Tunnel section recognition and deformation detection method based on laser scanning point cloud data

Abstract

The invention provides a tunnel section identification and deformation detection method based on laser scanning point cloud data, and the method comprises the steps: directly processing the point cloud data in a three-dimensional space through polar coordinate conversion and local calculation weighted average modes to extract a central axis, so as to completely reserve the three-dimensional information of the point cloud data, and more accurately reflect the real spatial form of a tunnel. The problem that geometric feature information of the top and the bottom of the tunnel is lost due to projection is avoided. Tunnel section interception and section point cloud data determination are carried out on the basis, global and local deformation discrimination is carried out in combination with adaptive shape fitting, tunnel section deformation is accurately monitored and discriminated, possible tiny deformation of a tunnel in a local area can be sensitively captured, and even under the condition that overall deformation is not obvious, the tunnel section deformation can be accurately monitored and discriminated. Potential safety hazards can be found in advance, all-around and precise monitoring and judgment of tunnel section deformation are achieved, and powerful technical support is provided for long-term safe operation of a tunnel.

Description

Technical Field

[0001] The present invention relates to the technical field of nondestructive detection of tunnel defects, and in particular to a tunnel section recognition and deformation detection method based on laser scanning point cloud data. Background Art

[0002] Tunnel engineering plays an important role in modern transportation. With the development and expansion of cities and transportation, the depth and width of tunnel engineering such as subway tunnels, high-speed rail tunnels, underground tunnels, and underwater tunnels are constantly expanding. With the influence of traffic volume, travel time, changes in soil pressure of tunnel surrounding rock, concrete shrinkage, rainwater erosion, and the quality of the tunnel lining itself (such as uneven thickness, misalignment, and voids), it is particularly important to monitor deformation, cracks, and falling off in the tunnel. Accurate, rapid, and automated monitoring of tunnel defects through automated detection methods is the key to ensuring driving safety.

[0003] At present, tunnel section deformation monitoring is mainly based on tunnel point cloud data to analyze and warn of tunnel deformation. For example, based on laser scanning point cloud data, cross-section extraction is performed based on the central axis, and then image processing is performed on the section. Existing research focuses on cutting sections based on the central axis, but the extraction of the central axis itself is a technical problem. The axis inside the tunnel does not extend in a straight line, and its height fluctuations and angle changes pose challenges to the mathematical method of extracting the spatial central axis.

[0004] Among the existing methods, Du Liming et al. proposed to project the tunnel point cloud data onto a horizontal plane and calculate the projection point set of the central axis on the plane, and then obtain the final spatial central axis by searching for high and low points based on the point set. However, as mentioned above, the internal direction of the tunnel itself is a complex nonlinear structure. For example, for non-circular tunnels with curves or complex internal structures, when the tunnel point cloud data is projected onto a horizontal plane, the depth information in the three-dimensional space will be lost. In actual situations, the tunnel is a complex geometric body with a three-dimensional spatial structure. The projection operation will cause the geometric feature information of the top and bottom of the tunnel to be lost or deformed in the vertical direction (depth direction). For tunnels with different slopes and undulating tops or bottoms, these features will be "flattened" onto the plane after projection, so that the central axis subsequently calculated based on the plane projection point set cannot accurately reflect the true shape of the tunnel in the vertical direction. Summary of the invention

[0005] In view of the existing problems in tunnel section identification and deformation monitoring, the purpose of the present invention is to provide a tunnel section identification and deformation detection method based on laser scanning point cloud data. The central axis is extracted by directly processing the point cloud data through polar coordinate conversion and local calculation weighted averaging in three-dimensional space, so as to completely retain the three-dimensional information of the point cloud data and avoid the problem of loss of geometric feature information of the top and bottom of the tunnel caused by projection; on this basis, the tunnel section is intercepted and the section point cloud data is determined, and global and local deformation discrimination is performed in combination with adaptive shape fitting, so as to realize accurate monitoring and discrimination of tunnel section deformation.

[0006] According to the first aspect of the present invention, a tunnel section recognition and deformation detection method based on laser scanning point cloud data is proposed, which comprises the following steps:

[0007] Step 1: Obtain point cloud data of any two adjacent sites in the tunnel, complete the registration of point cloud data of all laser scanning sites through point cloud registration, and obtain tunnel point cloud data;

[0008] Step 2: De-noise the point cloud data of the entire tunnel and perform spatial simplification of the point cloud data to obtain voxelized point cloud data;

[0009] Step 3: Extract the local central axis points of the tunnel based on polar coordinate transformation, and obtain the central axis of the tunnel after global central axis fitting optimization;

[0010] Step 4: Based on the obtained tunnel centerline as a reference, determine a plane perpendicular to the centerline according to the spacing as a section intercepting plane, and calculate the section intercepting plane equation;

[0011] Step 5: Segment the points in the point cloud data that are within a certain distance range on both sides of each cross-section interception plane as the original point cloud data of the cross-section interception plane, and sort and number them after removing isolated points to obtain cross-section point cloud data;

[0012] Step 6: Perform adaptive shape analysis on the cross-section point cloud data and perform adaptive shape fitting to obtain the shape and size parameters of the fitted tunnel cross section;

[0013] Step 7: Based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation of the tunnel section is evaluated to determine the magnitude and direction of the deformation change;

[0014] Step 8: Locally divide the tunnel section and monitor the local deformation according to the changes of the divided sub-areas.

