Deformation Analysis Method of Initial Support for Shotcrete and Anchor in Drill-and-Blast Tunnel Based on LiDAR and Improved CSF Algorithm

By developing a LiDAR acquisition system adapted to drill-and-blast tunnels and improving the CSF algorithm, combining RTK control points for natural registration and local surface density estimation, and designing the GDef algorithm for full-section three-dimensional deformation quantification, the problem of high-precision automation in monitoring the deformation of the initial support of the shotcrete in drill-and-blast tunnels was solved, achieving a segmentation accuracy of 96.12% and a deformation detection error of less than 2.5 mm.

CN123116408APending Publication Date: 2026-11-17CHINA RAILWAY TUNNEL SURVEY DESIGN & RES INST CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611348409.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies cannot achieve high-precision, automated deformation monitoring of shotcrete and anchor supports in drill-and-blast tunnels. Especially in environments with high humidity, high dust, and strong vibration, the accuracy of point cloud acquisition and ROI extraction is insufficient, and the deformation quantification algorithm loses local details, failing to meet the safety requirements of tunnel construction.

Method used

Develop a LiDAR acquisition system adapted to drill-and-blast tunnels, combine it with RTK control points for natural registration, improve the CSF algorithm to extract the initial support cloud of shotcrete anchors through local surface density estimation and virtual settlement, design the GDef algorithm for full-section three-dimensional deformation quantification, and construct an integrated detection framework.

Benefits of technology

It achieves high-precision point cloud acquisition and automated ROI extraction in harsh environments, with a segmentation accuracy of 96.12% and a deformation detection error of less than 2.5mm, meeting the engineering monitoring needs of drill-and-blast tunnels and providing accurate deformation data support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116408A_ABST
    Figure CN123116408A_ABST
Patent Text Reader

Abstract

This invention discloses a method for analyzing the deformation of the initial support of shotcrete in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm. The method includes: acquiring and registering multi-temporal point clouds of the drill-and-blast tunnel; extracting the initial support point cloud of shotcrete based on the improved CSF algorithm; and quantifying the three-dimensional deformation of the initial support of shotcrete based on the GDef deformation algorithm. This includes designing the GDef deformation algorithm, using early initial support point clouds from multiple temporal sequences as a reference and later initial support point clouds as the detection object. By extracting cross-sectional feature points and calculating point cloud deviations, the algorithm achieves three-dimensional deformation quantification of the entire cross-section of the initial support, preserving local millimeter-level deformation details. The input of the GDef deformation algorithm is the multi-temporal initial support point cloud, including a reference initial support point cloud and a detection initial support point cloud, and the output is a set of initial support point clouds with deformation values. The deformation detection results are verified and analyzed. The invention also discloses a system, apparatus, electronic equipment, and a computer-readable storage medium.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of tunnel construction safety monitoring technology, specifically to a method and system for analyzing the initial support deformation of shotcrete and anchor in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm. Specifically, it relates to a method for analyzing the initial support deformation of shotcrete and anchor in drill-and-blast tunnels, applicable to high-precision monitoring of full-section deformation in drill-and-blast construction scenarios such as mountainous highway tunnels. Background Technology

[0002] Drill and blast method is the mainstream excavation technology for highway tunnels in mountainous areas. Shotcrete and anchor support, as the core structure of the New Austrian Tunneling Method (NATM), plays a crucial role in transferring stress in the surrounding rock and preventing tunnel collapse. However, due to factors such as blasting disturbance and complex geological conditions (uneven thrust in soft rock), shotcrete and anchor support inevitably deforms. If this deformation is not monitored accurately and in a timely manner, it will exceed the safety limits of the secondary support, leading to a major collapse accident. Therefore, deformation monitoring of shotcrete and anchor support is a core aspect of tunnel construction safety.

[0003] Current tunnel deformation monitoring technology is trending towards a shift from single-point contact monitoring to three-dimensional non-contact monitoring. Traditional monitoring methods rely primarily on equipment such as total stations, levels, and measuring rods, which can only acquire discrete single-point displacement data. They cannot capture the deformation characteristics of the entire cross-section, resulting in poor data representativeness and difficulty in detecting large-scale, non-uniform initial support deformation, leading to significant blind spots in safety monitoring. 3D laser scanning (LiDAR / Terrestrial Laser Scanning TLS) technology, with its advantages of automation, high precision, and large information capacity, has become a new technological direction for tunnel deformation monitoring and has been widely applied in fields such as autonomous driving, geological surveying, and engineering inspection. However, it still faces key technological bottlenecks in drill-and-blast tunnel monitoring.

[0004] The construction environment of drill-and-blast tunnels is characterized by enclosed spaces, dense machinery and equipment, severe obstruction, and high dust / vibration levels. This results in the raw point cloud acquired by LiDAR containing a large amount of data related to non-ROIs (Regions of Interest), such as construction equipment, personnel, and pipelines. These non-ROIs severely interfere with the segmentation and deformation analysis of the initial support structure. Existing point cloud ROI extraction methods suffer from extremely low efficiency due to manual cropping, poor detail retention due to reliance on complex prior knowledge, and high computational costs due to the need for deep learning semantic segmentation. None of these methods are suitable for the engineering requirements of drill-and-blast tunnels. Furthermore, while the existing Bouguer Simulation Filter (CSF) algorithm has been applied to the monitoring of initial supports in subway tunnels, it requires the use of cylindrical projection to unfold the point cloud. However, due to construction interference in drill-and-blast tunnels, the cylindrical fitting center is difficult to locate accurately, rendering this method unsuitable. Additionally, traditional tunnel deformation calculation methods rely on elliptical / circular geometric fitting, which easily loses local deformation details and cannot achieve millimeter-level full-section 3D deformation quantification.

[0005] The existing technologies closest to this invention can be mainly divided into three categories: LiDAR tunnel point cloud acquisition technology, CSF algorithm tunnel application technology, and tunnel deformation quantization calculation technology. These three together constitute the basis for the improvement of this invention, and the specific solutions are as follows:

[0006] LiDAR tunnel point cloud acquisition technology: Existing technologies use commercial TLS equipment (such as RIEGLVZ400) to acquire tunnel point clouds. By registering with the design surface, tunnel over-excavation and under-excavation can be identified, achieving a cross-sectional monitoring accuracy of about 4mm. However, this equipment has not been adapted and optimized for the high humidity, dust, and strong vibration environment of drill-and-blast tunnels. Furthermore, multi-stage point cloud registration relies on traditional methods, and the problem of filtering non-ROI data has not been solved, directly affecting subsequent deformation analysis.

[0007] CSF algorithm application technology in tunnels: Existing technology combines the CSF algorithm with moving laser scanning (MLS) for deformation monitoring in subway tunnels. The RANSAC algorithm is used to determine the center of the tunnel cross-section, and cylindrical projection is used to unfold the 3D point cloud into a 2D plane before extracting the ROI, thus achieving deformation monitoring of the initial support of the subway tunnel. However, this scheme relies on the regular geometric model of cylindrical projection. Subway tunnel construction environments are regular, and auxiliary facilities are fixed in distribution. In contrast, drill-and-blast tunnel construction involves significant interference and uneven point cloud structure, making it difficult to locate the projection center. This significantly reduces the adaptability and extraction accuracy of the method.

[0008] Tunnel deformation quantification calculation technology: Existing technologies are mainly divided into two categories. One is to use the least squares method and RANSAC algorithm to fit the tunnel cross section into a standard geometric shape such as an ellipse / circle, so as to achieve a cross section detection accuracy of about 1.5mm for subway tunnels. However, this method has low efficiency in processing large-scale point clouds and is prone to losing local deformation details. The other is to use the minimum distance projection algorithm to analyze the spatiotemporal displacement characteristics of the fitted curve. However, it cannot adapt to the non-uniform multi-directional deformation of the initial support of the shotcrete anchor in the drill-and-blast method tunnel, and the deformation quantification accuracy is insufficient.

[0009] In summary, existing technologies have achieved tunnel acquisition of LiDAR point clouds, extraction of tunnel ROI using CSF algorithm, and basic quantification of tunnel deformation, respectively. However, they have not been integrated and improved for the complex environment of drill-and-blast tunnels, and have not yet formed a complete technical system of "precise point cloud acquisition → automatic ROI extraction → millimeter-level quantification of full-section deformation". Therefore, they cannot meet the engineering requirements for monitoring the deformation of the initial support of shotcrete in drill-and-blast tunnels.

