Three-dimensional tunnel deformation monitoring method and system based on radar point cloud comparison

Abstract

The invention provides a three-dimensional tunnel deformation monitoring method and system based on radar point cloud comparison, and belongs to the technical field of tunnel deformation monitoring. According to the invention, the laser radar is driven for multiple times to move along the same-direction route of the tunnel axis to collect the three-dimensional point cloud data, so that high-precision monitoring of the tunnel structure can be realized; comparing the radar position value of the point cloud data matrix with the overall point cloud data to obtain a first deformation matrix and a second deformation matrix; after the first deformation matrix and the second deformation matrix are subtracted, the deformation condition of the tunnel can be accurately recognized by judging whether the absolute value of each element in the deformation matrixes is larger than a threshold value or not, and detailed deformation information of the tunnel can be further obtained by using an unmanned aerial vehicle to mount a laser radar and an orthographic camera for flight scanning. The monitoring efficiency is improved, and the risk that personnel enter the tunnel for monitoring is reduced; the deformation condition of the tunnel can be visually displayed, and subsequent analysis and processing are facilitated.

Description

Technical Field

[0001] The invention relates to a three-dimensional tunnel deformation monitoring method and system based on radar point cloud comparison, and belongs to the technical field of tunnel deformation monitoring. Background Art

[0002] During the construction of highway tunnels, affected by the special geographical environment and complex geology, mud and water gushing or landslides often occur in the tunnels, which will bring very serious consequences to the tunnel engineering and personal property. Therefore, tunnel deformation monitoring technology is a crucial part of tunnel engineering. It aims to ensure the safety and stability of the tunnel structure during construction and operation. At present, it mainly uses advanced geological detection to make a macroscopic description and early warning of the geological environment, so as to formulate construction plans under complex geological conditions in advance. However, the existing detection equipment cannot meet the real-time monitoring and early warning of disasters.

[0003] For this purpose, patent CN116291724A proposes a real-time early warning detection method and system for highway tunnel construction, which discloses the following steps: turn on the laser radar to perform a large-scale scan of the initial support stage of the tunnel, the area where the monitoring measurement points are placed on the arch is the key area, and the key area is repeatedly scanned, and the data obtained by the radar scanning is processed to obtain the arch surface model; turn on the structured light scanner to scan and measure the key monitoring area to obtain a three-dimensional reconstruction model; monitor the falling stones and dripping water in the tunnel through the event camera; compare the information obtained by the laser radar, structured light scanner and event camera with the data obtained by the monitoring measurement during the construction process, and when the abnormal value reaches the set threshold, the system issues an early warning message. The system can monitor the deformation of the tunnel face, initial support deformation, dripping water and falling rocks in real time during the tunnel construction process, and transmit the data to the monitoring system or mobile device for the administrator to make a decision; but the applicant found that when the above technical solution is adopted, the setting of the early warning threshold is the key to the system's early warning function, but the appropriate threshold is often difficult to determine, and manual intervention is still required, and the untimely or inaccurate manual operation still has a hazard to the overall performance and safety of the system monitoring. Summary of the invention

[0004] In order to improve the accuracy of alarm threshold setting and the accuracy of tunnel deformation monitoring, the present invention provides a three-dimensional tunnel deformation monitoring method and system based on radar point cloud comparison. The technical solution is as follows:

[0005] The tunnel deformation monitoring method of the present invention comprises:

[0006] Step 1: Use laser radar to collect three-dimensional point cloud data of the tunnel and record the radar measurement position to generate the first point cloud data matrix;

[0007] Step 2: After a preset time interval, repeat the above data collection process to generate a second point cloud data matrix; extract the position value of the radar in the first point cloud data matrix and compare it with the installation position value of the radar in the second point cloud data matrix, and obtain the deformation matrix of the radar in the tunnel as the first deformation matrix; extract the first point cloud data matrix and compare it with the second point cloud data matrix, and obtain the deformation matrix of the tunnel as the second deformation matrix;

[0008] Step 3: Calculate the difference between the first deformation matrix and the second deformation matrix, check whether the absolute value of each element in the deformation matrix exceeds a set threshold, and determine whether there is a deformation area;

