Three-dimensional laser point cloud detection method and system for wide and thick lining belt in water endowed area of tunnel

Abstract

The invention provides a three-dimensional laser point cloud detection method and system for a tunnel water-endowed area lining wide and thick belt, and the method comprises the steps: detecting the tunnel water-endowed area lining wide and thick belt through a rolling ball method, scanning the tunnel point cloud through three-dimensional laser, and collecting the laser point cloud data of the tunnel water-endowed area lining wide and thick belt; establishing a high-fidelity tunnel twinborn model; simulating tunnel deformation conditions through finite elements, constructing a finite element simulation tunnel deformation analysis model, and outputting a tunnel deformation true value of the tunnel twinning model; building a tunnel simulation scanning virtual simulation platform according to the tunnel twinborn model and the laser scanning characteristic data; performing simulated three-dimensional laser scanning on the tunnel twin model in a virtual environment to obtain large sample detection data; training a deformation detection neural network model in an auxiliary manner according to the large sample detection data; performing tunnel point cloud registration through a neural network model; a tunnel section point cloud is obtained by fitting the central axis of the tunnel, and tunnel deformation analysis is carried out; and the tunnel deformation analysis processing accuracy is obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional laser point cloud detection of the lining wide and thick belt in a tunnel water-bearing area and the laser detection tunnel engineering safety technology; more specifically, the present invention relates to a method and system for three-dimensional laser point cloud detection of the lining wide and thick belt in a tunnel water-bearing area. Background Art

[0002] Tunnel engineering safety technology is very crucial. Due to the slow deformation of tunnels and the difficulty in obtaining effective experimental detection data, the research on tunnel deformation detection technology is limited; problems such as how to detect the lining wide and thick belt in a tunnel water-bearing area and collect data of the lining wide and thick belt in a tunnel water-bearing area, how to simulate the tunnel deformation situation, simulate the tunnel deformation to analyze the true value of tunnel deformation, how to build a tunnel simulation scanning virtual simulation platform and simulate three-dimensional laser scanning in a virtual environment to obtain large-sample detection data, how to more efficiently optimize and train a neural network model and perform tunnel point cloud registration and tunnel deformation analysis remain to be solved; therefore, it is necessary to propose a three-dimensional laser point cloud detection system and method for the lining wide and thick belt in a tunnel water-bearing area to at least partially solve the problems existing in the prior art. Summary of the Invention

[0003] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further elaborated in the Detailed Description section; the Summary of the Invention section of the present invention does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0004] To at least partially solve the above problems, the present invention provides a method for three-dimensional laser point cloud detection of the lining wide and thick belt in a tunnel water-bearing area, including: S10, detecting the lining wide and thick belt in a tunnel water-bearing area by the rolling ball method, scanning the tunnel point cloud by three-dimensional laser, and collecting the laser point cloud data of the lining wide and thick belt in a tunnel water-bearing area; establishing a high-fidelity tunnel twin model; S20, simulating the tunnel deformation situation by finite element, constructing a finite element simulation tunnel deformation analysis model, and outputting the true value of tunnel deformation of the tunnel twin model; S30, building a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; performing simulated three-dimensional laser scanning on the tunnel twin model in a virtual environment to obtain large-sample detection data; S40, assisting in training a deformation detection neural network model according to the large-sample detection data; performing tunnel point cloud registration through the neural network model; obtaining tunnel cross-section point cloud by fitting the tunnel central axis and performing tunnel deformation analysis.

[0005] Preferably, S10 includes: S101. Detect the thick lining belt of the water-rich area of the tunnel by the rolling ball method. Through the strain and humidity sensing optical fiber, estimate the damaged defect area and obtain the humidity information of the defect area. S102. During the process of the spherical body rolling around the water-rich area of the tunnel in all directions, collect the laser point cloud data of the thick lining belt of the water-rich area of the tunnel by three-dimensional laser scanning of the tunnel point cloud. S103. Establish a high-fidelity tunnel twin model based on the damaged defect area, the humidity information of the defect area, and the laser point cloud data of the thick lining belt of the water-rich area of the tunnel.

[0006] Preferably, S20 includes: S201. Establish the finite element simulation metadata of the sampling nodes of the thick lining belt of the water-rich area. The finite element simulation metadata includes the rock and soil geological structure, the water content of the rock and soil, the thickness of the rock and soil, and the tunnel deformation amount. S202. According to the analysis information of the tunnel deformation amount of the single metadata, intelligently allocate the proportion among the influence weights of the geological structure, the water content of the rock and soil, and the thickness of the rock and soil that affect the tunnel deformation amount. S203. Construct a finite element simulation tunnel deformation analysis model; according to the finite element simulation tunnel deformation analysis model, output the true value of the tunnel deformation of the tunnel twin model.

[0007] Preferably, S30 includes: S301. Build a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data. S302. Conduct simulated three-dimensional laser scanning on the tunnel twin model in the virtual environment to obtain a large sample of detection data.

[0008] Preferably, S40 includes: S401. Assist in training the deformation detection neural network model according to the large sample of detection data; the deformation detection neural network model includes the Geotransformer neural network model. S402. Perform tunnel point cloud registration through the neural network model; obtain the tunnel cross-section point cloud by fitting the tunnel central axis and conduct tunnel deformation analysis. S403. Determine the tunnel deformation position in the tunnel cross-section point cloud and infer the tunnel deformation evolution trend. Obtaining the tunnel cross-section point cloud by fitting the tunnel central axis and conducting tunnel deformation analysis includes: inputting the collected tunnel cross-section data; selecting a small number of points to construct an error equation and solve it; distinguishing inliers and outliers according to the threshold; determining whether the number of inliers meets 90%; if the number of inliers does not meet 90%, then remove the outliers and integrate the inliers; output the inliers that meet the conditions and conduct tunnel cross-section analysis based on the inliers.