[0010] The existing technology has four major objective technical defects, as follows:

[0011] 1. Poor adaptability of point cloud acquisition system: Commercial TLS equipment has not been protected and hardware optimized for the harsh environment of high humidity, dust and strong vibration in drill-and-blast tunnels. Moreover, the multi-stage point cloud registration method is cumbersome and the natural registration accuracy is insufficient, making it impossible to achieve high-precision multi-time series point cloud acquisition and alignment.

[0012] 2. Insufficient tunnel adaptability of CSF algorithm: The existing CSF application scheme for subway tunnels relies on cylindrical projection and preset geometric models, which cannot be applied to the construction interference and point cloud occlusion problems of drill-and-blast tunnels, resulting in low ROI extraction accuracy and poor robustness.

[0013] 3. There are shortcomings in point cloud ROI extraction methods: manual cropping is extremely inefficient, traditional filtering methods rely on prior knowledge and have poor detail retention, and deep learning segmentation requires a large amount of labeled data and is difficult to implement in engineering. None of these methods can achieve automatic and efficient extraction of ROI for shotcrete anchor initial support under complex point clouds in drill-and-blast tunnels.

[0014] 4. Insufficient accuracy and detail in deformation quantization algorithms: Traditional geometric fitting methods lose the local deformation details of the initial support of the shotcrete anchor, the minimum distance projection algorithm cannot adapt to non-uniform deformation, and there is a lack of full-section 3D deformation quantization algorithms for drill-and-blast tunnels, making it difficult to achieve millimeter-level high-precision deformation detection.

[0015] Therefore, there is an urgent need to develop a LiDAR point cloud deformation detection technology that is adapted to the complex construction environment of drill-and-blast tunnels. By improving the CSF algorithm, we can achieve automatic and high-precision extraction of the ROI of the initial support of the shotcrete and anchor, and design a dedicated deformation quantification algorithm to achieve millimeter-level full-section deformation detection. This will provide a precise and efficient monitoring solution for the safety of drill-and-blast tunnel construction, and it is also an important development direction for intelligent monitoring of tunnel engineering. Summary of the Invention

[0016] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method and system for analyzing the deformation of initial support of shotcrete anchors in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm, solving the following four core technical problems:

[0017] 1. How to develop a self-developed TLS acquisition system adapted to the harsh environment of drill-and-blast tunnels, and design a point cloud natural registration method based on the tunnel control network to achieve high-precision and automated acquisition and alignment of point clouds in multiple phases;

[0018] 2. How to improve the CSF algorithm, eliminate cylindrical projection and preset geometric model, adapt to the complex point cloud environment of drill-and-blast tunnels, and achieve automatic and high-precision extraction of the ROI of shotcrete and anchor initial support;

[0019] 3. How to design a dedicated GDef deformation quantization algorithm to break through the limitations of traditional geometric fitting, realize full-section 3D deformation calculation of the initial support of the shotcrete anchor in the drill-blast method tunnel, retain local deformation details and control the detection error to the millimeter level;

[0020] 4. How to optimize the key parameters of the CSF algorithm (Bouguer mesh resolution, rigidity, distance between Bouguer and initial support) to achieve a balance between ROI extraction accuracy and computational efficiency, and meet the actual application needs of engineering sites.

[0021] The first aspect of this invention is to provide a method for analyzing the deformation of the initial support of the shotcrete anchor in the drill-and-blast tunnel based on LiDAR and an improved CSF algorithm, comprising:

[0022] S1, Acquisition and registration of multi-temporal point clouds of drill-and-blast tunnels;

[0023] S2, Extraction of shotcrete initial support point cloud based on improved CSF algorithm, including: Based on the curved geometric features of shotcrete initial support in drill-and-blast tunnel, local surface density estimation is introduced to optimize CSF algorithm parameters, and the automatic extraction of shotcrete initial support point cloud in the initial support area is achieved by simulating the natural settlement of virtual fabric under gravity.

[0024] S3, quantifying the three-dimensional deformation of the initial support of shotcrete based on the GDef deformation algorithm, including: designing the GDef deformation algorithm, using the early initial support point cloud in multiple time series as a reference and the later initial support point cloud as the detection object, and realizing the three-dimensional deformation quantification of the entire section of the initial support by extracting cross-sectional feature points and calculating point cloud deviations, while preserving local millimeter-level deformation details. The input of the GDef deformation algorithm is the initial support point cloud in multiple time series, including the reference initial support point cloud. and detection of initial fulcrum cloud The output is a set of initial support point clouds with deformation.

[0025] S4. Verify and analyze the deformation detection results.

[0026] Preferably, S1 includes:

[0027] S11, A LiDAR point cloud acquisition system specifically designed for tunnel structure construction is set up for 3D information acquisition, including a laser scanning device, an industrial camera, protective equipment, and an edge computing industrial control computer; the protective equipment is used to protect the laser scanning device from dust, shock, and water, and the edge computing industrial control computer is used to realize the rapid processing of on-site point cloud data, adapting to the harsh environment of tunnel structure construction;

[0028] S12, High-precision acquisition of multi-time series point clouds, including: setting up RTK control points with known WGS-84 coordinates inside the tunnel; using the LiDAR point cloud acquisition system specifically designed for tunnel structure construction, performing laser scanning on the initial support area of ​​the shotcrete in the tunnel at the same control point at different time points during tunnel construction, acquiring massive amounts of three-dimensional point cloud data, the point cloud data containing spatial coordinates. Information reflects the geometric shape of the initial branch;

[0029] S13 performs registration and cropping of multi-temporal point clouds.

[0030] Preferably, S13 includes:

[0031] The resection method was adopted, and the spatial location of the TLS station was determined by measuring the RTK control points with a total station. The TLS local coordinate system of the point cloud was converted into the WGS-84 geographic coordinate system to achieve natural registration of multi-time series point clouds. The coordinate transformation corresponding to the conversion of the TLS local coordinate system of the point cloud into the WGS-84 geographic coordinate system adopts the Bursa-Wolf seven-parameter transformation model, and the formula is shown in the following formula (1):

[0032] (1);

[0033] in, Coordinates in the WGS-84 coordinate system. Coordinates in the TLS local coordinate system; , and For translation parameters, , and Rotational parameters, unit: rad. Scale factor;

[0034] The root mean square error (RMSE) is used to evaluate the point cloud registration accuracy, as shown in the following formula (2):

[0035] (2);

[0036] in, To register the number of control points, and These are the coordinates of the control points before and after registration;

[0037] The registered original point cloud was cropped to remove the secondary initial support and non-monitored areas outside the tunnel entrance, retaining the local tunnel point cloud including the shotcrete initial support. The cropping method was spatial cube cropping, and the boundary coordinates of the cropping area were defined. Point clouds that satisfy the following conditions as shown in equation (3) shall be retained:

[0038] (3).

[0039] Preferably, S2 includes:

[0040] S21, determine the improved CSF algorithm;

[0041] S22, Four-stage simulation of the extraction of initial support for shotcrete;

[0042] S23, Parameter optimization of the improved CSF algorithm, including: optimizing the core parameters of the simulated filtering through single-factor analysis experiments, and determining the optimal parameter combination for extracting the initial support of the tunnel using the drill-and-blast method. The optimal parameter combination includes:

[0043] The mesh resolution is 6mm; the fabric stiffness is 1; the distance between the fabric and the initial support surface is 1.5mm.

[0044] The optimization process of the core parameters adopts a comprehensive evaluation index. and calculation time Construct the objective function This achieves multi-objective optimization, as shown in equation (9):

[0045] (9);

[0046] in, , These are the weighting coefficients. for , To maximize computation time, in this embodiment, the optimized CSF algorithm achieves a segmentation accuracy of 96.12% for the initial shotcrete support. The segmentation accuracy is determined using the intersection-union ratio (IUU). The evaluation is shown in equation (10):

[0047] (10);

[0048] in, The correct number of initial fulcrums to be extracted. The number of non-initial pivots that were mistakenly extracted. This represents the number of initial fulcrums that were not extracted.