[0015] In a further embodiment, in step 1, the registration of point cloud data of all laser scanning sites is completed by point cloud registration to obtain tunnel point cloud data, including:

[0016] Step 1.1, based on the feature points, solve the rigid transformation rotation matrix and translation vector between the stations, transform the point cloud data, and complete the rough registration;

[0017] Step 1.2: Based on the rough registration, optimization is performed based on the ICP algorithm, using a point-to-point distance measurement method, and using the LM algorithm to iteratively optimize and solve the transformation matrix to minimize the error function.

[0018] In a further embodiment, in step 2, the entire tunnel point cloud data is subjected to denoising and point cloud data space simplification to obtain voxelized point cloud data, including:

[0019] Step 2.1, statistical filtering and radius filtering are performed on the tunnel point cloud data to remove the marked noise points;

[0020] Step 2.2, divide the point cloud data space into voxel grids, retain only one representative point for each voxel grid, and take the point closest to the center of the voxel grid as the representative point to obtain voxelized point cloud data.

[0021] In a further embodiment, in step 3, extracting the local central axis points of the tunnel based on polar coordinate transformation, and obtaining the central axis of the tunnel after global central axis fitting optimization, includes:

[0022] Step 3.1, convert the voxelized point cloud data from the Cartesian coordinate system to the polar coordinate system, and determine the polar diameter and polar angle of any point in the point cloud data;

[0023] Step 3.2, in each polar angle interval, select points within a certain range with the polar angle as the center and expand to the left and right, and calculate the local central axis point according to the polar diameter of the selected points, and use weighted average to calculate the polar diameter of the local central axis point;

[0024] Step 3.3, convert the local central axis points calculated in each polar angle interval back to the Cartesian coordinate system to obtain a set of discrete local central axis point sets;

[0025] Step 3.4, using the spline curve fitting method to perform curve fitting on the local central axis point set, and obtain a smooth curve as the preliminary central axis of the tunnel;

[0026] Step 3.5: Optimize the preliminary central axis to obtain the tunnel central axis.

[0027] In a further embodiment, in each polar angle interval, points within a certain range are selected with the polar angle as the center and the polar diameters of the selected points are calculated, and the polar diameters of the local central axis points are calculated by weighted average, including:

[0028] In each polar angle interval, with the polar angle as the center, the left and right expansions are Δθ ext , choose Δθ ext For points within the extended range, the polar diameter of the local central axis point is calculated using weighted average:

[0029] Assume that Δθ ext Any point P within the extended range i , whose polar diameter is ρ i , whose weight ω i According to point P i The distance to the polar angle center ray is determined by:

[0030]

[0031] Where, d i Indicates point P i The distance to the polar angle center ray, ∈ represents a minimum constant to prevent the denominator from being zero;

[0032] Polar diameter of the local midline point The calculation formula is:

[0033]

[0034] In a further embodiment, the step of optimizing the preliminary central axis to obtain the tunnel central axis includes:

[0035] For each point in the point cloud data within the set boundary range around the preliminary central axis, the distance deviation between each point and the preliminary central axis is calculated;

[0036] For points whose distance deviation exceeds the set distance threshold, they are marked as deviation points;

[0037] Divide the preliminary central axis and its surrounding point cloud data into local areas at equal length intervals;

[0038] Count the number of deviation points marked in each local area, and mark the local area where the number of deviation points exceeds the proportion as a problem segment;

[0039] For the central axis segment of the marked problem segment, take the two end points of the problem segment as the center, expand to both sides by a certain distance, and select point cloud data within a certain width in the direction perpendicular to the central axis as the candidate point cloud for recalculating the local central axis;

[0040] Within the selected candidate point cloud range, the weighted average is used to recalculate the local central axis segment and replace the central axis segment of the problem segment to update the preliminary central axis;

[0041] The central axis segments of all problem segments are updated, and the endpoints of the positions are smoothly replaced to obtain the central axis of the tunnel.

[0042] In a further embodiment, in step 6, adaptive shape analysis is performed on the cross-section point cloud data, and adaptive shape fitting is performed to obtain the shape and size parameters of the fitted tunnel cross section, including:

[0043] Step 6.1, using the weighted least squares method to fit the circular section based on the cross-section point cloud data, and determining the circular radius and coordinates through iterative optimization;

[0044] Step 6.2: Estimate the directions of the major axis and minor axis of the ellipse, as well as the initial length, based on the covariance matrix of the cross-section point cloud data, and determine the major axis and minor axis of the ellipse through iterative optimization.