[0009] Step 4: If there is a deformed area, use the drone to fly twice, mount the lidar and orthophoto camera respectively, scan the same tunnel slope, obtain the third point cloud data, and construct the radar 3D model and orthophoto model;

[0010] Step 5: Remove the false deformation area and locate the potential tunnel deformation area through filtering and denoising algorithm;

[0011] Step 6: Mapping the tunnel deformation area to the orthophoto model for intuitive display through a radar three-dimensional model to orthophoto model mapping algorithm;

[0012] Step 7: Based on the three-dimensional point cloud data, perform overall intrusion detection and initial local intrusion detection. The results are uploaded to the cloud center for processing and displayed through the terminal.

[0013] Optionally, the filtering and denoising algorithm in step 5 includes:

[0014] Step 51: segment the tunnel in the radar 3D model and remove non-tunnel areas;

[0015] Step 52: Calculate the distance between the two phases of radar point clouds to obtain a difference point cloud;

[0016] Step 53: removing outlier points from the difference point cloud;

[0017] Step 54: marking the tunnel deformation area in the three-dimensional model.

[0018] Optionally, the overall intrusion detection analysis cuts and slices the point cloud data, uses RANSAC or least squares method to perform circle fitting, obtains the coordinates of the center of the circle and the radius parameters, and compares them with the tunnel design parameters to determine whether the primary support is deformed and intrudes.

[0019] Optionally, the process of the initial branch local intrusion detection and analysis includes: analyzing supervoxels of slice point cloud data established by Octree, calculating voxel neighborhood curvature, and comparing curvature differences to determine the local intrusion position.

[0020] Optionally, the method further includes: providing a key monitoring area on the tunnel vault, setting monitoring points in the key monitoring area, and adding a supplementary light source to the structured light scanner to supplement the brightness of the measuring points.

[0021] The tunnel deformation monitoring system of the present invention is used to implement the tunnel deformation monitoring method as described in any one of the above items, and the system includes: a computer, an unmanned aerial vehicle, an information and data acquisition module, a power supply and power management module, and a data exchange module; the information and data acquisition module includes a laser radar and a structured light scanner;

[0022] The computer controls the operation of the drone and controls the information and data acquisition sensor to collect and process data to determine the safety level of the current environment; the output control interface of the computer is connected to the sound and light alarm device; the computer exchanges data with the information and data acquisition module through the data exchange module.

[0023] Optionally, the power supply and power management module includes a backup power supply for emergency use.

[0024] Optionally, the system also includes a black box for storing and recording key data information.

[0025] The present invention provides a tunnel deformation monitoring device, comprising a memory and a processor;

[0026] The memory is used to store computer programs;

[0027] The processor is used to implement the tunnel deformation monitoring method as described in any one of the above items when executing the computer program.

[0028] The present invention provides a computer-readable storage medium, characterized in that a computer program is stored on the storage medium, and when the computer program is executed by a processor, the tunnel deformation monitoring method as described in any one of the above items is implemented.

[0029] The beneficial effects of the present invention are:

[0030] The present invention can realize high-precision monitoring of tunnel structure by repeatedly driving the laser radar to move along the same direction route of the tunnel axis to collect three-dimensional point cloud data; improve the accuracy of deformation judgment, compare the radar position values ​​of the first point cloud data matrix and the second point cloud data matrix and the overall point cloud data, and obtain two first deformation matrices and second deformation matrices; after subtracting the first deformation matrix from the second deformation matrix, by judging whether the absolute value of each element in the deformation matrix is ​​greater than a threshold, the deformation of the tunnel can be accurately identified, and the detailed deformation information of the tunnel can be further obtained by using the unmanned aerial vehicle mounted with the laser radar and the orthographic camera for flight scanning, which not only improves the accuracy of deformation monitoring threshold setting, but also effectively improves the monitoring efficiency, while reducing the risk of personnel entering the tunnel for monitoring; deformation warnings can be obtained in time and the deformation of the tunnel can be intuitively displayed, which is convenient for subsequent analysis and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 Schematic diagram of the monitoring process of the tunnel deformation monitoring method in an embodiment of the present invention.