[0049] Preferably, S21 includes:

[0050] S211 utilizes the spatial characteristics of the non-RoI of the initial support of the drill-and-blast tunnel located inside the initial support, allowing the virtual cloth to settle naturally in the actual coordinate system without the need for tunnel point cloud inversion.

[0051] S212, Estimating Local Surface Density: Calculating the Surface Density of Tunnel Point Clouds , which represents the number of neighboring points in the spherical region centered at each point / the surface area of ​​the spherical region, as shown in the following formula (4):

[0052] (4);

[0053] in, For surface density, Let be the number of neighboring points within the spherical region. The initial mesh size of the virtual cloth is determined based on the surface density, where the radius of the spherical region is given.

[0054] S213, Equation of motion of virtual cloth particle: Considering the combined effect of gravity and spring force, the equation of motion of virtual cloth particle is as follows (5):

[0055] (5);

[0056] in, For particle mass, For the particle position vector, It is the gravitational acceleration vector. This is the spring force between adjacent particles. The collision force between particles and point clouds. The number of adjacent particles;

[0057] S214, Calculate the spring force using Hooke's Law. As shown in equation (6):

[0058] (6);

[0059] in, For spring stiffness, For spring deformation, This represents the position vector between adjacent particles. This is the original length of the spring;

[0060] S215, the collision force is calculated using an elastic collision model. As shown in equation (7):

[0061] (7);

[0062] in, For collision stiffness, The penetration depth of particles and point clouds. is the normal vector of the collision contact surface.

[0063] Preferably, S22 includes:

[0064] Initializing the virtual cloth includes: generating a virtual cloth mesh composed of interconnected particles above the tunnel model, wherein the particles are uniformly distributed and connected by springs to form a flexible structure that can adapt to the complex geometry of the initial support.

[0065] The virtual cloth undergoes gravity settlement, including: releasing the virtual cloth and allowing it to settle towards the surface of the initial support under simulated gravity; when the cloth particles come into contact with the outer surface of the shotcrete initial support, they are marked as fixed particles and stop moving.

[0066] Implementing elastic relaxation of the virtual cloth includes: using the spring force between adjacent particles to allow the movable particles to continue adjusting their positions until the internal tension of the cloth stabilizes, allowing the cloth to smoothly conform to the initial support surface and eliminating interference caused by local point cloud irregularities and occlusions.

[0067] Extracting the initial support point cloud of shotcrete includes: after the virtual cloth reaches a stable state, its shape is the outer surface contour of the initial support of shotcrete. By judging whether the spatial distance between the original point cloud and the virtual cloth is less than a set threshold, the initial support point cloud of shotcrete is extracted, as shown in Equation (8):

[0068] (8);

[0069] in, For point The shortest distance to the virtual cloth, The distance threshold is denoted as RoI, which is the region of interest, i.e., the initial pivot cloud.

[0070] Preferably, S23 includes:

[0071] S23, Parameter optimization of the improved CSF algorithm, including: optimizing the core parameters of the fabric simulation filter through single-factor analysis experiments, and determining the optimal parameter combination for the initial support extraction of the tunnel using the drill-and-blast method. The optimal parameter combination includes: a fabric grid resolution of 6 mm; a fabric stiffness of 1; and a distance of 1.5 mm between the fabric and the initial support surface.

[0072] The optimization process of the core parameters adopts a comprehensive evaluation index. and calculation time Construct the objective function This achieves multi-objective optimization, as shown in equation (9):

[0073] (9);

[0074] in, , These are the weighting coefficients. for , This represents the maximum computation time.

[0075] Segmentation accuracy uses intersection-union ratio (IU / U). The evaluation is shown in equation (10):

[0076] (10);

[0077] in, The correct number of initial fulcrums to be extracted. The number of non-initial pivots that were mistakenly extracted. This represents the number of initial fulcrums that were not extracted.

[0078] Preferably, S3 includes:

[0079] S31, Define the coordinate system, including: establishing the scanner's right-handed coordinate system, where... The axis points towards the tunnel wall. The axis runs along the tunnel excavation direction. The axis is vertically upward, providing a spatial reference for cross-section extraction and deformation calculation;

[0080] S32, Fitting the cylindrical surface of the reference initial pivot cloud, including: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The cylindrical surface is fitted, and the equation of the cylindrical surface is shown in equation (11) below:

[0081] (11);

[0082] in,( , ( ) represents the coordinates of the cylinder's center. Where is the radius of the cylinder;

[0083] The cylinder parameters are solved using the least squares method, and the objective function is shown in equation (12):

[0084] (12);

[0085] The optimal solution for the cylinder parameters is obtained by solving the above optimization problem using the Lagrange multiplier method.

[0086] S33, Match feature point pairs and calculate the bias, including:

[0087] For each valid reference point of the reference profile surface The Euclidean distance method is used for nearest neighbor search in the detection of initial pivot cloud. Finding distance The 3 most recent points , and As shown in equation (13):

[0088] (13);

[0089] calculate , and centroid Centroid The coordinates are calculated as shown in equation (14):

[0090] (14);

[0091] Calculate the centroid With reference point The spatial distance is considered, and the deformation type is determined by taking into account the normal direction, and the deformation amount is calculated. As shown in equation (15):

[0092] (15);

[0093] in, For reference point The unit normal vector at that location is positive if it points outward from the tunnel. For expansion and deformation, For shrinkage deformation;

[0094] The reference point normal vector is calculated using a local plane fitting method. of The plane is fitted to the neighboring points, and the plane equation is shown in equation (16):

[0095] (16);

[0096] Wherein, the normal vector is The unit normal vector is ;

[0097] S34, statistically analyze the deformation of the entire cross-section and visualize it; S34 includes:

[0098] Statistical analysis was performed on the deformation of all characteristic points of the initial support section, and key parameters were calculated, including: maximum value, minimum value, and average deformation. Standard deviation Root mean square error As shown in equations (17)-(19):

[0099] (17);

[0100] (18);

[0101] (19);

[0102] in, This represents the average deformation. Standard deviation, The root mean square error, This represents the number of feature points;

[0103] The deformation is converted into a color gradient to generate the initial support 3D deformation cloud map. The color mapping function is shown in equation (20) below:

[0104] (20);

[0105] in, The maximum allowable deformation is determined in accordance with the "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019.

[0106] Preferably, S4 includes:

[0107] S41, Verification simulation experiment, including: establishing a simplified CAD model of the initial support of shotcrete in a drill-and-blast tunnel, setting different degrees of radial convergence deformation, and introducing non- The model is then used to simulate scanning errors by adding Gaussian noise with zero mean and a standard deviation of 0.005m. The Gaussian noise generation formula is shown in equation (21) below:

[0108] (twenty one);

[0109] in, =0.005m is the standard deviation of noise;

[0110] S42, Verification field test, including: conducting field tests in a three-lane highway tunnel constructed by actual drill-and-blast method, collecting initial support point clouds at different construction stages, completing initial support extraction and deformation calculation, comparing the deformation detection results with the safe deformation limit of "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019, and judging the stability of the initial support structure; the deformation detection error is evaluated by absolute error and relative error, as shown in equations (22) and (23):

[0111] (twenty two);

[0112] (twenty three);

[0113] in, The actual deformation was measured using a high-precision total station. For absolute error, This is relative error;

[0114] S43, outlier filtering, including: for a small number of misclassified point clouds in the field detection, setting a filtering threshold of 150mm to remove outlier points that are more than the threshold away from the initial support surface of the fitting, as shown in equation (24):

[0115] (twenty four);

[0116] in, The filtering threshold is... For point The distance to the fitted cylindrical surface.

[0117] A second aspect of the present invention is to provide a deformation analysis system for the initial support of a drill-and-blast tunnel using shotcrete and anchor methods based on LiDAR and an improved CSF algorithm, comprising:

[0118] The acquisition and registration module (101) is used to acquire and register multi-temporal point clouds of drill-and-blast tunnels;

[0119] The point cloud extraction module (102) is used to extract the point cloud of the initial support of shotcrete based on the improved CSF algorithm. It includes: based on the curved geometric features of the initial support of shotcrete in the drill-and-blast tunnel, local surface density estimation is introduced to optimize the parameters of the CSF algorithm. By simulating the natural settlement of the virtual fabric under gravity, the automatic extraction of the point cloud of the initial support of shotcrete in the initial support area is realized.

