is Blender 3D modeling software. The specific steps for making a 3D virtual heading face model based on a semi-circle using Blender 3D modeling software are as follows:

[0027] (1) Generate a number of triangular pyramid modules within a semi-circular area, and combine the number of triangular pyramid modules into a set, which is denoted as set 1. Among them, the attributes such as the point-line-plane angle direction, density, and size of the generated triangular pyramid modules can be set by oneself. In this embodiment, the initial normal vector data of the four faces of the generated triangular pyramid modules are respectively (0, 3, 1), (0, 1, 2), (1, 3, 2), (3, 0, 2). The density parameter of the triangular pyramid modules within the semi-circular area is 0.01. When generating a number of triangular pyramid modules, a random variable with a normal distribution can also be added to make their point-line-plane spatial position data randomly distributed within a certain range to improve the overall effect of making the 3D virtual heading face model.

[0028] (2) Generate a semi-cylindrical body, which is used to simulate the tunnel space, and this semi-cylindrical body is called Object 2.

[0029] (3) Align the semi-circular part of the semi-cylindrical body with the set and have partial overlap, that is, through functions such as moving, scaling, and rotating in Blender 3D modeling software, align the top surface (i.e., the semi-circular part) of the semi-cylindrical body with Set 1 and make them have partial overlap. By adjusting the degree of overlap, the attitude of the generated 3D virtual tunnel face can be controlled to a certain extent.

[0030] (4) Perform a Boolean difference operation on Set 1 and Object 2. After the operation is completed, move Set 1 away to obtain a 3D virtual tunnel face model based on a semi-circle. Among them, the Boolean difference operation targets Set 1 (i.e., the triangular pyramid module set) and is applied to Object 2 (i.e., the semi-cylindrical body). In this embodiment, the process of the Boolean difference operation is actually based on a semi-cylindrical body, and the triangular pyramid blocks in Set 1 are used as knives to dig out grooves of different sizes on the cylinder. At the same time, convex parts will be formed on the inner side of the cylinder.

[0031] According to the above steps, a 3D virtual tunnel face model based on a semi-circle can be obtained. However, this 3D virtual tunnel face model still needs to be processed before it can be delivered to the manufacturer for 3D printing. In this embodiment, the bevel modifier in Blender 3D modeling software is used. Through this bevel modifier, the edges of the mesh can be cut into bevels, and the position and degree of the bevel can be controlled. After being modified by the bevel modifier, the concave and convex parts of the 3D virtual tunnel face model can be closer to the state in reality. There are few completely sharp parts in the connections in the real world, so the bevel treatment is beneficial to endow the non-natural model with a sense of reality, and the bevel design of the edges and corners is also convenient for 3D printing. The processing of the 3D virtual tunnel face model also needs to trim it to a state convenient for 3D printing, so a ring of hole walls is added. The diameter of the semi-circle of this hole wall is 1 m, the thickness is 0.5 mm, and the longitudinal distance of the hole wall is 0.1 m. After the above trimming, the 3D virtual tunnel face model can be printed by 3D printing technology to obtain a 3D printed specimen.

[0032] Step 2, use the multi-view digital photography method to perform photogrammetry on the 3D printed specimen to obtain multiple photos of the 3D printed specimen.

[0033] The multi-view digital photography method is a new field extended from the binocular digital photography method. By using one or more cameras to take pictures of the same object from multiple angles, different photos of this object in different directions can be obtained, that is, the two-dimensional image information of the simulated tunnel face of the 3D printed specimen can be obtained, so as to meet the requirements of three-dimensional reconstruction.