[0033] Figure 2 Schematic diagram of the structure of a tunnel deformation monitoring system in an embodiment of the present invention.

[0034] Figure 3 Schematic diagram of the positions of radar and reflection detection in an embodiment of the present invention.

[0035] Figure 4 Schematic diagram of deformation monitoring by radar laser in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0037] Embodiment 1:

[0038] This embodiment provides a tunnel deformation monitoring method, including:

[0039] Step 1: Use laser radar to collect three-dimensional point cloud data of the tunnel and record the radar measurement position to generate the first point cloud data matrix;

[0040] Step 2: After a preset time interval, repeat the above data collection process to generate a second point cloud data matrix; extract the position value of the radar in the first point cloud data matrix and compare it with the installation position value of the radar in the second point cloud data matrix, and obtain the deformation matrix of the radar in the tunnel as the first deformation matrix; extract the first point cloud data matrix and compare it with the second point cloud data matrix, and obtain the deformation matrix of the tunnel as the second deformation matrix;

[0041] Step 3: Calculate the difference between the first deformation matrix and the second deformation matrix, check whether the absolute value of each element in the deformation matrix exceeds the set threshold, and determine whether there is a deformation area;

[0042] Step 4: If there is a deformed area, use the drone to fly twice, mount the lidar and orthophoto camera respectively, scan the same tunnel slope, obtain the third point cloud data, and construct the radar 3D model and orthophoto model;

[0043] Step 5: Remove the false deformation area and locate the potential tunnel deformation area through filtering and denoising algorithm;

[0044] Step 6: Use the radar 3D model to orthophoto model mapping algorithm to map the tunnel deformation area to the orthophoto model for intuitive display;

[0045] Step 7: Based on the three-dimensional point cloud data, perform overall intrusion detection and initial local intrusion detection. The results are uploaded to the cloud center for processing and displayed through the terminal.

[0046] Embodiment 2

[0047] This embodiment provides a three-dimensional tunnel deformation monitoring method based on radar point cloud comparison, see Figure 1 , the method comprising:

[0048] Step 1: Data collection;

[0049] The laser radar is driven to move multiple times along a route in the same direction as the tunnel axis, wherein there are several equidistant benchmark reset points on the route, wherein the tunnel is used as the installation benchmark, the laser radar is installed on the tunnel, the three-dimensional point cloud data of the tunnel is collected, and the measurement position points of the radar are recorded, and the first point cloud data matrix of the tunnel is generated according to the three-dimensional point cloud data and uploaded to the cloud center.

[0050] In this embodiment, the laser radar performs a large-scale scan of the initial stage of the tunnel to obtain multi-phase high-precision three-dimensional point cloud data about the initial branch structure of the tunnel, and sequentially performs filtering and noise reduction, tunnel point cloud data interpolation, tunnel point cloud data simplification and tunnel point cloud posture adjustment on the tunnel three-dimensional point cloud data to obtain tunnel point cloud data whose direction is parallel to a certain coordinate axis of the Cartesian coordinate system; and analyzes the coordinates and reflection intensity of each point, removes data points with large errors, obtains high-quality point cloud data about the initial branch structure of the tunnel, and thus obtains complete tunnel three-dimensional point cloud data, and uses the repeatedly monitored and adopted numerical range as the error range value as the threshold, and uploads the data to the cloud server. In actual operation, at least two and no more than five scans must be performed.

[0051] The measurement and modeling of this application are mainly carried out through the Cartesian coordinate system, that is, the three-dimensional user coordinate system (UCS), which satisfies the right-hand rule. For example, it is stipulated that the tunnel section profile is parallel to the XOY plane of the Cartesian coordinate system, perpendicular to the Z axis and along the positive direction of the Z axis of the Cartesian coordinate system. The Delaunay triangulation algorithm is preferably used for modeling. The main steps include:

[0052] (1) First, monitor all actual edges of the tunnel and generate boundary point cloud data (calculate the bounding box of the area, generate two super triangles as the initial triangular mesh, and generate boundary points based on the boundary curve). Then, traverse all faces of the tunnel to collect the coordinates of its boundary points in the parameter area, where the parameter area is the XOZ plane of the tunnel bottom plane coordinates. Then, generate a Delaunay triangulated mesh on the parameter area according to the above algorithm.