[0120] The three-dimensional deformation quantization module (103) is used to quantify the three-dimensional deformation of the initial support of shotcrete based on the GDef deformation algorithm. This includes: designing the GDef deformation algorithm, using the early initial support point cloud in multiple time series as a reference and the later initial support point cloud as the detection object; extracting cross-sectional feature points and calculating point cloud deviations to achieve three-dimensional deformation quantization of the entire cross-section of the initial support, while preserving local millimeter-level deformation details. The input to the GDef deformation algorithm is the initial support point cloud in multiple time series, including the reference initial support point cloud. and detection of initial fulcrum cloud The output is a set of initial support point clouds with deformation.

[0121] The verification and analysis module (104) is used to verify and analyze the deformation detection results.

[0122] A third aspect of the present invention provides an electronic device including a processor and a memory, the memory storing a plurality of instructions, the processor being configured to read the instructions and execute the method as described in the first aspect.

[0123] A fourth aspect of the present invention provides a computer-readable storage medium storing a plurality of instructions which can be read by a processor and executed as described in the first aspect.

[0124] The beneficial effects of the method and system of the present invention are as follows:

[0125] 1. Strong environmental adaptability and high acquisition accuracy: Existing technologies use general-purpose laser scanning equipment, which has low acquisition accuracy in the harsh environment of drill-and-blast tunnels, and multi-time series point cloud registration relies on complex algorithms; This invention develops a dedicated acquisition device and configures protective equipment to adapt to the high humidity, high dust, and strong vibration environment of tunnels. Combined with RTK control points, it achieves natural registration with a registration accuracy better than 3.6mm, and significantly improves the stability and accuracy of point cloud acquisition.

[0126] 2. Automated initial support extraction with high segmentation accuracy: Existing CSF algorithms rely on cylindrical projection and center fitting, resulting in low extraction accuracy in drill-and-blast tunnels. Furthermore, traditional point cloud processing methods are inefficient and lack robustness. The improved CSF algorithm of this invention abandons projection and center fitting, and achieves automated extraction of the initial support region through parameter optimization and surface density estimation, achieving a segmentation accuracy of 96.12%. No manual intervention is required, significantly improving processing efficiency and engineering robustness.

[0127] 3. High-precision deformation calculation while preserving local details: Existing technologies are based on circle / ellipse fitting, which loses local deformation details and results in low detection accuracy. The GDef algorithm designed in this invention adopts a feature point matching method of "reference point-centroid" to achieve three-dimensional deformation quantification of the entire cross-section of the initial support. The deformation detection error is less than 2.5mm, which can accurately capture local millimeter-level deformation features and meet the needs of refined monitoring of drill-and-blast tunnels.

[0128] 4. High degree of integration and strong engineering applicability: Existing technologies have not formed a complete detection framework, and each link is independent of each other, making it difficult to adapt to the engineering site; This invention constructs an integrated detection framework of "collection-extraction-calculation-verification", realizing automated and rapid detection. The deformation results are visualized in the form of cloud maps, which makes it easy for engineering technicians to intuitively judge the stability of the initial support, providing accurate data support for tunnel construction safety decisions. It has strong engineering applicability and operability.

[0129] 5. No need for preset geometric models, adaptable to non-standard deformations: Existing tunnel point cloud processing methods mostly rely on preset circular / cylindrical geometric models, which are insufficient for handling non-standard structural deformations; the improved CSF and GDef algorithms of this invention, through flexible cloth fitting and cylindrical surface reference fitting, do not require strict standard geometric models, and can adapt to non-standard initial support deformations caused by blasting and geological conditions in drill-and-blast tunnels, thus having a wider range of applications. Attached Figure Description

[0130] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of the present invention, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0131] Figure 1 This is a flowchart of the deformation analysis method for the initial support of the drill-and-blast tunnel shotcrete based on LiDAR and the improved CSF algorithm according to an embodiment of the present invention.

[0132] Figure 2 Here is a detailed flowchart of the initial support deformation analysis method for the drill-and-blast tunnel shotcrete anchor based on LiDAR and improved CSF algorithm according to an embodiment of the present invention.

[0133] Figure 3 A schematic diagram of the system architecture for deformation analysis of initial support shotcrete anchor in drill-and-blast tunnel based on LiDAR and improved CSF algorithm according to an embodiment of the present invention.

[0134] Figure 4This is a structural diagram of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0135] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0136] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0137] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0138] The core improvement and innovation of this invention compared to existing technologies lies in the construction of an integrated LiDAR high-precision deformation detection solution that combines data acquisition, extraction, and calculation, specifically addressing the complex construction environment of drill-and-blast tunnels. This solution solves the problems of poor adaptability, insufficient accuracy, and loss of detail inherent in existing technologies. Specific innovations are as follows:

[0139] 1. Development of a dedicated LiDAR acquisition system for drill-and-blast tunnels: Designing a dedicated 3D information acquisition device consisting of laser scanning equipment, protective equipment, and edge computing control mechanism for tunnels with high humidity, high dust, and strong vibration environments. Combining the resection method of RTK control points and coordinate transformation, it achieves natural registration of multi-temporal point clouds with a registration accuracy better than 3.6mm, solving the problems of low acquisition accuracy and large registration error of multi-temporal point clouds of general equipment.

[0140] 2. Improved Fabric Simulation Filtering (CSF) Algorithm: Abandoning the cylindrical projection and center-fitting methods of the traditional CSF algorithm, this algorithm utilizes the spatial characteristics of the initial support of the drill-and-blast tunnel to directly realize virtual fabric settlement in the actual coordinate system. Local surface density estimation is introduced to optimize the fabric mesh size. The optimal parameter combination (6mm resolution, stiffness 1, and fabric spacing 1.5mm) is determined through single-factor experiments, enabling automated extraction of the initial support region with a segmentation accuracy of 96.12%. This solves the problems of poor adaptability and low extraction efficiency of existing methods for drill-and-blast tunnels.

[0141] 3. Design of GDef 3D Deformation Calculation Algorithm: Breaking through the limitations of traditional circle / ellipse fitting, the reference profile is determined by fitting a cylindrical surface. The feature point pair matching method of "reference point - nearest three centroids" is adopted to calculate the spatial deformation of the entire cross section of the initial support, retaining local millimeter-level deformation details. The deformation detection error is less than 2.5mm, which solves the problems of loss of local details and insufficient accuracy of existing deformation calculation methods.

[0142] 4. Construct an integrated deformation detection framework: Integrate the entire process of "multi-time series point cloud acquisition and registration → improved CSF algorithm for initial support extraction → GDef algorithm for deformation quantification → result verification and visualization" to achieve automated, rapid, and millimeter-level detection of the deformation of the initial support of shotcrete in drill-and-blast tunnels. It adapts to the actual application needs of engineering sites and solves the problems of low integration and poor engineering practicality of existing technologies.