[0045] In a further embodiment, in step 7, based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation assessment of the tunnel section is performed to determine the magnitude and direction of the deformation change, including:

[0046] For the tunnel sections at the same location scanned at different times, the section parameters are extracted in the same way: for circular sections, the coordinates of the center of the circle and the radius at different times are recorded; for elliptical sections, the length of the major axis, the length of the minor axis, the center coordinates, and the angle between the major axis and the coordinate axis at different times are recorded;

[0047] For circular sections, the radius change differences and circular coordinate change differences at different periods are compared with the preset circular section difference threshold. If the circular section difference threshold is exceeded, it is determined that the circular tunnel section has global deformation.

[0048] For the elliptical section, the difference in the change of the major axis or minor axis length and the difference in the change of the major axis angle at different periods are compared with the preset elliptical section difference threshold. If the elliptical section difference threshold is exceeded, it is determined that the elliptical tunnel section has global deformation.

[0049] In a further embodiment, in step 8, the tunnel section is locally divided, and local deformation is monitored according to changes in the divided sub-areas, including:

[0050] For a circular section, the circular section is divided into several sectors with equal angles. For each sector, the local parameters of the sector are calculated: point cloud density and average radius. Then, the point cloud density changes and average radius changes of the same sector in different periods are compared. If the point cloud density change rate exceeds the set density threshold or the average radius change exceeds the radius deviation threshold, it is judged that the sector has local deformation.

[0051] For each circular section, the number of points in the sector area is counted as the point cloud density, and the average distance from the point to the center of the circle is calculated as the average radius.

[0052] In a further embodiment, for an elliptical cross section, the elliptical cross section is divided into a plurality of rectangular sub-regions along the major axis and the minor axis, and in each rectangular sub-region, the distribution characteristics of the point cloud are calculated; then, the distribution characteristics of the point cloud in the same rectangular sub-region at different times are compared, and if the distribution characteristics of the point cloud exceed a set deviation threshold, it is determined that the rectangular sub-region has local deformation;

[0053] Among them, for each elliptical section, the standard deviation of the coordinates of the points in the major and minor axes is calculated as the distribution feature of the point cloud, indicating the degree of discreteness in the major and minor axes. By comparing the changes in the standard deviation of the major and minor axes in the same rectangular sub-area at different times, if the change in the major axis standard deviation or the minor axis standard deviation exceeds the set deviation threshold, it is determined that there is local deformation in the rectangular sub-area.

[0054] The tunnel section recognition and deformation detection method based on laser scanning point cloud data of the above embodiment of the present invention directly processes the point cloud data in three-dimensional space by polar coordinate conversion and local calculation weighted average to extract the central axis, so as to completely retain the three-dimensional information of the point cloud data, more accurately reflect the real spatial form of the tunnel, and avoid the problem of loss of geometric feature information of the top and bottom of the tunnel caused by projection. On this basis, the tunnel section is intercepted and the section point cloud data is determined, and global and local deformation discrimination is performed in combination with adaptive shape fitting to achieve accurate monitoring and discrimination of tunnel section deformation, which can keenly capture the possible slight deformation of the tunnel in the local area. Even when the overall deformation is not obvious, potential safety hazards can be discovered in advance, and all-round and accurate monitoring and discrimination of tunnel section deformation can be achieved, providing strong technical support for the long-term safe operation of the tunnel.

[0055] At the same time, the method based on polar coordinates and local weighted average proposed in the present invention can calculate the local central axis according to the natural distribution of point clouds in the curved area, and through the subsequent global central axis fitting and optimization steps, it can better adapt to the changes of complex-shaped tunnels such as curved and non-circular shapes. For tunnels with multi-layer structures or irregular cavities inside, since the present invention directly processes point cloud data in three-dimensional space, it can take into account the distribution of point clouds at different levels and around the cavities. By adjusting parameters such as the scope and weight of local calculations, it can more flexibly obtain the central axis that can represent the complex tunnel structure, control the accuracy of central axis extraction, and have better applicability for tunnels of complex shapes, providing a more reliable basis for subsequent section identification and deformation analysis.

[0056] Compared with the prior art, the present invention has the following significant advantages:

[0057] (1) The method of the present invention directly processes point cloud data in three-dimensional space, avoiding the problem of losing vertical direction (depth information) when projecting point cloud data onto a horizontal plane as in traditional methods. The present invention can fully utilize the three-dimensional geometric features of point cloud data to perform operations such as tunnel centerline extraction, cross-section recognition, and deformation detection, thereby more accurately reflecting the real spatial form of the tunnel. The method combines the use of local weighted average and curve fitting after polar coordinate conversion, rather than relying solely on local features such as high and low points, and comprehensively utilizing the overall distribution of point cloud data, which can better mine useful information from the data when processing complex tunnel structures and irregular point cloud distributions;

[0058] (2) In the process of extracting the central axis, the weighted average method is used to consider factors such as the distance from the point to the polar angle center ray, thereby reducing the influence of abnormal points (such as noise points caused by scanning equipment errors, environmental interference, or irregular points caused by local unevenness of the tunnel wall and auxiliary facilities) on the extraction of the central axis, making the acquisition of the central axis more robust.

[0059] (3) For section recognition, the accuracy and reliability of section recognition can be improved by accurately determining the section interception plane (based on the tangent direction vector of the central axis) and adaptive fitting methods (such as weighted least squares fitting of circular and elliptical sections). This can reduce the fitting error caused by interference factors in the point cloud data (such as noise and local irregular shapes).