[0053] (2) The internal points and triangles generated by each face in the parameter region are projected into three-dimensional space, added to the triangular mesh, and the tunnel point cloud data is projected onto the XOZ plane for dimensionality reduction. The bounding box of the point cloud data is calculated, and the four vertices of the bounding box are added to P to obtain P'. The following algorithm is used for inserting new point cloud data at each step:

[0054] (a) In the existing triangulation, the set of all triangles whose circumcircles contain vertex v forms a "star polygon" (the line connecting any vertex of the polygon to v is inside the polygon).

[0055] (b) For the star-shaped polygon above, delete all the triangles inside it to form a "hole". Connect the vertices of the hole boundary with the newly added vertex v to obtain a new triangle to replace the deleted triangle in the triangulation. At this time, a new Delaunay triangulation containing vertex v is obtained.

[0056] (c) When the last vertex is inserted into the triangulation and all operations are completed, a Delaunay triangulation of the point set P' is obtained. Now delete the four vertices of the bounding box added in the first step and remove all the triangles connected to them. The remaining triangles constitute the Delaunay triangulation T of the point set P, and the Delaunay triangulated mesh of all boundary points is obtained.

[0057] (d) For the triangles inside the region, according to certain conditions, the centroids are inserted into the triangular mesh and adjusted according to the Bowyer-Watson algorithm to obtain a new Delaunay triangulation. This process continues until there are no more centroids to be inserted, at which point a complete Delaunay triangulated mesh of the region is obtained.

[0058] After repeated calculations and rereadings, it is determined that each point on the surface has two sides. A local number is used for the triangular mesh in the parameter area and a global number is used for the overall mesh in the three-dimensional space. The final clustering result is obtained and a complete tunnel model is generated in the Cartesian coordinate system.

[0059] It should be noted that the rock mass of the tunnel excavation face in this embodiment is not completely on the same plane, so it is not possible to directly use the plane projection points for triangulation to reconstruct the three-dimensional curved surface of the excavation face. According to the characteristic of the three-dimensional laser scanner that emits lasers from the scanning center to the surroundings, spherical projection is first used, followed by triangulation. According to the mapping relationship, a triangular mesh model of the spatial excavation surface can be established. The triangular mesh model is obtained by surface reconstruction, and the characteristic parameters of each triangular facet are statistically analyzed to find triangular faces with the same or similar parameters to form a structural surface, thereby identifying the structural surface and using the three-dimensional laser scanner to obtain the roughness parameters of the rock mass structural surface. Finally, the relevant geological parameters of the rock mass are obtained by using the RANSAC algorithm and vector algebra method, thereby extracting the identification of the tunnel rock mass structural surface.

[0060] Step 2: Data comparison;

[0061] After a preset time interval, repeat the above data collection process to generate a second point cloud data matrix; extract the position value of the radar in the first point cloud data matrix and compare it with the installation position value of the radar in the second point cloud data matrix to obtain the deformation matrix of the radar in the tunnel as the first deformation matrix;

[0062] Then, the first point cloud data matrix is ​​extracted and compared with the second point cloud data matrix, and the deformation matrix of the tunnel is obtained as the second deformation matrix;

[0063] Step 3: Deformation analysis;

[0064] Calculate the difference between the first deformation matrix and the second deformation matrix, and check whether the absolute value of each element in the deformation matrix exceeds the set threshold; if so, execute the subsequent steps;

[0065] It should be noted here that the radar sensor in the present invention is directly installed on the inner wall of the tunnel for effective monitoring. However, due to the influence of actual road surface collapse and tunnel side wall collapse, some radar equipment will be displaced, resulting in inaccurate monitoring data. Therefore, multiple calibration test points of radar sensors are set in the tunnel, that is, a sample set of images / point clouds in a given area A. The next point is the farthest point p, which must satisfy the following:

[0066]

[0067] This principle can be explained geometrically as follows: the next sampling point is the center of the largest empty circle / sphere located within the image / point cloud boundary, which is characterized by a correlation function that decreases with decreasing distance. In this case, intuitively, the point farthest from the sample set is the point with the highest information content,

[0068] The given data sample is:

[0069]

[0070] Specify sampling point N:

[0071] Initialize the set S t ={s1}, randomly select a point and delete this point from the original set;

[0072] Calculate the set S t The distance measurement set with the S set point is obtained t The minimum distance corresponding to the points in ;

[0073] Calculate S t The maximum value of the minimum distance corresponding to the point, and add this point to S t , delete this point in S;

[0074] Repeat the above steps until the number of St points reaches N.