[0143] Specifically, a LiDAR point cloud acquisition and registration method for drill-and-blast tunnels was developed. A 3D information acquisition device consisting of a laser scanning device, an industrial camera, protective equipment, and an edge computing industrial control computer was developed. The laser scanning device parameters are a 360° horizontal and 100° vertical scanning field of view, a horizontal angular resolution of 0.0007° and a vertical angular resolution of 0.0008°, and a ranging accuracy of 10mm at 100m. RTK control points with WGS-84 coordinates were deployed inside the tunnel. The station positions were determined by the resection method, and the TLS local coordinate system was converted to the WGS-84 coordinate system to achieve natural registration of multi-time series point clouds with a registration accuracy better than 3.6mm. The registered point cloud was then trimmed to retain the local point cloud of the initial support of the shotcrete. Furthermore, an improved cloth simulation filtering (CSF) algorithm was developed for the extraction of initial support point clouds of shotcrete in drill-and-blast tunnels. This algorithm eliminates the need to invert the tunnel point cloud, directly simulating virtual cloth gravity settlement in the actual coordinate system. Local surface density estimation is introduced, using the number of neighboring points at the center of the sphere divided by the surface area of ​​the sphere as the surface density to determine the initial mesh size of the virtual cloth. Optimal parameters were obtained through single-factor experiments: cloth mesh resolution 6mm, cloth stiffness 1, and distance between the cloth and the initial support surface 1.5mm. Through four stages—virtual cloth initialization, gravity settlement, elastic relaxation, and initial support point cloud extraction—automatic extraction of the shotcrete initial support region was achieved, with a segmentation accuracy of 96.12%. Thirdly, a GDef three-dimensional deformation calculation algorithm was developed. Its characteristic is that it uses the multi-time-series shotcrete initial support point cloud of drill-and-blast tunnels as input, establishing a scanner right-hand coordinate system (…). The axis points towards the tunnel wall. The axis is along the excavation direction. (Axis vertically upward); Perform cylindrical surface fitting on the early reference initial support point cloud to determine the center parameters. and radius Filter valid reference point clouds; for each reference point In the later detection point cloud, the three nearest points are found and their centroids are calculated. ,calculate and The spatial distance is the deformation amount, and the expansion / contraction deformation type is determined by the normal direction; the initial support point cloud set with deformation amount is output, and the deformation detection error is less than 2.5mm. Fourth, a deformation analysis method for shotcrete initial support of drill-blast tunnel based on LiDAR and cloth simulation filtering is developed, and the following steps are performed in sequence: (1) Complete the acquisition and registration of tunnel multi-time series point clouds; (2) Extract the shotcrete initial support point cloud based on the improved CSF algorithm; (3) Calculate the three-dimensional deformation amount of the initial support based on the GDef algorithm; (4) Perform statistical analysis and visualization on the deformation results, and set a 150mm filter threshold to remove abnormal points and complete the deformation detection; This method achieves millimeter-level detection of initial support deformation, with a segmentation accuracy of 96.12% and a deformation detection error of less than 2.5mm. The test results were verified through simulation experiments and field tests. In the simulation experiments, a CAD model of the initial support with radial convergence deformation and Gaussian noise was established. Field tests were conducted in a three-lane drill-and-blast highway tunnel. The test results were compared with the safe deformation limits in the "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019 to determine the stability of the initial support structure. The following technical problems were solved:

[0144] 1. How to develop a self-developed TLS acquisition system adapted to the harsh environment of drill-and-blast tunnels, and design a point cloud natural registration method based on the tunnel control network to achieve high-precision and automated acquisition and alignment of point clouds in multiple phases;

[0145] 2. How to improve the CSF algorithm, eliminate cylindrical projection and preset geometric model, adapt to the complex point cloud environment of drill-and-blast tunnels, and achieve automatic and high-precision extraction of the ROI of shotcrete and anchor initial support;

[0146] 3. How to design a dedicated GDef deformation quantization algorithm to break through the limitations of traditional geometric fitting, realize full-section 3D deformation calculation of the initial support of the shotcrete anchor in the drill-blast method tunnel, retain local deformation details and control the detection error to the millimeter level;

[0147] 4. How to optimize the key parameters of the CSF algorithm (Bouguer mesh resolution, rigidity, distance between Bouguer and initial support) to achieve a balance between ROI extraction accuracy and computational efficiency, and meet the actual application needs of engineering sites.

[0148] To achieve the following technical objectives:

[0149] 1. Develop a self-developed TLS acquisition system adapted to the tunnel environment of the drill-and-blast method, and achieve natural registration of point clouds in multiple phases with a registration accuracy better than 3.6mm;

[0150] 2. Improve the CSF algorithm and optimize its key parameters to achieve automatic extraction of the ROI of the initial support of the shotcrete anchor, with a segmentation accuracy of over 96.12%;

[0151] 3. The GDef deformation quantization algorithm is proposed to realize full-section 3D deformation detection of the initial support of the shotcrete anchor. The deformation detection error is less than 2.5mm, and local deformation details are preserved.

[0152] 4. A complete technical framework of "point cloud acquisition - registration - ROI extraction - deformation quantification - verification" is formed, providing a practical and efficient millimeter-level deformation monitoring solution for drill-and-blast tunnel construction, thereby improving the safety level of tunnel construction.

[0153] The first aspect of this invention is to provide a method for analyzing the deformation of initial supports of shotcrete in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm. This method belongs to the category of deformation analysis methods for initial supports of shotcrete in drill-and-blast tunnels based on LiDAR and simulated filtering. It constructs an integrated technical framework of "multi-time-series point cloud acquisition and registration → initial support extraction based on the improved CSF algorithm → deformation quantization based on the GDef algorithm → deformation result output and analysis," achieving high-precision deformation detection of initial supports of shotcrete in drill-and-blast tunnels. The overall process of this method is as follows: Figure 1-2 As shown, it includes:

[0154] S1, Acquisition and registration of multi-temporal point clouds of drill-and-blast tunnels;

[0155] In a preferred embodiment, S1 includes:

[0156] S11, A LiDAR point cloud acquisition system specifically designed for tunnel structure construction is set up for 3D information acquisition, including a laser scanning device, an industrial camera, protective equipment, and an edge computing industrial control computer; the protective equipment is used to protect the laser scanning device from dust, shock, and water, and the edge computing industrial control computer is used to realize the rapid processing of on-site point cloud data, adapting to the harsh environment of tunnel structure construction;

[0157] In this embodiment, a LiDAR point cloud acquisition system specifically designed for tunnel structure construction is set up for 3D information acquisition in the high humidity, high dust, and strong vibration construction environment of the drill-and-blast tunnel method. The core technical parameters of the laser scanning equipment are shown in Table 1, which shows the technical specifications of the laser scanning equipment.

[0158] Table 1

[0159]

[0160] S12, High-precision acquisition of multi-temporal point clouds, including: setting up RTK control points with known WGS-84 coordinates inside the tunnel; using the LiDAR point cloud acquisition system specifically designed for tunnel structure construction, performing laser scanning on the initial support area of ​​the shotcrete in the tunnel at the same control point at different time points during tunnel construction (multi-temporal), acquiring massive amounts of three-dimensional point cloud data, the point cloud data containing spatial coordinates. Information reflects the geometric shape of the initial branch.

[0161] S13, performs multi-temporal point cloud registration and cropping, including:

[0162] The resection method was adopted, and the spatial location of the TLS station was determined by measuring the RTK control points with a total station. The TLS local coordinate system of the point cloud was converted into the WGS-84 geographic coordinate system to achieve natural registration of multi-time series point clouds. The coordinate transformation corresponding to the conversion of the TLS local coordinate system of the point cloud into the WGS-84 geographic coordinate system adopts the Bursa-Wolf seven-parameter transformation model, and the formula is shown in the following formula (1):

[0163] (1);

[0164] in, Coordinates in the WGS-84 coordinate system. Coordinates in the TLS local coordinate system; , and For translation parameters, , and Rotational parameters, unit: rad. Scale factor;

[0165] The root mean square error (RMSE) is used to evaluate the point cloud registration accuracy, as shown in the following formula (2):

[0166] (2);

[0167] in, To register the number of control points, and The coordinates of the control points before and after registration are shown. The registration accuracy of this method is better than 3.6mm.

[0168] The registered original point cloud was cropped to remove non-monitored areas such as the secondary initial support and outside the tunnel entrance, retaining only the local tunnel point cloud including the shotcrete initial support. The cropping method employed a spatial cube cropping approach, defining the boundary coordinates of the cropping region. Point clouds that satisfy the following conditions as shown in equation (3) shall be retained:

[0169] (3);

[0170] S2, Extraction of shotcrete initial support point cloud based on improved CSF algorithm, including: Based on the curved geometric features of shotcrete initial support in drill-and-blast tunnel, local surface density estimation is introduced to optimize CSF algorithm parameters, and the automatic extraction of shotcrete initial support point cloud in the initial support area is achieved by simulating the natural settlement of virtual fabric under gravity.

[0171] In this embodiment, the cylindrical projection and center-fitting methods of the traditional CSF algorithm are abandoned. Instead, the curved surface geometry of the initial support of shotcrete in drill-and-blast tunnels is combined with the local surface density estimation to optimize the CSF algorithm parameters. By simulating the natural settlement of the virtual support under gravity, the automatic extraction of the initial support area is achieved. The specific process is as follows:

[0172] S21, Determine the improved CSF algorithm; the improved CSF algorithm includes:

[0173] S211 utilizes the spatial characteristics of the non-RoI of the initial support of the drill-and-blast tunnel located inside the initial support, allowing the virtual cloth to settle naturally directly in the actual coordinate system without the need for tunnel point cloud inversion, thus simplifying the preprocessing process.