[0034] When performing photogrammetry using multi-view digital photography technology, it can be completed by a handheld mobile device. The three-dimensional information photography acquisition method for the tunnel face based on mobile phone photography has the advantages of fast acquisition speed, simple and convenient operation. Only 4 - 12 photos need to be taken on site, and it can be completed in 1 - 2 minutes. When taking photos in this embodiment, the 3D printed specimen is photographed in sequence from left to right. Among the multiple photos of the 3D printed specimen obtained, at least 50% or more of the adjacent photos have the same part, so that the selection of feature points can be satisfied. In order to present a better three-dimensional reconstruction effect, the number of photos taken is preferably more than 5 - 6 and not more than 12.

[0035] Step 3: Perform three-dimensional reconstruction based on the multiple photos of the 3D printed specimen to obtain a preliminary point cloud model, and crop and edit the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen.

[0036] In this step, the software used for three-dimensional reconstruction is RealityCapture, which is one of the most advanced photogrammetry software and can automatically extract accurate three-dimensional models from a set of photos or laser scan data. The point cloud editing software used in this step is CloudCompare, which is an open-source and free three-dimensional point cloud (mesh) processing software with functions of processing triangular meshes and calibrating images, and is equipped with basic tools for manually editing and rendering three-dimensional point clouds and triangular meshes, integrating many point cloud processing algorithms and display enhancement tools. This step specifically includes the following steps:

[0037] (1) Import the multiple photos of the 3D printed specimen into RealityCapture, click "Align Images" to perform data calculation of feature points and camera positions, and obtain a preliminary three-dimensional point cloud distribution;

[0038] (2) Add details to the preliminary three-dimensional point cloud distribution. In this embodiment, general-level details are generated through the NormlDetail function in the RealityCapture software, and the surrounding shape of the three-dimensional point cloud generation can be controlled by adjusting the acquisition frame.

[0039] (3) Color the three-dimensional point cloud after adding details. In this embodiment, the three-dimensional point cloud data model is colored through the Colorize function in the RealityCapture software. At this time, the obtained three-dimensional point cloud model is already relatively close to the overall state of the original 3D printed specimen.

[0040] (4) Select the relatively ideal 3D point cloud data and save it as a.ply data model file, and import it into the CloudCompare point cloud editing software. Use the CloudCompare software to perform editing operations such as position adjustment and cropping on the non-test area of the 3D point cloud, and finally obtain the reconstructed point cloud model. Specifically, the imported 3D point cloud model may contain many irrelevant points. Use the cropping function to crop off other parts and only leave the core area. The cropping operation can be performed from three angles of the three views of the 3D point cloud respectively. Finally, store the 3D reconstructed point cloud coordinates of the edited 3D printed specimen in.txt format.

[0041] Step 4: Use the algorithm to be verified to identify the point cloud information of the 3D printed specimen, obtain the discontinuous surface grouping, independent structural surface information, trace recognition result and base vector distribution of the 3D printed specimen, and compare the base vector distribution with the original feature data of the 3D printed specimen to complete the verification of the algorithm to be verified.

[0042] The algorithm to be verified in this step is the fast k-means++ algorithm. In identifying the occurrence of the structural surface from the point cloud information of the 3D printed specimen obtained in the previous step through the fast k-means++ algorithm and extracting the structural surface trace information using the trace recognition algorithm, this process can be implemented through a MATLAB program developed based on the k-means++ contour algorithm. This automatic recognition program can perform visualization processing on the recognition results. The specific operation process is as follows:

[0043] (1) Import the.txt format point cloud file obtained in the previous step into the MATLAB program. After successful import, display the original point cloud;

[0044] (2) Click on occurrence recognition to perform occurrence recognition work. After identifying the main occurrences of the structural surface, group them based on the structural surface normal vector information;

[0045] (3) Generate independent discontinuous surfaces based on the structural surface grouping results and display the independent structural surfaces; among them, the generation of independent structural surfaces can be completed using the double clustering algorithm.

[0046] (4) Obtain the trace feature map of the 3D printed specimen through the operation program based on the trace recognition algorithm and generate a distribution map of the base vector density.