[0075] To simulate the detection of the tunnel, when the distance between the radar and the corresponding 2-3 reflective detection points exceeds the threshold, an alarm needs to be activated;

[0076] Step 4: Detailed inspection;

[0077] The drone is launched for two targeted flights. In the first flight, the drone will be equipped with a lidar to perform a detailed scan of the target tunnel slope to obtain the third point cloud data matrix. Then, in the second flight, the drone will be replaced with an orthophoto camera to photograph the same area to build a new local or overall radar 3D model and orthophoto model.

[0078] Step 5: Filtering and modeling;

[0079] The filtering and denoising algorithm is used to process the data for overall limit intrusion detection and initial local limit intrusion detection. The results are uploaded to the cloud center for processing and displayed through the terminal.

[0080] The filtering and denoising algorithm processing specifically includes the following steps:

[0081] Step 51: region segmentation: using the RANSAC algorithm to accurately segment the tunnel region in the radar 3D model to eliminate non-tunnel regions that may cause interference;

[0082] Step 52: calculation of difference point cloud; Based on the point cloud data obtained by the two radar scans, this embodiment will calculate the distance between them, so as to identify the difference point cloud. In this process, according to the actual needs and accuracy requirements, cloud difference algorithms such as C2C algorithm and M3C2 algorithm can be selected for calculation. Since the geographic reference information of the two scans is consistent, the point cloud data has been automatically aligned, and this embodiment only needs to directly calculate the distance between the two point clouds.

[0083] Step 53: outlier filtering;

[0084] In order to further improve the accuracy of the data, this embodiment uses the SOR outlier filtering algorithm to remove outlier noise in the difference point cloud. This algorithm can effectively remove the random high-frequency noise generated during the radar point cloud acquisition process, while retaining low-frequency noise and real difference information.

[0085] Step 54: Cluster analysis and deformation area identification;

[0086] Despite the processing of the SOR outlier removal algorithm, there may still be some low-frequency noise in the difference point cloud. In addition, since the point cloud data is stored in a one-dimensional linear manner, the real difference area cannot be directly displayed on the two-dimensional plane. In order to solve these problems, we will use a clustering algorithm to further process the difference point cloud. Specifically, we first use the RANSAC algorithm to automatically classify the difference point cloud; then, according to the definition of small area noise set based on actual segmentation experience, filter out the group with very small area and the group with the largest area (where the small area group is regarded as noise and the large area group is regarded as the background without deformation); finally, output the remaining deformation area point cloud, that is, the tunnel deformation area, and construct the final three-dimensional model based on it, and frame the location of the tunnel deformation area.

[0087] After the above processing is completed, the analysis data will be uploaded to the cloud center for further processing and analysis. The cloud center will integrate and optimize the data and generate intuitive image results, which will then be sent to the terminal for display. In this way, the deformation of the tunnel can be intuitively understood, and corresponding repair measures can be taken in time to ensure the safe operation of the tunnel.

[0088] Generally, the deformation area is mapped to the orthophoto map for intuitive display through the radar 3D model to orthophoto model mapping algorithm. The goal of mapping the deformation area from the radar 3D model to the orthophoto model is to obtain an intuitive feeling of the deformation area in the orthophoto model, so the latitude and longitude accuracy requirements are not high. The main error of the two types of models comes from the elevation error of the orthophoto model (usually in the order of tens of meters). To eliminate this error, it is necessary to calculate the elevation of the final 3D model and the orthophoto model respectively, and then subtract the elevation difference from the elevation of the deformation area to obtain the deformation area on the orthophoto model.

[0089] A key monitoring area is set up on the tunnel vault, monitoring points are set up in the key monitoring area, and supplementary light sources are added to the structured light scanner to supplement the brightness of the measuring points.