[0009] The present invention provides a three-dimensional laser point cloud detection system for the lining wide and thick belt in a tunnel water-bearing area, including: A twin model subsystem for the lining of the tunnel water-bearing area, which detects the wide and thick belt of the tunnel lining in the water-bearing area by the rolling ball method, scans the tunnel point cloud by three-dimensional laser, and collects the laser point cloud data of the wide and thick belt of the tunnel lining in the water-bearing area; establishes a high-fidelity tunnel twin model; A finite element simulation tunnel deformation subsystem, which simulates the tunnel deformation situation by finite element, constructs a finite element simulation tunnel deformation analysis model, and outputs the true value of the tunnel deformation of the tunnel twin model; A tunnel simulation scanning virtual simulation subsystem, which builds a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; conducts simulated three-dimensional laser scanning on the tunnel twin model in a virtual environment to obtain a large sample of detection data; A tunnel point cloud registration and deformation analysis subsystem, which assists in training a deformation detection neural network model according to the large sample of detection data; performs tunnel point cloud registration through the neural network model; obtains the tunnel cross-section point cloud by fitting the tunnel central axis and conducts tunnel deformation analysis.

[0010] Preferably, the twin model subsystem for the lining of the tunnel water-bearing area includes: A rolling ball method for detecting tunnel defects module, which detects the wide and thick belt of the tunnel lining in the water-bearing area by the rolling ball method, estimates the damaged defect area through strain humidity sensing optical fiber, and obtains the humidity information of the defect area; A three-dimensional laser scanning tunnel point cloud module, which collects the laser point cloud data of the wide and thick belt of the tunnel lining in the water-bearing area by three-dimensional laser scanning of the tunnel point cloud during the rolling of a spherical body around the tunnel water-bearing area in all directions; A high-fidelity tunnel twin model architecture module, which establishes a high-fidelity tunnel twin model according to the damaged defect area, the humidity information of the defect area, and the laser point cloud data of the wide and thick belt of the tunnel lining in the water-bearing area.

[0011] Preferably, the finite element simulation tunnel deformation subsystem includes: A finite element simulation module for the lining in the water-bearing area, which establishes the finite element simulation metadata of the sampling nodes of the wide and thick belt of the lining in the water-bearing area. The finite element simulation metadata includes the rock and soil geological structure, the water content of the rock and soil, the thickness of the rock and soil, and the tunnel deformation amount; A tunnel deformation amount weight intelligent distribution module, which intelligently distributes the proportion among the geological structure influence weight, the rock and soil water content influence weight, and the rock and soil thickness influence weight that affect the tunnel deformation amount according to the tunnel deformation amount analysis information of a single metadata; A simulated tunnel deformation analysis model module, which constructs a finite element simulation tunnel deformation analysis model; and outputs the true value of the tunnel deformation of the tunnel twin model according to the finite element simulation tunnel deformation analysis model.

[0012] Preferably, the tunnel simulation scanning virtual simulation subsystem includes: Tunnel scanning simulation platform: build a tunnel simulation scanning virtual simulation platform based on the tunnel twin model and laser scanning feature data; The simulated 3D laser scanning module performs simulated 3D laser scanning on the tunnel twin model in a virtual environment to obtain large sample detection data.

[0013] Preferably, the tunnel point cloud registration deformation analysis subsystem includes: The detection data network model training module assists in training the deformation detection neural network model based on large sample detection data; the deformation detection neural network model includes the Geotransformer neural network model; The tunnel point cloud registration module performs tunnel point cloud registration through a neural network model; obtains tunnel cross-section point clouds by fitting the tunnel centerline, and performs tunnel deformation analysis; The tunnel deformation determination and inference module determines the tunnel deformation position in the tunnel section point cloud and infers the tunnel deformation evolution trend; The tunnel cross-section point cloud is obtained by fitting the tunnel centerline, and the tunnel deformation analysis includes: inputting the collected tunnel cross-section data; selecting a small number of points to construct the error equation and solving it; distinguishing the in-points and out-points according to the threshold value; determining whether the number of in-points meets 90%; if the number of in-points does not meet 90%, eliminating the out-points and integrating the in-points; outputting the in-points that meet the conditions, and performing tunnel cross-section analysis based on the in-points.

[0014] Compared with the prior art, the present invention has at least the following beneficial effects: A three-dimensional laser point cloud detection method and system for the lining wide and thick belt in a tunnel water-rich area of the present invention detects the lining wide and thick belt in the tunnel water-rich area through the rolling ball method, scans the tunnel point cloud by three-dimensional laser, and collects the laser point cloud data of the lining wide and thick belt in the tunnel water-rich area; establishes a high-fidelity tunnel twin model; simulates the tunnel deformation situation through finite element, constructs a finite element simulation tunnel deformation analysis model, and outputs the tunnel deformation true value of the tunnel twin model; builds a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; performs simulated three-dimensional laser scanning on the tunnel twin model in the virtual environment to obtain a large sample of detection data; assists in training a deformation detection neural network model according to the large sample of detection data; performs tunnel point cloud registration through the neural network model; obtains tunnel cross-section point cloud by fitting the tunnel central axis for tunnel deformation analysis; a digital twin method for tunnel deformation detection research, establishes a high-fidelity tunnel twin model; simulates the deformation situation of the tunnel through finite element and outputs the deformation true value of the tunnel twin model; builds a virtual simulation platform to realize three-dimensional laser scanning of the tunnel model in the virtual environment to obtain a large sample of detection data and assist in training a deformation detection method; the deformation detection uses the Geotransformer neural network to realize tunnel point cloud registration, obtains tunnel cross-section point cloud by fitting the tunnel central axis, and realizes tunnel deformation analysis; improves the safety of tunnel engineering, can solve the problem that it is difficult to obtain effective experimental detection data due to slow tunnel deformation; can detect the lining wide and thick belt in the tunnel water-rich area and collect the data of the lining wide and thick belt in the tunnel water-rich area; can more accurately simulate the tunnel deformation situation; can simulate tunnel deformation to analyze the tunnel deformation true value; can build a tunnel simulation scanning virtual simulation platform and simulate three-dimensional laser scanning in the virtual environment to obtain a large sample of detection data; can more efficiently optimize and train the neural network model and perform tunnel point cloud registration, significantly improving the accuracy of tunnel deformation analysis and processing; has important technical significance and remarkable effects.

[0015] A three-dimensional laser point cloud detection method and system for the lining wide and thick belt in a tunnel water-rich area according to the present invention, other advantages, objectives and features of the present invention will be partially reflected by the following description, and partially will also be understood by those skilled in the art through the research and practice of the present invention. Brief Description of the Drawings

[0016] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings: Figure 1 It is a diagram of an embodiment of a three-dimensional laser point cloud detection method for the lining wide and thick belt in a tunnel water-rich area according to the present invention.