[0060] It should be understood that all combinations of the aforementioned concepts and the additional concepts described in more detail below can be considered as part of the inventive subject matter of the present disclosure as long as such concepts are not mutually inconsistent. In addition, all combinations of the claimed subject matter are considered as part of the inventive subject matter of the present disclosure.

[0061] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The accompanying drawings are not intended to be drawn to scale. In the accompanying drawings, each identical or nearly identical component shown in each figure may be represented by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings.

[0063] Figure 1 It is a flow chart of a method for tunnel section recognition and deformation detection based on laser scanning point cloud data according to an embodiment of the present invention.

[0064] Figure 2 It is a schematic diagram of a process of obtaining a tunnel centerline based on polar coordinate conversion and processing according to an embodiment of the present invention.

[0065] Figure 3 It is a schematic diagram of a process of globally optimizing a preliminary central axis according to an embodiment of the present invention.

[0066] Figure 4 It is a schematic diagram of the global deformation determination process of a tunnel section according to an embodiment of the present invention.

[0067] Figure 5 It is a schematic diagram of a process flow for determining local deformation of a tunnel section according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.

[0069] Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed by the present invention are not limited to any implementation. In addition, some aspects disclosed by the present invention can be used alone or in any appropriate combination with other aspects disclosed by the present invention.

[0070] {Example 1}

[0071] Combination Figure 1 As shown, according to an embodiment of the present invention, a tunnel section recognition and deformation detection method based on laser scanning point cloud data comprises the following steps:

[0072] Step 1: Obtain point cloud data of any two adjacent sites in the tunnel, complete the registration of point cloud data of all laser scanning sites through point cloud registration, and obtain tunnel point cloud data;

[0073] Step 2: De-noise the point cloud data of the entire tunnel and perform spatial simplification of the point cloud data to obtain voxelized point cloud data;

[0074] Step 3: Extract the local central axis points of the tunnel based on polar coordinate transformation, and obtain the central axis of the tunnel after global central axis fitting optimization;

[0075] Step 4: Based on the obtained tunnel centerline as a reference, determine a plane perpendicular to the centerline according to the spacing as a section intercepting plane, and calculate the section intercepting plane equation;

[0076] Step 5: Segment the points in the point cloud data that are within a certain distance range on both sides of each cross-section interception plane as the original point cloud data of the cross-section interception plane, and sort and number them after removing isolated points to obtain cross-section point cloud data;

[0077] Step 6: Perform adaptive shape analysis on the cross-section point cloud data and perform adaptive shape fitting to obtain the shape and size parameters of the fitted tunnel cross section;

[0078] Step 7: Based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation of the tunnel section is evaluated to determine the magnitude and direction of the deformation change;

[0079] Step 8: Locally divide the tunnel section and monitor the local deformation according to the changes of the divided sub-areas.

[0080] In a further embodiment, in step 1, the registration of point cloud data of all laser scanning sites is completed by point cloud registration to obtain tunnel point cloud data, including:

[0081] Step 1.1, based on the feature points, solve the rigid transformation rotation matrix and translation vector between the stations, transform the point cloud data, and complete the rough registration;

[0082] Step 1.2: Based on the rough registration, optimization is performed based on the ICP algorithm, using a point-to-point distance measurement method, and using the LM algorithm to iteratively optimize and solve the transformation matrix to minimize the error function.

[0083] In a further embodiment, in step 2, the entire tunnel point cloud data is subjected to denoising and point cloud data space simplification to obtain voxelized point cloud data, including:

[0084] Step 2.1, statistical filtering and radius filtering are performed on the tunnel point cloud data to remove the marked noise points;

[0085] Step 2.2, divide the point cloud data space into voxel grids, retain only one representative point for each voxel grid, and take the point closest to the center of the voxel grid as the representative point to obtain voxelized point cloud data.

[0086] In a further embodiment, in combination Figure 2 As shown, in step 3, the local central axis point of the tunnel is extracted based on polar coordinate transformation, and the central axis of the tunnel is obtained after global central axis fitting optimization, including:

[0087] Step 3.1, convert the voxelized point cloud data from the Cartesian coordinate system to the polar coordinate system, and determine the polar diameter and polar angle of any point in the point cloud data;

[0088] Step 3.2, in each polar angle interval, select points within a certain range with the polar angle as the center and expand to the left and right, and calculate the local central axis point according to the polar diameter of the selected points, and use weighted average to calculate the polar diameter of the local central axis point;

[0089] Step 3.3, convert the local central axis points calculated in each polar angle interval back to the Cartesian coordinate system to obtain a set of discrete local central axis point sets;

[0090] Step 3.4, using the spline curve fitting method to perform curve fitting on the local central axis point set, and obtain a smooth curve as the preliminary central axis of the tunnel;

[0091] Step 3.5: Optimize the preliminary central axis to obtain the tunnel central axis.