[0090] Line structured light is projected onto the top of the tunnel, and the deformed fringes in the monitoring area on the top of the tunnel are collected by a camera, and the key monitoring area is obtained through image preprocessing.

[0091] The Hessian matrix is ​​used to solve the vector characteristics of the laser line, and the sub-pixel center point of the laser line is extracted using Gaussian curve fitting correction in the normal direction. The point chain topological relationship is constructed through the KD tree to establish connections between the originally independent points.

[0092] The light plane equation is matched with the polar line constraint to perform three-dimensional reconstruction, and the structural laser is used to collect distance information for fusion to obtain millimeter-level deformation monitoring data. During the monitoring process, a time interval is set, and the result of each scan reconstruction is compared with the previous scan result to determine whether deformation has occurred.

[0093] Under the same detection conditions, data collection is carried out, and the distance between measuring stations is the main factor affecting the detection accuracy. Before the data collection step, the distance measurement accuracy experiment is first carried out based on two different scanning modes of the three-dimensional laser scanner, ultra high and super high. That is, the distance measurement is set to 5m, 10m, 15m, 20m, 25m, 30m, 35m, 40m, 45m, 50m, 55m, and 60m for twelve groups of different distances to conduct experiments in order to obtain a reasonable distance between measuring stations. The specific experimental method is: set an initial point outside the tunnel, place prisms at the initial point and twelve groups of different distance measurement distances in the tunnel, fix the total station at the initial point and observe, and record the distance from the prism to the total station at different distance measurement points. Then replace the prism of each measuring point with the target ball of the laser scanner, set up a three-dimensional laser scanner at the original initial point, scan using two different resolution modes of the three-dimensional laser scanner, obtain the coordinates of the center point of the target ball after software processing of the scanning results, and calculate the distance from the center point of the target ball to the three-dimensional laser scanner. Compare the distance from the 3D laser scanner to the target sphere and the distance from the prism to the total station, and calculate the mean error of each measurement distance under different resolution modes according to the mean error formula. Mean error, also known as root mean square error, is a data standard for measuring observation accuracy. The mean error calculation formula is as follows.

[0094]

[0095] Among them, RMSD is the mean square error, is the residual value, and n is the number of times.

[0096] When the scanner uses ultra high mode, the scanning time is slightly longer than super high, but the ranging accuracy is higher, so it is advisable to use ultra high resolution for scanning and measurement. At the same resolution, when the ranging is within 20m, the ranging accuracy of the 3D laser scanner is higher, and the mean error is less than 3mm. This error range meets the accuracy requirements of railway point cloud data processing, so the Z+F IMAGER 5010C ground 3D laser scanner can be used to implement tunnel point cloud acquisition.

[0097] The process of overall limit intrusion detection and analysis in this embodiment includes:

[0098] (1) The slice point cloud data is clipped in the elevation direction, and the slice projection of the clipped segmented contour is fitted with a circle by RANSAC or least squares method;

[0099] Get the center coordinates and radius R parameters of the fitted circle;

[0100] Compare and analyze the obtained parameters such as the center coordinates and radius R with the tunnel design parameters to determine whether the initial support of the tunnel has deformed and encroached on the limit.

[0101] The process of detecting and analyzing the local encroachment of the initial support includes: according to the target area where the initial support has overall encroachment, establish supervoxels for the sliced point cloud data of the target area through Octree, calculate the curvature of each voxel within the neighborhood range, and determine the local encroachment position of the initial support by comparing the difference in curvature within the neighborhood. Among them, project the tunnel slices, and divide the tunnel point cloud data into equally spaced slices along the direction parallel to the tunnel central axis. Perform a dimensionality reduction projection of each slice towards the slice midplane, replace the thick slice with the slice midplane without thickness, and retain the sliced point cloud data at the same time.

[0102] In this embodiment, re-slice the tunnel point cloud, set the slice step size to the side length Lstep2 of the unit to be generated, and the direction of each slice is along the tangent direction of the two-dimensional horizontal midline of the slice, and the slice thickness is taken as d.