[0017] Figure 2 It is a diagram of an embodiment of a three-dimensional laser point cloud detection system for the lining wide and thick belt in a tunnel water-rich area according to the present invention.

[0018] Figure 3 This is another embodiment diagram of a three-dimensional laser point cloud detection system for the wide and thick lining of a tunnel water-bearing area according to the present invention. Specific implementation manner

[0019] The following further elaborates on the present invention in conjunction with the accompanying drawings and embodiments, so that those skilled in the art can implement it with reference to the description. As shown in the figure, the present invention provides a method for detecting the three-dimensional laser point cloud of the wide and thick lining of a tunnel water-bearing area, including: S10. Detect the wide and thick lining of the tunnel water-bearing area by the rolling ball method, collect the laser point cloud data of the wide and thick lining of the tunnel water-bearing area by three-dimensional laser scanning the tunnel point cloud; establish a high-fidelity tunnel twin model. S20. Simulate the tunnel deformation situation by finite element method, construct a finite element simulation tunnel deformation analysis model, and output the true value of the tunnel deformation of the tunnel twin model. S30. Build a virtual simulation platform for tunnel simulation scanning according to the tunnel twin model and the laser scanning feature data; perform simulated three-dimensional laser scanning on the tunnel twin model in the virtual environment to obtain a large sample of detection data. S40. Assist in training a deformation detection neural network model according to the large sample of detection data; perform tunnel point cloud registration through the neural network model; obtain the tunnel cross-section point cloud by fitting the tunnel central axis, and perform tunnel deformation analysis.

[0020] The principle and effect of the above technical solution are as follows: A three-dimensional laser point cloud detection method for the lining wide and thick belt in a tunnel water-rich area includes: detecting the lining wide and thick belt in the tunnel water-rich area by the rolling ball method, scanning the tunnel point cloud by three-dimensional laser, and collecting the laser point cloud data of the lining wide and thick belt in the tunnel water-rich area; establishing a high-fidelity tunnel twin model; simulating the tunnel deformation by finite element, constructing a finite element simulation tunnel deformation analysis model, and outputting the true value of the tunnel deformation of the tunnel twin model; building a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; performing simulated three-dimensional laser scanning on the tunnel twin model in the virtual environment to obtain a large sample of detection data; assisting in training a deformation detection neural network model according to the large sample of detection data; performing tunnel point cloud registration through the neural network model; obtaining tunnel cross-section point cloud by fitting the tunnel central axis and performing tunnel deformation analysis; a digital twin method for tunnel deformation detection research, establishing a high-fidelity tunnel twin model; simulating the tunnel deformation by finite element and outputting the true value of the deformation of the tunnel twin model; building a virtual simulation platform to realize three-dimensional laser scanning of the tunnel model in the virtual environment to obtain a large sample of detection data and assist in training a deformation detection method; the deformation detection uses the Geotransformer neural network to realize tunnel point cloud registration, obtains tunnel cross-section point cloud by fitting the tunnel central axis, and realizes tunnel deformation analysis; improves the safety of tunnel engineering, can solve the problem that it is difficult to obtain effective experimental detection data due to slow tunnel deformation; can detect the lining wide and thick belt in the tunnel water-rich area and collect data of the lining wide and thick belt in the tunnel water-rich area; can more accurately simulate the tunnel deformation situation; can simulate tunnel deformation and analyze the true value of tunnel deformation; can build a tunnel simulation scanning virtual simulation platform and perform simulated three-dimensional laser scanning in the virtual environment to obtain a large sample of detection data; can more efficiently optimize the training of the neural network model and perform tunnel point cloud registration, significantly improving the accuracy of tunnel deformation analysis and processing; has important technical significance and remarkable effects.

[0021] In one embodiment, S10 includes: S101, detecting the lining wide and thick belt in the tunnel water-rich area by the rolling ball method, estimating the damaged defect area through strain and humidity sensing optical fibers, and obtaining the humidity information of the defect area; S102, during the process of the spherical body rolling around the tunnel water-rich area in all directions, scanning the tunnel point cloud by three-dimensional laser and collecting the laser point cloud data of the lining wide and thick belt in the tunnel water-rich area; S103, establishing a high-fidelity tunnel twin model according to the damaged defect area, the humidity information of the defect area, and the laser point cloud data of the lining wide and thick belt in the tunnel water-rich area.

[0022] The principle and effect of the above technical solution are as follows: As Figure 3As shown in the figure, the rolling ball method is used to detect the thick and wide belt of the lining in the water-rich area of the tunnel. Through the strain and humidity sensing optical fiber, the damaged defect area is estimated, and the humidity information of the defect area is obtained. During the process of the spherical body rolling around the water-rich area of the tunnel in all directions (left, right, up, and down), the 3D laser scanning tunnel point cloud is used to collect the laser point cloud data of the thick and wide belt of the lining in the water-rich area of the tunnel. Based on the damaged defect area, the humidity information of the defect area, and the laser point cloud data of the thick and wide belt of the lining in the water-rich area of the tunnel, a high-fidelity tunnel twin model is established. By using the rolling ball method to detect the thick and wide belt of the lining in the water-rich area of the tunnel, the strain optical fiber detection information and the humidity and moisture-sensing detection information are obtained. According to the strain optical fiber detection information and the humidity and moisture-sensing detection information, the damaged defect area is estimated, and the humidity information of the defect area is obtained, including: setting a spherical body with a radius of hr, and arranging the strain and humidity sensing optical fiber on the surface of the spherical body; the spherical body rolls around the water-rich area of the tunnel in all directions (left, right, up, and down), the area contacted by the spherical body is the estimated damaged defect area, and the space not contacted by the spherical body is the effective protection space; the strain and humidity sensing optical fiber includes the grating optical fiber band 1011 on the surface of the spherical body, the humidity-sensitive material 1012 on the surface of the spherical body, and the optical fiber humidity signal detection module 1013; the grating optical fiber band on the surface of the spherical body detects the area contacted by the spherical body according to the fiber strain, and obtains the fiber strain signal of the spherical body contacting the optical fiber; the humidity causes changes in the humidity-sensitive material on the surface of the spherical body, and according to the changes in the humidity-sensitive material on the surface of the spherical body, a change signal is formed, and the change signal of the spherical body contacting the humidity-sensitive material is obtained; the optical fiber humidity signal detection module detects the fiber strain signal of the spherical body contacting the optical fiber and the change signal of the spherical body contacting the humidity-sensitive material, obtains the strain optical fiber detection information and the humidity and moisture-sensing detection information, and according to the strain optical fiber detection information and the humidity and moisture-sensing detection information, estimates the damaged defect area and obtains the humidity information of the defect area.