[0174] S212, Estimating Local Surface Density: Calculating the Surface Density of Tunnel Point Clouds , which represents the number of neighboring points in the spherical region centered at each point / the surface area of ​​the spherical region, as shown in the following formula (4):

[0175] (4);

[0176] in, For surface density, Let be the number of neighboring points within the spherical region. The radius of the spherical region is taken as , in this embodiment. The initial mesh size of the virtual cloth is determined based on the surface density, balancing extraction accuracy and computational efficiency.

[0177] S213, Equation of motion of virtual cloth particle: Considering the combined effect of gravity and spring force, the equation of motion of virtual cloth particle is as follows (5):

[0178] (5);

[0179] in, For particle mass, For the particle position vector, It is the gravitational acceleration vector. This is the spring force between adjacent particles. The collision force between particles and point clouds. This represents the number of adjacent particles.

[0180] S214, Calculate the spring force using Hooke's Law. As shown in equation (6):

[0181] (6);

[0182] in, For spring stiffness, For spring deformation, This represents the position vector between adjacent particles. This is the original length of the spring.

[0183] S215, the collision force is calculated using an elastic collision model. As shown in equation (7):

[0184] (7);

[0185] in, For collision stiffness, The penetration depth of particles and point clouds. is the normal vector of the collision contact surface.

[0186] S22, a four-stage simulation of the extraction of the initial support for shotcrete is performed, including:

[0187] Initializing the virtual cloth includes: generating a virtual cloth mesh composed of interconnected particles above the tunnel model, wherein the particles are uniformly distributed and connected by springs to form a flexible structure that can adapt to the complex geometry of the initial support.

[0188] The virtual cloth undergoes gravity settlement, including: releasing the virtual cloth and allowing it to settle towards the surface of the initial support under simulated gravity; when the cloth particles come into contact with the outer surface of the shotcrete initial support, they are marked as fixed particles and stop moving.

[0189] Implementing elastic relaxation of the virtual cloth includes: using the spring force between adjacent particles to allow the movable particles to continue adjusting their positions until the internal tension of the cloth stabilizes, allowing the cloth to smoothly conform to the initial support surface and eliminating interference caused by local point cloud irregularities and occlusions.

[0190] Extracting the initial support point cloud of shotcrete includes: after the virtual cloth reaches a stable state, its shape is the outer surface contour of the initial support of shotcrete. By judging whether the spatial distance between the original point cloud and the virtual cloth is less than a set threshold, the initial support point cloud of shotcrete is extracted, as shown in Equation (8):

[0191] (8);

[0192] in, For point The shortest distance to the virtual cloth, As the distance threshold, this embodiment takes... RoI represents the region of interest, i.e., the initial pivot cloud.

[0193] S23, Parameter optimization of the improved CSF algorithm, including: optimizing the core parameters of the simulated filtering through single-factor analysis experiments, and determining the optimal parameter combination for extracting the initial support of the tunnel using the drill-and-blast method. The optimal parameter combination includes:

[0194] Mesh resolution: 6mm (matching the average neighbor distance of point cloud acquisition, balancing accuracy and efficiency);

[0195] Cloth stiffness: 1 (the cloth particles move only once, and the displacement is 1 / 2 of the vertical distance of the particle, which effectively suppresses noise and adapts to the complex contours of tunnels).

[0196] Distance between the fabric and the initial support surface: 1.5mm (set according to the "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019, effectively including the initial support point cloud and avoiding the introduction of non-standard points). data).

[0197] The optimization process of the core parameters adopts a comprehensive evaluation index. and calculation time Construct the objective function This achieves multi-objective optimization, as shown in equation (9):

[0198] (9);

[0199] in, , These are the weighting coefficients. for , This represents the maximum computation time.

[0200] Segmentation accuracy uses intersection-union ratio (IU / U). The evaluation is shown in equation (10):

[0201] (10);

[0202] in, The correct number of initial fulcrums to be extracted. The number of non-initial pivots that were mistakenly extracted. This represents the number of initial fulcrums that were not extracted.

[0203] S3, quantifying the three-dimensional deformation of the initial support of shotcrete based on the GDef (Get Deformation of Two Tunnels) deformation algorithm, including: designing the GDef (Get Deformation of Two Tunnels) deformation algorithm, using the early initial support point cloud in multiple time series as a reference and the later initial support point cloud as the detection object, and realizing the three-dimensional deformation quantification of the entire section of the initial support by extracting cross-sectional feature points and calculating point cloud deviations, while preserving local millimeter-level deformation details. The input of the GDef (Get Deformation of Two Tunnels) deformation algorithm is the initial support point cloud in multiple time series, including the reference initial support point cloud. and detection of initial fulcrum cloud The output is a set of initial support point clouds with deformation. The specific steps are as follows:

[0204] S31, Define the coordinate system, including: establishing the scanner's right-handed coordinate system, where... The axis points towards the tunnel wall. The axis runs along the tunnel excavation direction. The axis is vertically upward, providing a spatial reference for cross-section extraction and deformation calculation;

[0205] S32, Fitting the cylindrical surface of the reference initial pivot cloud, including: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The cylindrical surface is fitted, and the equation of the cylindrical surface is shown in equation (11) below:

[0206] (11);

[0207] in,( , ( ) represents the coordinates of the cylinder's center. Where is the radius of the cylinder;

[0208] The cylinder parameters are solved using the least squares method, and the objective function is shown in equation (12):

[0209] (12);

[0210] The optimal solution for the cylinder parameters is obtained by solving the above optimization problem using the Lagrange multiplier method.

[0211] S33, Match feature point pairs and calculate the bias, including:

[0212] For each valid reference point of the reference profile surface The Euclidean distance method is used for nearest neighbor search in the detection of initial pivot cloud. Finding distance The 3 most recent points , and As shown in equation (13):

[0213] (13);

[0214] calculate , and centroid Centroid The coordinates are calculated as shown in equation (14):

[0215] (14);

[0216] Calculate the centroid With reference point The spatial distance is considered, and the deformation type is determined by taking into account the normal direction, and the deformation amount is calculated. As shown in equation (15):

[0217] (15);

[0218] in, For reference point The unit normal vector at that location is positive if it points outward from the tunnel. For expansion and deformation, For shrinkage deformation;

[0219] The reference point normal vector is calculated using a local plane fitting method. of The plane is fitted to the neighboring points, and the plane equation is shown in equation (16):

[0220] (16);

[0221] Wherein, the normal vector is The unit normal vector is .

[0222] S34, statistically analyze the deformation of the entire cross section and visualize it;

[0223] In this embodiment, S34 includes:

[0224] Statistical analysis was performed on the deformation of all characteristic points of the initial support section, and key parameters were calculated, including: maximum value, minimum value, and average deformation. Standard deviation Root mean square error As shown in equations (17)-(19):

[0225] (17);

[0226] (18);

[0227] (19);

[0228] in, This represents the average deformation. Standard deviation, The root mean square error, This represents the number of feature points.

[0229] The deformation is converted into a color gradient to generate the initial support 3D deformation cloud map. The color mapping function is shown in equation (20) below:

[0230] (20);

[0231] in, The maximum allowable deformation is determined in accordance with the "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019.

[0232] S4, Verify and analyze the deformation detection results;

[0233] In this embodiment, S4 includes:

[0234] S41, Verification simulation experiment, including: establishing a simplified CAD model of the initial support of shotcrete in the drill-and-blast tunnel, setting different degrees of radial convergence deformation (2cm, 4cm, 6cm), introducing non-RoI models such as construction equipment and ventilation ducts, and adding Gaussian noise with zero mean and standard deviation of 0.005m to simulate scanning error. The Gaussian noise generation formula is shown in the following formula (21):

[0235] (twenty one);

[0236] in, =0.005m is the noise standard deviation.