[0047] By analyzing the distribution diagram of the basis vector density, it can be seen that there are three obvious directions in the recognition result of the basis vector density. In step 1, when generating the Boolean triangular pyramid module, the directions of the three faces of the triangular pyramid are preset and randomized with a normal distribution. After comparison, the initial position information of the triangular pyramid: (0, 3, 1), (0, 1, 2), (1, 3, 2), (3, 0, 2) is consistent with the main occurrence information of the structural plane shown by the basis vector density, thus verifying the reliability of the fast k-means++ algorithm.

[0048] It is not difficult to find that the present invention designs a unique three-dimensional tunnel surrounding rock space structure heading face model, manufactures it through 3D printing technology, takes pictures of the above 3D heading face model based on the photogrammetry method of multi-camera digital photography, remodels the photographing results, identifies the 3D printed heading face model through the automatic recognition algorithm of rock mass structural planes, analyzes the recognition results, completes the verification of the recognition algorithm, and thus moves the verification research and analysis process of the automatic recognition algorithm of rock mass structural planes indoors.

Claims

1. A verification method for an algorithm for identifying the spatial structure of tunnel surrounding rock based on 3D printing, characterized in that, It includes the following steps: Use 3D modeling software to create a 3D virtual tunnel face model based on a semi - circle, and print it through 3D printing technology to obtain a 3D printed specimen; Use the multi - view digital photography method to conduct photogrammetry on the 3D printed specimen to obtain multiple photos of the 3D printed specimen; perform 3D reconstruction based on the multiple photos of the 3D printed specimen to obtain a preliminary point cloud model, and crop and edit the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen; Use the algorithm to be verified to identify the point cloud information of the 3D printed specimen, obtain the discontinuous surface grouping, independent structural surface information, trace recognition results, and base vector distribution of the 3D printed specimen, and compare the base vector distribution with the original feature data of the 3D printed specimen to complete the verification of the algorithm to be verified.

2. The verification method of the tunnel surrounding rock spatial structure recognition algorithm based on 3D printing according to claim 1, characterized in that The step of using 3D modeling software to create a 3D virtual tunnel face model based on a semi - circle specifically includes: Use 3D modeling software to generate several triangular pyramid modules within a semi - circular area, record the characteristic parameters of the triangular pyramid modules, and combine several triangular pyramid modules into a set; Use 3D modeling software to generate a semi - circular cylinder; Align the semi - circular part of the semi - circular cylinder with the set and perform partial overlap; Perform a Boolean difference operation on the set and the semi - circular cylinder. After the operation is completed, move the set away to obtain a 3D virtual tunnel face model based on a semi - circle; the Boolean difference operation targets the set and is applied to the semi - circular cylinder.

3. The verification method of the tunnel surrounding rock spatial structure recognition algorithm based on 3D printing according to claim 1, characterized in that, When using the multi - view digital photography method to conduct photogrammetry on the 3D printed specimen, the 3D printed specimen is photographed in sequence from left to right. Among them, at least 50% or more of the adjacent photos in the multiple photos of the 3D printed specimen are the same.

4. The verification method of the tunnel surrounding rock spatial structure recognition algorithm based on 3D printing according to claim 1, characterized in that The step of performing 3D reconstruction based on the multiple photos of the 3D printed specimen to obtain a preliminary point cloud model specifically includes: Import the multiple photos of the 3D printed specimen into 3D reconstruction software, perform data calculation of feature points and camera positions to obtain a preliminary 3D point cloud distribution; Add details and color to the preliminary 3D point cloud distribution to obtain a preliminary point cloud model.

5. The verification method of the tunnel surrounding rock spatial structure recognition algorithm based on 3D printing according to claim 1, characterized in that The step of cropping and editing the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen is specifically to import the preliminary point cloud model into point cloud editing software, adjust the position and crop the non - test area of the 3D point cloud of the preliminary point cloud model to obtain the point cloud information of the 3D printed specimen.

Images