[0023] In one embodiment, S20 includes: S201, establishing the finite element simulation metadata of the sampling nodes of the thick and wide belt of the lining in the water-rich area, and the finite element simulation metadata includes the rock and soil geological structure, the water content of the rock and soil, the thickness of the rock and soil, and the tunnel deformation; S202, according to the analysis information of the tunnel deformation of the single metadata, intelligently allocating the proportion among the influence weights of the geological structure, the water content of the rock and soil, and the thickness of the rock and soil that affect the tunnel deformation; S203, constructing a finite element simulation tunnel deformation analysis model; according to the finite element simulation tunnel deformation analysis model, outputting the true value of the tunnel deformation of the tunnel twin model.

[0024] The principles and effects of the above technical solution are as follows: Establish the finite element simulation metadata of the sampling nodes in the lining wide-thickness zone of the water-bearing area. The finite element simulation metadata includes the geological structure of rock strata and soil, the water content of rock strata and soil, the thickness of rock strata and soil, and the tunnel deformation amount. According to the tunnel deformation amount analysis information of single metadata, intelligently allocate the proportion among the influence weights of geological structure, water content of rock strata and soil, and thickness of rock strata and soil on the tunnel deformation amount. Construct a finite element simulation tunnel deformation analysis model. According to the finite element simulation tunnel deformation analysis model, output the true value of tunnel deformation of the tunnel twin model. According to the tunnel deformation amount analysis information of single metadata, establish the finite element simulation metadata of the sampling nodes in the lining wide-thickness zone of the water-bearing area. The finite element simulation metadata includes the geological structure of rock strata and soil, the water content of rock strata and soil, the thickness of rock strata and soil, and the tunnel deformation amount. According to the finite element simulation metadata, establish the metadata simulation analysis relationship among the geological structure of rock strata and soil, the water content of rock strata and soil, the thickness of rock strata and soil, and the tunnel deformation amount. The metadata simulation analysis relationship includes: mainly obtain the rock type data, soil type data, and rock-soil ratio data in the geological structure of rock strata and soil. The rock type is analyzed by calculating the rock strength, the soil type is analyzed by calculating the water absorption and softness, and the rock-soil ratio is analyzed by calculating the statistical average of the content between rock and soil. The lower the rock strength, the higher the water absorption and softness of the soil, and the lower the rock-soil ratio, the greater the influence of the geological structure on the tunnel deformation amount, which is set as the influence weight of the geological structure on the tunnel deformation amount. When the water content of rock strata and soil increases, the pressure borne by the lining wide-thickness zone of the water-bearing area increases, and the plastic deformation of the tunnel section increases, which is set as the influence weight of the water content of rock strata and soil on the tunnel deformation amount. The greater the thickness of rock strata and soil, the greater the pressure borne by the tunnel and the lining wide-thickness zone of the water-bearing area, and the greater the influence on the tunnel deformation amount, which is set as the influence weight of the thickness of rock strata and soil on the tunnel deformation amount. Respectively, simulate and analyze the tunnel deformation amount according to single metadata to obtain the tunnel deformation amount analysis information of single metadata. According to the tunnel deformation amount analysis information of single metadata, allocate the influence weights according to the proportion of the tunnel deformation amount size, and intelligently allocate the proportion among the influence weights of geological structure, water content of rock strata and soil, and thickness of rock strata and soil on the tunnel deformation amount, and construct a finite element simulation tunnel deformation analysis model.

[0025] In one embodiment, S30 includes: S301, build a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; S302, perform simulated three-dimensional laser scanning on the tunnel twin model in the virtual environment to obtain a large sample of detection data.

[0026] The principle and effect of the above technical solution are as follows: based on the tunnel twin model and laser scanning feature data, a tunnel simulation scanning virtual simulation platform is built; the tunnel twin model is simulated by three-dimensional laser scanning in a virtual environment to obtain large sample detection data; based on the tunnel twin model and laser scanning feature data, a tunnel simulation scanning virtual simulation platform is built, including: through a physical three-dimensional laser scanning device in the tunnel, the scanning data of the lining width and thickness of multiple sections of the water-bearing area in the tunnel are collected; the physical three-dimensional laser scanning device is set on the initial tunnel centerline; a strain temperature-sensitive optical fiber fitting the initial tunnel centerline is set on the initial tunnel centerline, and the scanning baseline of the physical three-dimensional laser scanning device is verified to be consistent with the initial tunnel centerline; through the physical three-dimensional laser scanning device, the data of the lining width and thickness of multiple sections of the water-bearing area in the tunnel are collected; a one-to-one correspondence between the physical three-dimensional laser scanning and the simulated three-dimensional laser scanning data of the lining width and thickness of the water-bearing area is established, and a tunnel simulation scanning virtual simulation platform is built.

[0027] In one embodiment, S40 includes: S401, assisting in training a deformation detection neural network model based on large sample detection data; the deformation detection neural network model includes a Geotransformer neural network model; S402, performing tunnel point cloud registration through a neural network model; obtaining a tunnel cross-section point cloud by fitting the tunnel centerline, and performing tunnel deformation analysis; S403, determining the tunnel deformation position in the tunnel cross-section point cloud, and inferring the tunnel deformation evolution trend; like Figure 1 As shown in the figure, the tunnel section point cloud is obtained by fitting the tunnel central axis, and the tunnel deformation analysis includes: inputting the collected tunnel section data; selecting a small number of points to construct the error equation and solving it; distinguishing the in-points and out-points according to the threshold value; determining whether the number of in-points meets 90%; if the number of in-points does not meet 90%, then eliminating the out-points and integrating the in-points; outputting the in-points that meet the conditions, and performing tunnel section analysis based on the in-points.