[0092] In a further embodiment, in each polar angle interval, points within a certain range are selected with the polar angle as the center and the polar diameters of the selected points are respectively extended to the left and right, and the local central axis point is calculated according to the polar diameters of the selected points, and the polar diameters of the local central axis point are calculated by weighted average, including:

[0093] In each polar angle interval, with the polar angle as the center, the left and right expansions are Δθ ext , choose Δθ ext For points within the extended range, the polar diameter of the local central axis point is calculated using weighted average:

[0094] Assume that Δθ ext Any point P within the extended range i , whose polar diameter is ρ i , whose weight ω i According to point P i The distance to the polar angle center ray is determined by:

[0095]

[0096] Where, d i Indicates point P i The distance to the polar angle center ray, ∈ represents a minimum constant to prevent the denominator from being zero;

[0097] Polar diameter of the local midline point The calculation formula is:

[0098]

[0099] In a further embodiment, in combination Figure 3 As shown, the preliminary central axis is optimized to obtain the tunnel central axis, including:

[0100] For each point in the point cloud data within the set boundary range around the preliminary central axis, the distance deviation between each point and the preliminary central axis is calculated;

[0101] For points whose distance deviation exceeds the set distance threshold, they are marked as deviation points;

[0102] Divide the preliminary central axis and its surrounding point cloud data into local areas at equal length intervals;

[0103] Count the number of deviation points marked in each local area, and mark the local area where the number of deviation points exceeds the proportion as a problem segment;

[0104] For the central axis segment of the marked problem segment, take the two end points of the problem segment as the center, expand to both sides by a certain distance, and select point cloud data within a certain width in the direction perpendicular to the central axis as the candidate point cloud for recalculating the local central axis;

[0105] Within the selected candidate point cloud range, the weighted average is used to recalculate the local central axis segment and replace the central axis segment of the problem segment to update the preliminary central axis;

[0106] The central axis segments of all problem segments are updated, and the endpoints of the positions are smoothly replaced to obtain the central axis of the tunnel.

[0107] In a further embodiment, in combination Figure 4 As shown, in step 6, adaptive shape analysis is performed on the cross-section point cloud data, and adaptive shape fitting is performed to obtain the shape and size parameters of the fitted tunnel cross section, including:

[0108] Step 6.1, using the weighted least squares method to fit the circular section based on the cross-section point cloud data, and determining the circular radius and coordinates through iterative optimization;

[0109] Step 6.2: Estimate the directions of the major axis and minor axis of the ellipse, as well as the initial length, based on the covariance matrix of the cross-section point cloud data, and determine the major axis and minor axis of the ellipse through iterative optimization.

[0110] In a further embodiment, in combination Figure 4 As shown, in step 7, based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation of the tunnel section is evaluated to determine the magnitude and direction of the deformation change, including:

[0111] For the tunnel sections at the same location scanned at different times, the section parameters are extracted in the same way: for circular sections, the coordinates of the center of the circle and the radius at different times are recorded; for elliptical sections, the length of the major axis, the length of the minor axis, the center coordinates, and the angle between the major axis and the coordinate axis at different times are recorded;

[0112] For circular sections, the radius change differences and circular coordinate change differences at different periods are compared with the preset circular section difference threshold. If the circular section difference threshold is exceeded, it is determined that the circular tunnel section has global deformation.

[0113] For the elliptical section, the difference in the change of the major axis or minor axis length and the difference in the change of the major axis angle at different periods are compared with the preset elliptical section difference threshold. If the elliptical section difference threshold is exceeded, it is determined that the elliptical tunnel section has global deformation.

[0114] In a further embodiment, in combination Figure 5 As shown, in step 8, the tunnel section is locally divided, and the local deformation is monitored according to the changes of the divided sub-areas, including:

[0115] For a circular section, the circular section is divided into several sectors with equal angles. For each sector, the local parameters of the sector are calculated: point cloud density and average radius. Then, the point cloud density changes and average radius changes of the same sector in different periods are compared. If the point cloud density change rate exceeds the set density threshold or the average radius change exceeds the radius deviation threshold, it is judged that the sector has local deformation.

[0116] For each circular section, the number of points in the sector area is counted as the point cloud density, and the average distance from the point to the center of the circle is calculated as the average radius.

[0117] In a further embodiment, for an elliptical cross section, the elliptical cross section is divided into a plurality of rectangular sub-regions along the major axis and the minor axis, and in each rectangular sub-region, the distribution characteristics of the point cloud are calculated; then, the distribution characteristics of the point cloud in the same rectangular sub-region at different times are compared, and if the distribution characteristics of the point cloud exceed a set deviation threshold, it is determined that the rectangular sub-region has local deformation;

[0118] Among them, for each elliptical section, the standard deviation of the coordinates of the points in the major and minor axes is calculated as the distribution feature of the point cloud, indicating the degree of discreteness in the major and minor axes. By comparing the changes in the standard deviation of the major and minor axes in the same rectangular sub-area at different times, if the change in the major axis standard deviation or the minor axis standard deviation exceeds the set deviation threshold, it is determined that there is local deformation in the rectangular sub-area.

[0119] {Example 2}

[0120] In this embodiment, the tunnel section recognition and deformation detection method based on laser scanning point cloud data of the aforementioned embodiment of the present invention is further described in conjunction with specific examples.