[0103] However, for any cross-section, the points within the cross-section are discrete. At this time, set the thickness d of the cross-section slice. Then, the point cloud within the range of d / 2 before and after the position of the point on the X-axis of the tunnel is within this cross-section, that is, when the cross-section intercepts the thickness d, the point set of the point cloud of this cross-section satisfies the following equation, the point is a point passing through the tunnel central axis.

[0104]

[0105] Among them, 0 < d ≤ Lstep2. The slice thickness d should not be too small. If the slice thickness d is too small, the amount of sliced point data will be too small. Perform a dimensionality reduction projection of each slice towards the slice midplane, that is, convert the Z coordinates of all points of each slice to the Z coordinate zi of the slice center. Then the Z-direction coordinate range of the slice is (zi - d / 2, zi + d / 2).

[0106] (2) First, perform the detection of the overall encroachment of the initial support. Since highway tunnels mostly present as three-centered circles or five-centered circles, first crop the obtained tunnel point cloud slice data in the elevation direction and perform circle fitting respectively. Perform circle fitting on the sliced projection of the cropped segmented contour through RANSAC or the least squares method, etc., to obtain parameters such as the center coordinates and radius R of the fitted circle. Compare and analyze the obtained parameters such as the center coordinates and radius R with the tunnel design parameters, and then it can be determined whether the initial support of the tunnel has deformed and encroached on the limit.

[0107] Specifically, a circle is fitted on the segment of the slice projection, and the fitting radius R is obtained. The design radius parameter of the primary branch of this part of the tunnel is Rs, and the reserved deformation is d. If the fitting radius R is smaller than the difference between the design radius Rs and the reserved deformation d, it means that the initial branch section of the tunnel has an overall deformation intrusion; if the fitting radius R is larger than the difference between the design radius Rs and the reserved deformation d, it means that the initial branch section of the tunnel has not undergone an overall deformation intrusion at this time. Multi-phase tunnel initial branch point cloud data can be collected to realize deformation intrusion monitoring of the tunnel initial branch and analyze its deformation rate.

[0108] (3) Then, the local intrusion detection of the initial branch is performed. After obtaining the target area where the initial branch has overall intrusion, the supervoxel is established for the point cloud of the target area through Octree, and the curvature of each voxel in the neighborhood is calculated. By comparing the difference in curvature within the neighborhood, the local intrusion detection of the initial branch is achieved.

[0109] Specifically, by establishing voxels, the cross-section of the primary branch of the tunnel is downsampled, so that the radius of curvature of each voxel can be calculated through the center of the circle fitted by the tunnel cross-section, and then the radius of curvature of each voxel in the neighborhood of the voxel is compared to obtain its curvature. By setting the curvature change threshold, the area with obvious curvature change is obtained, which is the area where the local encroachment of the primary branch of the tunnel occurs.

[0110] (4) After the overall and local intrusion of the primary support is detected, it is necessary to monitor or deal with it in time according to the severity of the intrusion. The monitoring of the deformation rate of the intrusion mainly includes: based on the multi-phase high-precision three-dimensional point cloud data of the tunnel primary support structure, voxels are established through Octree for each phase of the primary support point cloud data, and each voxel is encoded. After the multi-phase primary support point cloud data is processed in this way, the voxels with the same encoding are compared and analyzed, and the spatial coordinate difference of the center of gravity of the point cloud voxels is calculated. The difference is the intrusion deformation value. According to the ratio of the intrusion deformation value to the intrusion deformation time, the intrusion deformation rate is obtained.

[0111] Embodiment three:

[0112] This embodiment provides a tunnel deformation monitoring system, which is used to implement the tunnel deformation monitoring method of embodiment 1 or embodiment 2.

[0113] The system of this embodiment includes a computer, a drone, an information and data acquisition module, a power supply and power management module, a data exchange module, a black box, a drone and an audible and visual alarm device.

[0114] The computer controls the information and data acquisition sensors to collect and process data to determine the safety level of the current environment; the information and data acquisition sensors include laser radar and structured light scanner; the output control interface of the computer is connected to the sound and light alarm device to feedback the system warning level signal.

[0115] The power supply and power management module provides power and power management for sensors and functional devices; the power supply and power management module has a backup power supply for emergency use.