[0028] The principle and effect of the above technical solution are as follows: according to large sample detection data, auxiliary training of deformation detection neural network model; deformation detection neural network model includes Geotransformer neural network model; tunnel point cloud registration is performed through neural network model; tunnel cross-section point cloud is obtained by fitting the tunnel central axis, and tunnel deformation analysis is performed; tunnel deformation position in tunnel cross-section point cloud is determined, and tunnel deformation evolution trend is inferred; tunnel cross-section point cloud is obtained by fitting the tunnel central axis, and tunnel deformation analysis includes: inputting collected tunnel cross-section data; selecting a small number of points to construct error equations and solving them; distinguishing between internal and external points according to the threshold value; determining whether the number of internal points meets 90%; if the number of internal points does not meet 90%, the external points are eliminated and the internal points are integrated; outputting the internal points that meet the conditions, and performing tunnel cross-section analysis based on the internal points; Determining the deformation position of the tunnel in the tunnel section point cloud and inferring the deformation evolution trend of the tunnel includes: obtaining multi-dimensional simulated tunnel deformation position index data through simulated three-dimensional laser scanning; comparing and verifying the multi-dimensional simulated tunnel deformation position index data with the scanning data of the lining width and thickness strips of multiple sections of the water-bearing area in the tunnel, and comparing and verifying the accuracy of the simulated three-dimensional laser scanning; when the verification error of the simulated three-dimensional laser scanning accuracy is lower than the set verification error, fitting the tunnel centerline with the multi-dimensional simulated tunnel deformation position index data to obtain the tunnel section point cloud, conduct tunnel deformation analysis, determine the tunnel deformation position in the tunnel section point cloud, and infer the deformation evolution trend of the tunnel.

[0029] like Figure 2 As shown, the present invention provides a three-dimensional laser point cloud detection system for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel, comprising: The tunnel water conservancy area lining twin model subsystem detects the width and thickness of the tunnel water conservancy area lining by the rolling ball method, collects the laser point cloud data of the width and thickness of the tunnel water conservancy area lining by 3D laser scanning of the tunnel point cloud, and establishes a high-fidelity tunnel twin model; The finite element simulation tunnel deformation subsystem simulates tunnel deformation through finite element simulation, builds a finite element simulation tunnel deformation analysis model, and outputs the true value of tunnel deformation of the tunnel twin model; The tunnel simulation scanning virtual simulation subsystem builds a tunnel simulation scanning virtual simulation platform based on the tunnel twin model and laser scanning feature data; simulates three-dimensional laser scanning of the tunnel twin model in a virtual environment to obtain large sample detection data; The tunnel point cloud registration deformation analysis subsystem assists in training the deformation detection neural network model based on large sample detection data; performs tunnel point cloud registration through the neural network model; obtains the tunnel cross-section point cloud by fitting the tunnel centerline, and performs tunnel deformation analysis.

[0030] The principles and effects of the above technical solution are as follows: The present invention provides a three-dimensional laser point cloud detection system for the lining wide and thick belt in a tunnel water-bearing area, including: a twin model subsystem for the lining in the tunnel water-bearing area, which detects the lining wide and thick belt in the tunnel water-bearing area by the rolling ball method, scans the tunnel point cloud by three-dimensional laser, and collects the laser point cloud data of the lining wide and thick belt in the tunnel water-bearing area; establishes a high-fidelity tunnel twin model; a finite element simulation tunnel deformation subsystem, which simulates the tunnel deformation situation by finite element, constructs a finite element simulation tunnel deformation analysis model, and outputs the true tunnel deformation value of the tunnel twin model; a tunnel simulation scanning virtual simulation subsystem, which builds a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; performs simulated three-dimensional laser scanning on the tunnel twin model in a virtual environment to obtain a large sample of detection data; a tunnel point cloud registration and deformation analysis subsystem, which assists in training a deformation detection neural network model according to the large sample of detection data; performs tunnel point cloud registration through the neural network model; obtains the tunnel cross-section point cloud by fitting the tunnel central axis and conducts tunnel deformation analysis; a digital twin method for tunnel deformation detection, which establishes a high-fidelity tunnel twin model; simulates the tunnel deformation situation by finite element and outputs the true deformation value of the tunnel twin model; builds a virtual simulation platform to realize three-dimensional laser scanning of the tunnel model in a virtual environment to obtain a large sample of detection data and assist in training the deformation detection method; the deformation detection uses the Geotransformer neural network to realize tunnel point cloud registration, obtains the tunnel cross-section point cloud by fitting the tunnel central axis, and realizes tunnel deformation analysis; improves the safety of tunnel engineering, can solve the problem that it is difficult to obtain effective experimental detection data due to slow tunnel deformation; can detect the lining wide and thick belt in the tunnel water-bearing area and collect the data of the lining wide and thick belt in the tunnel water-bearing area; can more accurately simulate the tunnel deformation situation; can simulate tunnel deformation and analyze the true tunnel deformation value; can build a tunnel simulation scanning virtual simulation platform and perform simulated three-dimensional laser scanning in a virtual environment to obtain a large sample of detection data; can more efficiently optimize and train the neural network model and perform tunnel point cloud registration, significantly improving the accuracy of tunnel deformation analysis and processing; has important technical significance and remarkable effects.

[0031] In one embodiment, the twin model subsystem for the lining in the tunnel water-bearing area includes: A rolling ball method for detecting tunnel defects module, which detects the lining wide and thick belt in the tunnel water-bearing area by the rolling ball method, estimates the damaged defect area through strain and humidity sensing optical fibers, and obtains the humidity information of the defect area; A three-dimensional laser scanning tunnel point cloud module, which collects the laser point cloud data of the lining wide and thick belt in the tunnel water-bearing area by three-dimensional laser scanning of the tunnel point cloud during the rolling of a spherical body around the tunnel water-bearing area in all directions; A high-fidelity tunnel twin model architecture module, which establishes a high-fidelity tunnel twin model according to the damaged defect area, the humidity information of the defect area, and the laser point cloud data of the lining wide and thick belt in the tunnel water-bearing area.