[0237] S42, Verification field test, including: In a three-lane highway tunnel constructed by actual drill-and-blast method, the method of this invention is applied to conduct field tests, collect the initial support point cloud at different construction stages, complete the initial support extraction and deformation calculation, compare the deformation detection results with the safe deformation limit of the "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019, and judge the stability of the initial support structure. The deformation detection error is evaluated by absolute error and relative error, as shown in Equation (22) HE (23):

[0238] (twenty two);

[0239] (twenty three);

[0240] in, The actual deformation was measured using a high-precision total station. For absolute error, The deformation detection error of this method is less than 2.5 mm, which is a relative error.

[0241] S43, outlier filtering, including: for a small number of misclassified point clouds caused by ventilation duct obstruction during on-site detection, a filtering threshold of 150mm is set to remove outliers that are more than 150mm away from the initial support surface of the fitting, as shown in equation (24):

[0242] (twenty four);

[0243] in, The filtering threshold is... For point The distance to the fitted cylindrical surface.

[0244] like Figure 3 As shown, a second aspect of the present invention is to provide a deformation analysis system for the initial support of a drill-and-blast tunnel using shotcrete and anchor methods based on LiDAR and an improved CSF algorithm, comprising:

[0245] The acquisition and registration module 101 is used to acquire and register multi-temporal point clouds of drill-and-blast tunnels;

[0246] The point cloud extraction module 102 is used to extract the point cloud of the initial support of shotcrete based on the improved CSF algorithm. It includes: based on the curved geometric features of the initial support of shotcrete in the drill-and-blast tunnel, local surface density estimation is introduced to optimize the CSF algorithm parameters, and the automatic extraction of the point cloud of the initial support of shotcrete in the initial support area is achieved by simulating the natural settlement of the virtual support under gravity.

[0247] The three-dimensional deformation quantization module 103 is used to quantify the three-dimensional deformation of the initial support of shotcrete based on the GDef deformation algorithm. This includes: designing the GDef deformation algorithm, using early initial support point clouds from multiple time series as a reference and later initial support point clouds as the detection object; extracting cross-sectional feature points and calculating point cloud deviations to achieve three-dimensional deformation quantization of the entire cross-section of the initial support, while preserving local millimeter-level deformation details. The input to the GDef deformation algorithm is the initial support point clouds from multiple time series, including the reference initial support point cloud. and detection of initial fulcrum cloud The output is a set of initial support point clouds with deformation.

[0248] The verification and analysis module 104 is used to verify and analyze the deformation detection results.

[0249] The present invention also provides a memory that stores a plurality of instructions for implementing the methods as described in the embodiments and application embodiments.

[0250] like Figure 4 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301. The memory 302 stores a plurality of instructions, which can be loaded and executed by the processor to enable the processor to perform the methods as described in the embodiments and application embodiments.

[0251] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing the deformation of initial support of shotcrete anchors in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm, characterized in that, include: S1, Acquisition and registration of multi-temporal point clouds of drill-and-blast tunnels; S2, Extraction of shotcrete initial support point cloud based on improved CSF algorithm, including: Based on the curved geometric features of shotcrete initial support in drill-and-blast tunnel, local surface density estimation is introduced to optimize CSF algorithm parameters, and the automatic extraction of shotcrete initial support point cloud in the initial support area is achieved by simulating the natural settlement of virtual fabric under gravity. S3, quantifying the three-dimensional deformation of the initial support of shotcrete based on the GDef deformation algorithm, including: designing the GDef deformation algorithm, using the early initial support point cloud in multiple time series as a reference and the later initial support point cloud as the detection object, and realizing the three-dimensional deformation quantification of the entire section of the initial support by extracting cross-sectional feature points and calculating point cloud deviations, while preserving local millimeter-level deformation details. The input of the GDef deformation algorithm is the initial support point cloud in multiple time series, including the reference initial support point cloud. and detection of initial fulcrum cloud The output is a set of initial support point clouds with deformation. S4. Verify and analyze the deformation detection results.

2. The method for analyzing the initial support deformation of a drill-and-blast tunnel using shotcrete and anchor based on LiDAR and an improved CSF algorithm, as described in claim 1, is characterized in that... S1 includes: S11, A LiDAR point cloud acquisition system specifically designed for tunnel structure construction is set up for 3D information acquisition, including a laser scanning device, an industrial camera, protective equipment, and an edge computing industrial control computer; the protective equipment is used to protect the laser scanning device from dust, shock, and water, and the edge computing industrial control computer is used to realize the rapid processing of on-site point cloud data, adapting to the harsh environment of tunnel structure construction; S12, High-precision acquisition of multi-time series point clouds, including: setting up RTK control points with known WGS-84 coordinates inside the tunnel; using the LiDAR point cloud acquisition system specifically designed for tunnel structure construction, performing laser scanning on the initial support area of ​​the shotcrete in the tunnel at the same control point at different time points during tunnel construction, acquiring massive amounts of three-dimensional point cloud data, the point cloud data containing spatial coordinates. Information reflects the geometric shape of the initial branch; S13 performs registration and cropping of multi-temporal point clouds.

3. The method for analyzing the initial support deformation of a drill-and-blast tunnel using shotcrete and anchor based on LiDAR and an improved CSF algorithm, as described in claim 2, is characterized in that... S13 includes: The resection method was adopted, and the spatial location of the TLS station was determined by measuring the RTK control points with a total station. The TLS local coordinate system of the point cloud was converted into the WGS-84 geographic coordinate system to achieve natural registration of multi-time series point clouds. The coordinate transformation corresponding to the conversion of the TLS local coordinate system of the point cloud into the WGS-84 geographic coordinate system adopts the Bursa-Wolf seven-parameter transformation model, and the formula is shown in the following formula (1): (1); in, Coordinates in the WGS-84 coordinate system. Coordinates in the TLS local coordinate system; , and For translation parameters, , and Rotational parameters, unit: rad. Scale factor; The root mean square error (RMSE) is used to evaluate the point cloud registration accuracy, as shown in the following formula (2): (2); in, To register the number of control points, and These are the coordinates of the control points before and after registration; The registered original point cloud was cropped to remove the secondary initial support and non-monitored areas outside the tunnel entrance, retaining the local tunnel point cloud including the shotcrete initial support. The cropping method was spatial cube cropping, and the boundary coordinates of the cropping area were defined. Point clouds that satisfy the following conditions as shown in equation (3) shall be retained: (3)。 4. The method for analyzing the deformation of the initial support of the shotcrete anchor in the drill-and-blast tunnel based on LiDAR and an improved CSF algorithm according to claim 3, characterized in that, S2 includes: S21, determine the improved CSF algorithm; S22, Four-stage simulation of the extraction of initial support for shotcrete; S23, Parameter optimization of the improved CSF algorithm, including: optimizing the core parameters of the simulated filtering through single-factor analysis experiments, and determining the optimal parameter combination for extracting the initial support of the tunnel using the drill-and-blast method. The optimal parameter combination includes: The mesh resolution is 6mm; the fabric stiffness is 1; the distance between the fabric and the initial support surface is 1.5mm. The optimization process of the core parameters adopts a comprehensive evaluation index. and calculation time Construct the objective function This achieves multi-objective optimization, as shown in equation (9): (9); in, , These are the weighting coefficients. for , The maximum computation time is used; the segmentation accuracy is calculated using the intersection-union ratio (IUU). The evaluation is shown in equation (10): (10); in, The correct number of initial fulcrums to be extracted. The number of non-initial pivots that were mistakenly extracted. This represents the number of initial fulcrums that were not extracted.

5. The method for analyzing the deformation of initial support of shotcrete anchor in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm according to claim 4, characterized in that, S21 includes: S211 utilizes the spatial characteristics of the non-RoI of the initial support of the drill-and-blast tunnel located inside the initial support, allowing the virtual cloth to settle naturally in the actual coordinate system without the need for tunnel point cloud inversion. S212, Estimating Local Surface Density: Calculating the Surface Density of Tunnel Point Clouds , which represents the number of neighboring points in the spherical region centered at each point / the surface area of ​​the spherical region, as shown in the following formula (4): (4); in, For surface density, Let be the number of neighboring points within the spherical region. The initial mesh size of the virtual cloth is determined based on the surface density, where the radius of the spherical region is given. S213, Equation of motion of virtual cloth particle: Considering the combined effect of gravity and spring force, the equation of motion of virtual cloth particle is as follows (5): (5); in, For particle mass, For the particle position vector, It is the gravitational acceleration vector. This is the spring force between adjacent particles. The collision force between particles and point clouds. The number of adjacent particles; S214, Calculate the spring force using Hooke's Law. As shown in equation (6): (6); in, For spring stiffness, For spring deformation, This represents the position vector between adjacent particles. This is the original length of the spring; S215, the collision force is calculated using an elastic collision model. As shown in equation (7): (7); in, For collision stiffness, The penetration depth of particles and point clouds. is the normal vector of the collision contact surface.