[0116] The data exchange module exchanges the data that the system needs to exchange with the operator network or local area network through the network port.

[0117] The black box stores and records key data information.

[0118] The drone is connected to the computer via a wireless link to exchange data. After a disaster occurs, the drone can approach the dangerous tunnel section to take pictures, record and transmit data. It can conduct unmanned real-time monitoring in the tunnel, and the drone can conduct close exploration to avoid casualties caused by secondary disasters.

[0119] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A tunnel deformation monitoring method, characterized in that: The method comprises: Step 1: Use laser radar to collect three-dimensional point cloud data of the tunnel and record the radar measurement position to generate the first point cloud data matrix; Step 2: After a preset time interval, repeat the above data collection process to generate a second point cloud data matrix; extract the position value of the radar in the first point cloud data matrix and compare it with the installation position value of the radar in the second point cloud data matrix, and obtain the deformation matrix of the radar in the tunnel as the first deformation matrix; extract the first point cloud data matrix and compare it with the second point cloud data matrix, and obtain the deformation matrix of the tunnel as the second deformation matrix; Step 3: Calculate the difference between the first deformation matrix and the second deformation matrix, check whether the absolute value of each element in the deformation matrix exceeds a set threshold, and determine whether there is a deformation area; Step 4: If there is a deformed area, use the drone to fly twice, mount the lidar and orthophoto camera respectively, scan the same tunnel slope, obtain the third point cloud data, and construct the radar 3D model and orthophoto model; Step 5: Remove the false deformation area and locate the potential tunnel deformation area through filtering and denoising algorithm; Step 6: Mapping the tunnel deformation area to the orthophoto model for intuitive display through a radar three-dimensional model to orthophoto model mapping algorithm; Step 7: Based on the three-dimensional point cloud data, perform overall intrusion detection and initial local intrusion detection. The results are uploaded to the cloud center for processing and displayed through the terminal.

2. The tunnel deformation monitoring method according to claim 1, characterized in that: The filtering and denoising algorithm in step 5 includes: Step 51: segment the tunnel in the radar 3D model and remove non-tunnel areas; Step 52: Calculate the distance between the two phases of radar point clouds to obtain a difference point cloud; Step 53: removing outlier points from the difference point cloud; Step 54: marking the tunnel deformation area in the three-dimensional model.

3. The tunnel deformation monitoring method according to claim 1, characterized in that: The overall intrusion detection analysis cuts and slices the point cloud data, uses RANSAC or the least squares method to perform circle fitting, obtains the coordinates of the center of the circle and the radius parameters, and compares them with the tunnel design parameters to determine whether the primary support is deformed and intrudes.

4. The tunnel deformation monitoring method according to claim 1, characterized in that: The process of the local intrusion detection and analysis of the initial branch includes: analyzing the supervoxels of the slice point cloud data established by Octree, calculating the voxel neighborhood curvature, and comparing the curvature difference to determine the local intrusion position.

5. The tunnel deformation monitoring method according to claim 1, characterized in that: The method further includes: providing a key monitoring area on the tunnel vault, setting monitoring points in the key monitoring area, and adding a supplementary light source to the structured light scanner to supplement the brightness of the measuring points.

6. A tunnel deformation monitoring system, characterized in that: Used to implement the tunnel deformation monitoring method according to any one of claims 1 to 5, the system comprises: a computer, a drone, an information and data acquisition module, a power supply and power management module, and a data exchange module; the information and data acquisition module comprises a laser radar and a structured light scanner; The computer controls the operation of the drone and controls the information and data acquisition sensor to collect and process data to determine the safety level of the current environment; the output control interface of the computer is connected to the sound and light alarm device; the computer exchanges data with the information and data acquisition module through the data exchange module.

7. The tunnel deformation monitoring system according to claim 6, characterized in that: The power supply and power management module includes a backup power supply for emergency use.

8. The tunnel deformation monitoring system according to claim 6, characterized in that: The system also includes a black box for storing and recording key data information.

9. A tunnel deformation monitoring device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is used to implement the tunnel deformation monitoring method according to any one of claims 1 to 5 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the tunnel deformation monitoring method according to any one of claims 1 to 5 is implemented.

Images