[0032] The principle and effect of the above technical solution are as follows: As Figure 3 shown, the twin model subsystem for the lining of the tunnel water-bearing area includes: a module for detecting tunnel defects by the rolling ball method, which detects the wide and thick belt of the tunnel lining in the water-bearing area by the rolling ball method, estimates the damaged defect area through strain humidity sensing optical fibers, and obtains the humidity information of the defect area; a three-dimensional laser scanning tunnel point cloud module, which collects the laser point cloud data of the wide and thick belt of the tunnel lining in the water-bearing area by three-dimensional laser scanning of the tunnel point cloud during the rolling of the spherical body around the tunnel water-bearing area in all directions; a high-fidelity tunnel twin model architecture module, which establishes a high-fidelity tunnel twin model based on the damaged defect area, the humidity information of the defect area, and the laser point cloud data of the wide and thick belt of the tunnel lining in the water-bearing area; detecting the wide and thick belt of the tunnel lining in the water-bearing area by the rolling ball method, obtaining the strain optical fiber detection information and the humidity humidity-sensitive detection information, estimating the damaged defect area according to the strain optical fiber detection information and the humidity humidity-sensitive detection information, and obtaining the humidity information of the defect area includes: setting a spherical body with a radius of hr, and arranging strain humidity sensing optical fibers on the surface of the spherical body; the spherical body rolls around the tunnel water-bearing area in all directions, the area contacted by the spherical body is the estimated damaged defect area, and the space not contacted by the spherical body is the effective protection space; the strain humidity sensing optical fiber includes a grating optical fiber band 1011 on the surface of the spherical body, a humidity-sensitive material 1012 on the surface of the spherical body, and an optical fiber humidity signal detection module 1013; the grating optical fiber band on the surface of the spherical body detects the area contacted by the spherical body according to the optical fiber strain, and obtains the optical fiber strain signal of the spherical body contact; the humidity causes changes in the humidity-sensitive material on the surface of the spherical body, and a change signal is formed according to the changes in the humidity-sensitive material on the surface of the spherical body, and the change signal of the humidity-sensitive material contacted by the spherical body is obtained; the optical fiber humidity signal detection module detects the optical fiber strain signal of the spherical body contact and the change signal of the humidity-sensitive material contacted by the spherical body, obtains the strain optical fiber detection information and the humidity humidity-sensitive detection information, estimates the damaged defect area according to the strain optical fiber detection information and the humidity humidity-sensitive detection information, and obtains the humidity information of the defect area.

[0033] In one embodiment, the finite element simulation tunnel deformation subsystem includes: a finite element simulation module for the lining of the water-bearing area, which establishes the finite element simulation metadata of the sampling nodes of the wide and thick belt of the lining of the water-bearing area, and the finite element simulation metadata includes the geological structure of the rock and soil layer, the water content of the rock and soil layer, the thickness of the rock and soil layer, and the tunnel deformation amount; a tunnel deformation amount weight intelligent distribution module, which intelligently distributes the ratio among the geological structure influence weight, the rock and soil layer water content influence weight, and the rock and soil layer thickness influence weight that affect the tunnel deformation amount according to the analysis information of the tunnel deformation amount of a single metadata; a simulated tunnel deformation analysis model module, which constructs a finite element simulation tunnel deformation analysis model; and outputs the true value of the tunnel deformation of the tunnel twin model according to the finite element simulation tunnel deformation analysis model.

[0034] The principle and effect of the above technical solution are as follows: The finite element simulation tunnel deformation subsystem includes: a finite element simulation module for the lining in the water-rich area, which establishes the finite element simulation metadata of the sampling nodes in the wide-thickness belt of the lining in the water-rich area. The finite element simulation metadata includes the geological structure of rock and soil layers, the water content of rock and soil layers, the thickness of rock and soil layers, and the tunnel deformation amount; a tunnel deformation amount weight intelligent distribution module, which intelligently distributes the proportion among the influence weights of the geological structure, the water content of rock and soil layers, and the thickness of rock and soil layers on the tunnel deformation amount according to the analysis information of the tunnel deformation amount of a single metadata; a simulated tunnel deformation analysis model module, which constructs a finite element simulation tunnel deformation analysis model; and according to the finite element simulation tunnel deformation analysis model, outputs the true value of the tunnel deformation of the tunnel twin model. According to the analysis information of the tunnel deformation amount of a single metadata, establish the finite element simulation metadata of the sampling nodes in the wide-thickness belt of the lining in the water-rich area. The finite element simulation metadata includes the geological structure of rock and soil layers, the water content of rock and soil layers, the thickness of rock and soil layers, and the tunnel deformation amount; according to the finite element simulation metadata, establish the metadata simulation analysis relationship among the geological structure of rock and soil layers, the water content of rock and soil layers, the thickness of rock and soil layers, and the tunnel deformation amount. The metadata simulation analysis relationship includes: obtaining the main rock type data, soil type data, and rock-soil ratio data in the geological structure of rock and soil layers; analyzing the rock type according to the rock strength, analyzing the soil type according to the water absorption and softness, and analyzing the rock-soil ratio according to the statistical average of the content between rock and soil; the lower the rock strength, the higher the water absorption and softness of the soil, and the lower the rock-soil ratio, the greater the influence of the geological structure on the tunnel deformation amount, which is set as the influence weight of the geological structure on the tunnel deformation amount; when the water content of rock and soil layers increases, the pressure borne by the wide-thickness belt of the lining in the water-rich area increases, and the plastic deformation of the tunnel section increases, which is set as the influence weight of the water content of rock and soil layers on the tunnel deformation amount; the greater the thickness of rock and soil layers, the greater the pressure borne by the tunnel, the greater the pressure borne by the wide-thickness belt of the lining in the water-rich area, and the greater the influence on the tunnel deformation amount, which is set as the influence weight of the thickness of rock and soil layers on the tunnel deformation amount; respectively simulate and analyze the tunnel deformation amount according to a single metadata to obtain the analysis information of the tunnel deformation amount of a single metadata; according to the analysis information of the tunnel deformation amount of a single metadata, distribute the influence weight according to the proportion of the tunnel deformation amount size, and intelligently distribute the proportion among the influence weights of the geological structure, the water content of rock and soil layers, and the thickness of rock and soil layers on the tunnel deformation amount, and construct a finite element simulation tunnel deformation analysis model.

[0035] In one embodiment, the tunnel simulation scanning virtual simulation subsystem includes: A tunnel scanning simulation platform, which builds a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and the laser scanning feature data; A simulated three-dimensional laser scanning module, which performs simulated three-dimensional laser scanning on the tunnel twin model in a virtual environment to obtain a large sample of detection data.