[0121] 1. Original data collection

[0122] In this embodiment, a 3D laser scanner is selected, such as the high-precision FARO Focus 3D series, with a scanning accuracy of ±2mm and an angular resolution of 0.01°. The scanning range and resolution are reasonably set according to the length and width of the tunnel to ensure that sufficiently detailed point cloud data can be obtained.

[0123] Set up a scanning station at a certain distance (such as 20-30 meters) in the tunnel, and perform 360° full-scale scanning at each station. When scanning, ensure that the 3D laser scanner is placed horizontally, fixed with a tripod and adjusted to a stable state. Record the coordinate information of each station, including the X, Y, and Z axis coordinates, for subsequent data alignment.

[0124] The scanned point cloud data is transmitted and stored in the computer system in LAS format, and classified and stored according to date, site and other information to form a point cloud database.

[0125] (II) Point cloud data registration

[0126] In this embodiment, a coarse registration + fine registration method is adopted, based on a mixture of feature points and an improved ICP algorithm, to achieve high-precision registration.

[0127] As an example, the process of coarse registration is as follows:

[0128] In the point cloud data of adjacent sites, reliable feature points are selected through manual or automatic feature extraction algorithms (such as curvature-based features). Taking the selected n feature points as an example, for feature point i, its coordinates in site 1 are (x 1i ,y 1i , z 1i ), the coordinates in site 2 are (x 2i ,y 2i , z 2i ).

[0129] The initial rigid transformation matrix R from site 1 to site 2 is solved using the singular value decomposition (SVD) method. 0 and the translation vector T 0 The calculation process is as follows:

[0130] First calculate the centroid of the feature points:

[0131]

[0132] Then calculate the relative coordinates:

[0133]

[0134] Construct the covariance matrix H:

[0135]

[0136] Perform SVD decomposition on the covariance matrix H: H = U∑V T , then the rotation matrix R 0 =VU T , the translation vector

[0137] The point cloud data of site 1 is transformed by the rotation matrix and translation vector to complete the rough registration.

[0138] As an example, the process of fine registration is as follows:

[0139] Based on the rough registration, the ICP algorithm is used to further optimize the registration result. For the point P (x, y, z) after rough registration in site 1, find its nearest point Q (x', y', z') in the point cloud of site 2;

[0140] Define the error function:

[0141]

[0142] Where m is the number of points involved in the registration.

[0143] The point-to-point distance measurement method is adopted, and the Levenberg-Marquardt algorithm is used to optimize the transformation matrix R and T to minimize the error function. During the iteration process, the iteration termination condition is set, such as when the transformation matrix change between two adjacent iterations is less than a certain threshold or the number of iterations exceeds a certain value (such as 50 times), the iteration ends.

[0144] (III) Point cloud data denoising

[0145] In this step, the tunnel point cloud data is denoised and the point cloud data space is simplified to obtain voxelized point cloud data.

[0146] Firstly, statistical filtering and radius filtering are performed on the tunnel point cloud data to remove the marked noise points.

[0147] As an example, the statistical filtering process is as follows:

[0148] Calculate the average distance d between each point and its neighbors i , the neighborhood radius is set to r 1 (For example, 0.2 meters). For a point P in the point cloud data, its neighborhood point set is N(P), and the average distance calculation formula is:

[0149]

[0150] Then, according to the overall distribution of the point cloud data, the distance threshold D is set 1 , if the average distance d of a point iGreater than D 1 , it is initially marked as a noise point.

[0151] As an example, the radius filter process is as follows:

[0152] Based on statistical filtering, radius filtering is performed: for each point, take it as the center and set the radius to r 2 A sphere (such as 0.1 meters) is formed, and the number of points in the sphere is counted. If the counted number is less than a set threshold N (N can be determined according to the density of the tunnel point cloud, such as N=5), the point is marked as a noise point.

[0153] Then the two filtering results are combined to remove the marked noise points.

[0154] Finally, based on the noise-removed point cloud data, the point cloud data space is divided into voxel grids. For each voxel grid, only one representative point is retained, and the point closest to the center of the voxel grid is used as the representative point to obtain voxelized point cloud data.

[0155] (IV) Extraction of tunnel central axis

[0156] Combination Figure 3 As shown, the local central axis point of the tunnel is extracted based on polar coordinate transformation, and the central axis of the tunnel is obtained after global central axis fitting optimization, including:

[0157] Step 3.1, convert the voxelized point cloud data from the Cartesian coordinate system to the polar coordinate system, and determine the polar diameter and polar angle of any point in the point cloud data;

[0158] Step 3.2, in each polar angle interval, select points within a certain range with the polar angle as the center and expand to the left and right, and calculate the local central axis point according to the polar diameter of the selected points, and use weighted average to calculate the polar diameter of the local central axis point;

[0159] Step 3.3, convert the local central axis points calculated in each polar angle interval back to the Cartesian coordinate system to obtain a set of discrete local central axis point sets;

[0160] Step 3.4, using the spline curve fitting method to perform curve fitting on the local central axis point set, and obtain a smooth curve as the preliminary central axis of the tunnel;

[0161] Step 3.5: Optimize the preliminary central axis to obtain the tunnel central axis.