6. The method for analyzing the deformation of the initial support of the drill-and-blast tunnel using shotcrete and anchor based on LiDAR and an improved CSF algorithm as described in claim 5, characterized in that, S22 includes: Initializing the virtual cloth includes: generating a virtual cloth mesh composed of interconnected particles above the tunnel model, wherein the particles are uniformly distributed and connected by springs to form a flexible structure that can adapt to the complex geometry of the initial support. The virtual cloth undergoes gravity settlement, including: releasing the virtual cloth and allowing it to settle towards the surface of the initial support under simulated gravity; when the cloth particles come into contact with the outer surface of the shotcrete initial support, they are marked as fixed particles and stop moving. Implementing elastic relaxation of the virtual cloth includes: using the spring force between adjacent particles to allow the movable particles to continue adjusting their positions until the internal tension of the cloth stabilizes, allowing the cloth to smoothly conform to the initial support surface and eliminating interference caused by local point cloud irregularities and occlusions. Extracting the initial support point cloud of shotcrete includes: after the virtual cloth reaches a stable state, its shape is the outer surface contour of the initial support of shotcrete. By judging whether the spatial distance between the original point cloud and the virtual cloth is less than a set threshold, the initial support point cloud of shotcrete is extracted, as shown in Equation (8): (8); in, For point The shortest distance to the virtual cloth, The distance threshold is denoted as RoI, which is the region of interest, i.e., the initial pivot cloud.

7. The method for analyzing the deformation of initial support of shotcrete anchor in drill-and-blast tunnels based on LiDAR and an improved CSF algorithm according to claim 6, characterized in that, S23 includes: S23, Parameter optimization of the improved CSF algorithm, including: optimizing the core parameters of the fabric simulation filter through single-factor analysis experiments, and determining the optimal parameter combination for the initial support extraction of the tunnel using the drill-and-blast method. The optimal parameter combination includes: a fabric grid resolution of 6 mm; a fabric stiffness of 1; and a distance of 1.5 mm between the fabric and the initial support surface. The optimization process of the core parameters adopts a comprehensive evaluation index. and calculation time Construct the objective function This achieves multi-objective optimization, as shown in equation (9): (9); in, , These are the weighting coefficients. for , This represents the maximum computation time. Segmentation accuracy uses intersection-union ratio (IU / U). The evaluation is shown in equation (10): (10); in, The correct number of initial fulcrums to be extracted. The number of non-initial pivots that were mistakenly extracted. This represents the number of initial fulcrums that were not extracted.

8. The method for analyzing the initial support deformation of a drill-and-blast tunnel using shotcrete and anchor based on LiDAR and an improved CSF algorithm, as described in claim 7, is characterized in that... S3 includes: S31, Define the coordinate system, including: establishing the scanner's right-handed coordinate system, where... The axis points towards the tunnel wall. The axis runs along the tunnel excavation direction. The axis is vertically upward, providing a spatial reference for cross-section extraction and deformation calculation; S32, Fitting the cylindrical surface of the reference initial pivot cloud, including: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The cylindrical surface is fitted, and the equation of the cylindrical surface is shown in equation (11) below: (11); in,( , ( ) represents the coordinates of the cylinder's center. Where is the radius of the cylinder; The cylinder parameters are solved using the least squares method, and the objective function is shown in equation (12): (12); The optimal solution for the cylinder parameters is obtained by solving the above optimization problem using the Lagrange multiplier method. S33, Match feature point pairs and calculate the bias, including: For each valid reference point of the reference profile surface The Euclidean distance method is used for nearest neighbor search in the detection of initial pivot cloud. Finding distance The 3 most recent points , and As shown in equation (13): (13); calculate , and centroid Centroid The coordinates are calculated as shown in equation (14): (14); Calculate the centroid With reference point The spatial distance is considered, and the deformation type is determined by taking into account the normal direction, and the deformation amount is calculated. As shown in equation (15): (15); in, For reference point The unit normal vector at that location is positive if it points outward from the tunnel. For expansion and deformation, For shrinkage deformation; The reference point normal vector is calculated using a local plane fitting method. of The plane is fitted to the neighboring points, and the plane equation is shown in equation (16): (16); Wherein, the normal vector is The unit normal vector is ; S34, statistically analyze the deformation of the entire cross-section and visualize it; S34 includes: Statistical analysis was performed on the deformation of all characteristic points of the initial support section, and key parameters were calculated, including: maximum value, minimum value, and average deformation. Standard deviation Root mean square error As shown in equations (17)-(19): (17); (18); (19); in, This represents the average deformation. Standard deviation, The root mean square error, This represents the number of feature points; The deformation is converted into a color gradient to generate the initial support 3D deformation cloud map. The color mapping function is shown in equation (20) below: (20); in, The maximum allowable deformation is determined in accordance with the "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019.

9. The method for analyzing the initial support deformation of a drill-and-blast tunnel using shotcrete and anchor based on LiDAR and an improved CSF algorithm, as described in claim 8, is characterized in that... S4 includes: S41, Verification simulation experiment, including: establishing a simplified CAD model of the initial support of shotcrete in a drill-and-blast tunnel, setting different degrees of radial convergence deformation, and introducing non- The model is then used to simulate scanning errors by adding Gaussian noise with zero mean and a standard deviation of 0.005m. The Gaussian noise generation formula is shown in equation (21) below: (21); in, =0.005m is the standard deviation of noise; S42, Verification field test, including: conducting field tests in a three-lane highway tunnel constructed by actual drill-and-blast method, collecting initial support point clouds at different construction stages, completing initial support extraction and deformation calculation, comparing the deformation detection results with the safe deformation limit of "Technical Specification for Highway Tunnel Construction" JTG / T F60-2019, and judging the stability of the initial support structure; the deformation detection error is evaluated by absolute error and relative error, as shown in equations (22) and (23): (22); (23); in, The actual deformation was measured using a high-precision total station. For absolute error, This is relative error; S43, outlier filtering, including: for a small number of misclassified point clouds in the field detection, setting a filtering threshold of 150mm to remove outlier points that are more than the threshold away from the initial support surface of the fitting, as shown in equation (24): (24); in, The filtering threshold is... For point The distance to the fitted cylindrical surface.

10. A system for analyzing the initial support deformation of a drill-and-blast tunnel using shotcrete and anchoring based on LiDAR and an improved CSF algorithm, used to implement the method described in any one of claims 1-9, characterized in that, include: The acquisition and registration module (101) is used to acquire and register multi-temporal point clouds of drill-and-blast tunnels; The point cloud extraction module (102) is used to extract the point cloud of the initial support of shotcrete based on the improved CSF algorithm. It includes: based on the curved geometric features of the initial support of shotcrete in the drill-and-blast tunnel, local surface density estimation is introduced to optimize the parameters of the CSF algorithm. By simulating the natural settlement of the virtual fabric under gravity, the automatic extraction of the point cloud of the initial support of shotcrete in the initial support area is realized. The three-dimensional deformation quantization module (103) is used to quantify the three-dimensional deformation of the initial support of shotcrete based on the GDef deformation algorithm. This includes: designing the GDef deformation algorithm, using the early initial support point cloud in multiple time series as a reference and the later initial support point cloud as the detection object; extracting cross-sectional feature points and calculating point cloud deviations to achieve three-dimensional deformation quantization of the entire cross-section of the initial support, while preserving local millimeter-level deformation details. The input to the GDef deformation algorithm is the initial support point cloud in multiple time series, including the reference initial support point cloud. and detection of initial fulcrum cloud The output is a set of initial support point clouds with deformation. The verification and analysis module (104) is used to verify and analyze the deformation detection results.