[0036] The principle and effect of the above technical scheme are as follows: a tunnel simulation scanning virtual simulation subsystem includes: a tunnel scanning simulation platform, which builds a tunnel simulation scanning virtual simulation platform according to the tunnel twin model and laser scanning feature data; a simulated three-dimensional laser scanning module, which simulates three-dimensional laser scanning of the tunnel twin model in a virtual environment to obtain large sample detection data; according to the tunnel twin model and laser scanning feature data, a tunnel simulation scanning virtual simulation platform is built, including: through a physical three-dimensional laser scanning device in the tunnel, the scanning data of the lining width and thickness of multiple sections of the water-bearing area in the tunnel are collected; the physical three-dimensional laser scanning device is set on the initial tunnel centerline; a strain temperature-sensitive optical fiber that fits the initial tunnel centerline is set on the initial tunnel centerline, and the scanning baseline of the physical three-dimensional laser scanning device is verified to be consistent with the initial tunnel centerline; through the physical three-dimensional laser scanning device, the data of the lining width and thickness of multiple sections of the water-bearing area in the tunnel are collected; a one-to-one correspondence between the physical three-dimensional laser scanning and the simulated three-dimensional laser scanning data of the lining width and thickness of the water-bearing area is established, and a tunnel simulation scanning virtual simulation platform is built.

[0037] In one embodiment, the tunnel point cloud registration deformation analysis subsystem includes: The detection data network model training module assists in training the deformation detection neural network model based on large sample detection data; the deformation detection neural network model includes the Geotransformer neural network model; The tunnel point cloud registration module performs tunnel point cloud registration through a neural network model; obtains tunnel cross-section point clouds by fitting the tunnel centerline, and performs tunnel deformation analysis; The tunnel deformation determination and inference module determines the tunnel deformation position in the tunnel section point cloud and infers the tunnel deformation evolution trend; like Figure 1 As shown in the figure, the tunnel section point cloud is obtained by fitting the tunnel central axis, and the tunnel deformation analysis includes: inputting the collected tunnel section data; selecting a small number of points to construct the error equation and solving it; distinguishing the in-points and out-points according to the threshold value; determining whether the number of in-points meets 90%; if the number of in-points does not meet 90%, then eliminating the out-points and integrating the in-points; outputting the in-points that meet the conditions, and performing tunnel section analysis based on the in-points.

[0038] The principle and effects of the above technical solution are as follows: The tunnel point cloud registration deformation analysis subsystem includes: a detection data network model training module, which assists in training a deformation detection neural network model based on a large sample of detection data; the deformation detection neural network model includes a Geotransformer neural network model; a tunnel point cloud registration module, which performs tunnel point cloud registration through the neural network model; obtains tunnel cross-section point clouds by fitting the tunnel central axis for tunnel deformation analysis; a tunnel deformation determination and prediction module, which determines the tunnel deformation position in the tunnel cross-section point clouds and predicts the tunnel deformation evolution trend; obtaining tunnel cross-section point clouds by fitting the tunnel central axis for tunnel deformation analysis includes: inputting the collected tunnel cross-section data; selecting a small number of points to construct and solve the error equation; distinguishing inliers and outliers according to the threshold; determining whether the number of inliers meets 90%; if the number of inliers does not meet 90%, then removing the outliers and integrating the inliers; outputting the inliers that meet the conditions and performing tunnel cross-section analysis based on the inliers. Determining the tunnel deformation position in the tunnel cross-section point clouds and predicting the tunnel deformation evolution trend includes: obtaining multi-dimensional simulated tunnel deformation position index data through simulated 3D laser scanning; comparing and verifying the multi-dimensional simulated tunnel deformation position index data with the scanned data of the lining thickness bands of multiple water-bearing areas in the tunnel to verify the accuracy of the simulated 3D laser scanning; when the accuracy verification error of the simulated 3D laser scanning is lower than the set verification error, obtaining tunnel cross-section point clouds by fitting the tunnel central axis with the multi-dimensional simulated tunnel deformation position index data for tunnel deformation analysis, determining the tunnel deformation position in the tunnel cross-section point clouds, and predicting the tunnel deformation evolution trend.

[0039] Although the embodiments of the present invention have been disclosed as above, it is not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the examples shown and described herein.

Claims

1. A three-dimensional laser point cloud detection method for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel, characterized in that: include: S10, detect the lining width and thickness of the tunnel water supply area by the rolling ball method, and collect the laser point cloud data of the lining width and thickness of the tunnel water supply area by three-dimensional laser scanning of the tunnel point cloud; Build a high-fidelity tunnel twin model; S20, simulating the tunnel deformation by finite element method, constructing a finite element simulation tunnel deformation analysis model, and outputting the tunnel deformation true value of the tunnel twin model; S30, building a tunnel simulation scanning virtual simulation platform based on the tunnel twin model and laser scanning feature data; Conduct simulated 3D laser scanning of the tunnel twin model in a virtual environment to obtain large sample test data; S40, assisting in training a deformation detection neural network model based on large sample detection data; performing tunnel point cloud registration through the neural network model; The tunnel cross-section point cloud is obtained by fitting the tunnel centerline, and tunnel deformation analysis is performed.

2. A three-dimensional laser point cloud detection method for thick and wide strips of lining in a tunnel water-bearing area according to claim 1, characterized in that: S10 includes: S101, using the rolling ball method to detect the thickness of the lining in the water-bearing area of ​​the tunnel, using the strain humidity sensing optical fiber to estimate the damaged defect area and obtain the humidity information of the defect area; S102, when the spherical body rolls left and right and up and down around the tunnel water supply area, the tunnel point cloud is scanned by three-dimensional laser to collect the laser point cloud data of the lining width and thickness of the tunnel water supply area; S103, establishing a high-fidelity tunnel twin model based on the damaged defective area, the humidity information of the defective area and the laser point cloud data of the lining width and thickness of the tunnel water supply area.