[0162] In step 3.1, for any point P in the voxelized point cloud data, P = (x, y, z), its polar coordinates are converted to:

[0163] Pole diameter:

[0164] Polar Angle:

[0165] Then, in order to reduce the discreteness of the data, the polar coordinate data is discretized, and the polar angle range [0, 2π] is divided into a certain angle interval (such as 0.5°) to form multiple polar angle intervals.

[0166] Further, in step 3.2, within each polar angle interval, points within a certain range are selected with the polar angle as the center and the polar diameter of the selected points is used to calculate the local central axis point, and the polar diameter of the local central axis point is calculated by weighted average, including:

[0167] In each polar angle interval, with the polar angle as the center, the left and right expansions are Δθ ext (In this example, the value is 2°), select ext For points within the extended range, the polar diameter of the local central axis point is calculated using weighted average:

[0168] Assume that Δθ ext Any point P within the extended range i , whose polar diameter is ρ i , whose weight ω i According to point P i The distance to the polar angle center ray is determined by:

[0169]

[0170] Where, d i Indicates point P i The distance to the polar angle center ray, ∈ represents a minimum constant to prevent the denominator from being zero;

[0171] Polar diameter of the local midline point The calculation formula is:

[0172]

[0173] In the example of the present invention, in step 3.4, cubic spline curve fitting is adopted to obtain a smooth curve as the preliminary central axis of the tunnel.

[0174] Combination Figure 3 As shown, in step 3.5, the preliminary central axis is optimized to obtain the tunnel central axis, including:

[0175] For each point in the point cloud data within the set boundary range around the preliminary central axis, the distance deviation between each point and the preliminary central axis is calculated; for example, since the preliminary central axis is composed of a series of discrete points, the distance from any point to the nearest discrete point of the central axis is calculated using the Euclidean distance as the distance deviation;

[0176] For points whose distance deviation exceeds the set distance threshold (such as 0.1m), they are marked as deviation points;

[0177] Divide the preliminary central axis and its surrounding point cloud data into local areas at equal length intervals;

[0178] Count the number of deviation points marked in each local area, and mark the local area where the number of deviation points exceeds a ratio (such as a ratio setting of 20%) as a problem segment;

[0179] For the central axis segment of the marked problem segment, take the two end points of the problem segment as the center, expand a certain distance (for example, 2m) to both sides, and select point cloud data within a certain width (for example, 1m) in the direction perpendicular to the central axis as the candidate point cloud for recalculating the local central axis;

[0180] Within the selected candidate point cloud range, the weighted average is used to recalculate the local central axis segment and replace the central axis segment of the problem segment to update the preliminary central axis;

[0181] The central axis segments of all problem segments are updated, and the endpoints of the positions are smoothly replaced to obtain the central axis of the tunnel.

[0182] In this way, the preliminary central axis is optimized so that it better reflects the actual shape of the tunnel and provides an accurate benchmark for subsequent tunnel section identification and deformation detection.

[0183] (V) Point cloud segmentation and sorting

[0184] The points in the point cloud data that are within a certain distance range (such as ±0.2 meters) on both sides of each interception plane are segmented out as the original section point cloud data of the section.

[0185] The segmented original cross-section point cloud data is further processed to remove isolated points that are too far from the plane (determined by setting a comparison distance, such as 0.15 meters). These isolated points may be noise or incorrectly segmented points. The remaining point clouds are sorted and numbered.

[0186] (VI) Adaptive shape fitting of cross-section point cloud data

[0187] In this step, adaptive shape fitting is performed on the cross-section point cloud data to obtain the shape and size parameters of the fitted tunnel cross section.

[0188] As an optional embodiment, typical ellipse and circle in tunnel design are taken as examples for fitting.

[0189] Combination Figure 4 As shown in FIG. 1 , the process of adaptive shape fitting of cross-section point cloud data includes:

[0190] Step 6.1, using the weighted least squares method to fit the circular section based on the cross-section point cloud data, and determining the circular radius and coordinates through iterative optimization;

[0191] Step 6.2: Estimate the directions of the major axis and minor axis of the ellipse, as well as the initial length, based on the covariance matrix of the cross-section point cloud data, and determine the major axis and minor axis of the ellipse through iterative optimization.