3. The three-dimensional laser point cloud detection method for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel according to claim 1 is characterized in that: The S20 includes: S201, establishing finite element simulation metadata of sampling nodes of lining width and thickness belt in water-bearing area, wherein the finite element simulation metadata includes geological structure of rock layer soil, moisture content of rock layer soil, thickness of rock layer soil and deformation of tunnel; S202, intelligently allocating the proportions of the geological structure influence weight, the rock layer soil moisture content influence weight, and the rock layer soil thickness influence weight that affect the tunnel deformation based on the single metadata tunnel deformation analysis information; S203, constructing a finite element simulation tunnel deformation analysis model; outputting the tunnel deformation true value of the tunnel twin model according to the finite element simulation tunnel deformation analysis model.

4. The three-dimensional laser point cloud detection method for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel according to claim 1 is characterized in that: S30 includes: S301, building a tunnel simulation scanning virtual simulation platform based on the tunnel twin model and laser scanning feature data; S302, simulate three-dimensional laser scanning of the tunnel twin model in a virtual environment to obtain large sample detection data.

5. The three-dimensional laser point cloud detection method for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel according to claim 1 is characterized in that: S40 includes: S401, assisting in training a deformation detection neural network model based on large sample detection data; the deformation detection neural network model includes a Geotransformer neural network model; S402, performing tunnel point cloud registration through a neural network model; obtaining a tunnel cross-section point cloud by fitting the tunnel centerline, and performing tunnel deformation analysis; S403, determining the tunnel deformation position in the tunnel cross-section point cloud, and inferring the tunnel deformation evolution trend; The tunnel cross-section point cloud is obtained by fitting the tunnel centerline, and the tunnel deformation analysis includes: inputting the collected tunnel cross-section data; selecting a small number of points to construct the error equation and solving it; distinguishing the in-points and out-points according to the threshold value; determining whether the number of in-points meets 90%; if the number of in-points does not meet 90%, eliminating the out-points and integrating the in-points; outputting the in-points that meet the conditions, and performing tunnel cross-section analysis based on the in-points.

6. A three-dimensional laser point cloud detection system for the wide and thick strip of lining in the water-bearing area of ​​a tunnel, characterized in that: include: The tunnel water conservancy area lining twin model subsystem detects the width and thickness of the tunnel water conservancy area lining by the rolling ball method, collects the laser point cloud data of the width and thickness of the tunnel water conservancy area lining by 3D laser scanning of the tunnel point cloud, and establishes a high-fidelity tunnel twin model; The finite element simulation tunnel deformation subsystem simulates tunnel deformation through finite element simulation, builds a finite element simulation tunnel deformation analysis model, and outputs the true value of tunnel deformation of the tunnel twin model; The tunnel simulation scanning virtual simulation subsystem builds a tunnel simulation scanning virtual simulation platform based on the tunnel twin model and laser scanning feature data; simulates three-dimensional laser scanning of the tunnel twin model in a virtual environment to obtain large sample detection data; The tunnel point cloud registration deformation analysis subsystem assists in training the deformation detection neural network model based on large sample detection data; tunnel point cloud registration is performed through the neural network model; The tunnel cross-section point cloud is obtained by fitting the tunnel centerline, and tunnel deformation analysis is performed.

7. The three-dimensional laser point cloud detection system for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel according to claim 6 is characterized in that: The twin model subsystem of the tunnel water supply area lining includes: The rolling ball method is used to detect tunnel defect modules. The rolling ball method is used to detect the thickness of the lining in the water-bearing area of ​​the tunnel. The strain and humidity sensing optical fiber is used to estimate the damaged defect area and obtain the humidity information of the defect area. The 3D laser scanning tunnel point cloud module collects the laser point cloud data of the lining width and thickness of the tunnel water supply area through 3D laser scanning of the tunnel point cloud during the spherical body rolling left and right and up and down around the tunnel water supply area; The high-fidelity tunnel twin model architecture module establishes a high-fidelity tunnel twin model based on the damaged defect area, humidity information of the defect area and the laser point cloud data of the lining width and thickness of the tunnel water supply area.

8. The three-dimensional laser point cloud detection system for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel according to claim 6 is characterized in that: Finite element simulation tunnel deformation subsystem, including: The finite element simulation module for the lining of the water-bearing area is used to establish the finite element simulation metadata of the sampling nodes of the wide and thick belt of the lining of the water-bearing area. The finite element simulation metadata includes the geological structure of the rock layer and soil, the moisture content of the rock layer and soil, the thickness of the rock layer and soil, and the deformation of the tunnel; The intelligent allocation module of tunnel deformation weights intelligently allocates the proportions of geological structure weights, rock layer soil moisture content weights, and rock layer soil thickness weights that affect tunnel deformation based on the single metadata tunnel deformation analysis information; The tunnel deformation analysis model module is used to construct a finite element simulation tunnel deformation analysis model; based on the finite element simulation tunnel deformation analysis model, the true value of the tunnel deformation of the tunnel twin model is output.

9. The three-dimensional laser point cloud detection system for the wide and thick strip of the lining in the water-bearing area of ​​a tunnel according to claim 6 is characterized in that: Tunnel simulation scanning virtual simulation subsystem, including: Tunnel scanning simulation platform: build a tunnel simulation scanning virtual simulation platform based on the tunnel twin model and laser scanning feature data; The simulated 3D laser scanning module performs simulated 3D laser scanning on the tunnel twin model in a virtual environment to obtain large sample detection data.

10. The three-dimensional laser point cloud detection system for thick and wide strips of lining in a tunnel water supply area according to claim 6 is characterized in that: Tunnel point cloud registration deformation analysis subsystem, including: The detection data network model training module assists in training the deformation detection neural network model based on large sample detection data; the deformation detection neural network model includes the Geotransformer neural network model; The tunnel point cloud registration module performs tunnel point cloud registration through a neural network model; obtains the tunnel cross-section point cloud by fitting the tunnel centerline, and performs tunnel deformation analysis; The tunnel deformation determination and inference module determines the tunnel deformation position in the tunnel section point cloud and infers the tunnel deformation evolution trend; The tunnel cross-section point cloud is obtained by fitting the tunnel centerline, and the tunnel deformation analysis includes: inputting the collected tunnel cross-section data; selecting a small number of points to construct the error equation and solving it; distinguishing the in-points and out-points according to the threshold value; determining whether the number of in-points meets 90%; if the number of in-points does not meet 90%, eliminating the out-points and integrating the in-points; outputting the in-points that meet the conditions, and performing tunnel cross-section analysis based on the in-points.

Images