[0192] Further, in step 6.1, a weighted least square method is used to fit a circular section based on the cross-section point cloud data, and the circular radius and coordinates are determined by iterative optimization, including:

[0193] Obtain the cross-section point cloud data after preprocessing and segmentation, recorded as set S;

[0194] Initial center estimation: Calculate the center of gravity coordinates of cross-section point cloud data As the initial center estimate, the calculation formula is:

[0195]

[0196] Distance calculation and weight assignment: For each point in the set S, calculate its distance d to the initial center of the circle i , the calculation formula is:

[0197]

[0198] According to the distance d i The Gaussian weight function is used to assign weights to each point, namely:

[0199]

[0200] Among them, σ is a weight parameter, which can be adjusted according to the distribution of point cloud data and the requirements of fitting accuracy. Generally speaking, the larger σ is, the greater the weight of distant points is; the smaller σ is, the greater the weight of nearby points is;

[0201] Weighted Least Squares Fit: Use weighted least squares to fit the equation of a circle. Suppose the equation of the circle is:

[0202] (x-α) 2 +(yb) 2 =r 2 ;

[0203] The error function is defined as:

[0204]

[0205] Use the Levenberg-Marquardt algorithm to solve the center coordinates and radius of the circle by minimizing the error function;

[0206] Iterative optimization: Finally, the center coordinates and radius obtained by the Levenberg-Marquardt algorithm are used as new initial values, and the weight distribution and weighted least squares fitting are repeated until the changes in the center coordinates and radius are less than the set threshold (such as 10 -6 ), or the maximum number of iterations (such as 50) is reached, the center coordinates and radius obtained at this time are the final fitting results.

[0207] Further, in step 6.2, a weighted least square method is used to fit a circular section based on the section point cloud data, and the circular radius and coordinates are determined by iterative optimization, including:

[0208] Initial ellipse parameter estimation: Calculate the centroid coordinates of point cloud data As an initial estimate of the center of the elliptical section, the calculation method is the same as the calculation process of the centroid coordinates in the circular section mentioned above;

[0209] Long axis and short axis direction estimation: The direction of the long axis and short axis is estimated by calculating the covariance matrix of the point cloud data, specifically:

[0210] Calculate the covariance matrix C:

[0211] Perform eigenvalue decomposition on the covariance matrix C and obtain two eigenvalues ​​λ 1 , 2 and the corresponding eigenvector Then the major axis direction and the eigenvector Parallelism, minor axis direction and eigenvectors parallel;

[0212] Long axis and short axis length estimation: Estimate the length of the long axis and short axis according to the projection length of the point cloud data in the long axis and short axis directions. Project the point cloud data in the long axis and short axis directions, calculate the range of the projection points respectively, and obtain the initial length estimation value l of the long axis and short axis. 0 and 0 ;

[0213] Convert the point cloud data from the original coordinate system to a new coordinate system with the center of the ellipse as the origin and the major and minor axes as the coordinate axes: i (x i ,y i ), its coordinate in the new coordinate system (x′ i , y′ i )for:

[0214]

[0215] In the new coordinate system, for each point P′ i (x′ i , y′i ), calculate its distance to the center of the ellipse, the calculation formula is:

[0216]

[0217] Similarly, according to the distance d′ i Use Gaussian weight function to assign weight w to each point i :

[0218]

[0219] Then the ellipse equation is fitted, and the equation of the ellipse in the new coordinate system is:

[0220]

[0221] Define the error function E(a,b) as:

[0222]

[0223] The error function is minimized through a nonlinear optimization algorithm to solve the major semi-axis a and the minor semi-axis b of the ellipse.

[0224] Finally, the fitted ellipse parameters are converted from the new coordinate system back to the original coordinate system to obtain the center coordinates, major and minor axis lengths, and angle parameters between the major axis and the coordinate axis of the ellipse section in the original coordinate system. Then, using these parameters as new initial values, the coordinate transformation-weight distribution-ellipse fitting is repeated until the parameter change is less than the set threshold, and the final ellipse section fitting result is obtained.

[0225] (VII) Determination of global deformation of cross section

[0226] Combination Figure 4 As shown, in step 7, based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation of the tunnel section is evaluated to determine the magnitude and direction of the deformation change, including:

[0227] For the tunnel sections at the same location scanned at different times, the section parameters are extracted in the same way: for circular sections, the coordinates of the center of the circle and the radius at different times are recorded; for elliptical sections, the length of the major axis, the length of the minor axis, the center coordinates, and the angle between the major axis and the coordinate axis at different times are recorded;

[0228] For circular sections, the radius change differences and circular coordinate change differences at different periods are compared with the preset circular section difference threshold. If the circular section difference threshold is exceeded, it is determined that the circular tunnel section has global deformation.

[0229] For the elliptical section, the difference in the change of the major axis or minor axis length and the difference in the change of the major axis angle at different periods are compared with the preset elliptical section difference threshold. If the elliptical section difference threshold is exceeded, it is determined that the elliptical tunnel section has global deformation.

[0230] (VIII) Determination of global deformation of cross section

[0231] Combination Figure 5 As shown, in step 8, the tunnel section is locally divided, and the local deformation is monitored according to the changes of the divided sub-areas, including:

[0232] For a circular section, the circular section is divided into several sectors with equal angles. For example, each sector is divided into 12 sectors with a central angle of 30°. For each sector, the local parameters of the sector are calculated: point cloud density and average radius. Then, the point cloud density changes and average radius changes of the same sector in different periods are compared. If the point cloud density change rate exceeds the set density threshold or the average radius change exceeds the radius deviation threshold, it is judged that the sector has local deformation.

[0233] For each circular section, the number of points in the sector area is counted as the point cloud density, and the average distance from the point to the center of the circle is calculated as the average radius.

[0234] In a further embodiment, for an elliptical cross section, the elliptical cross section is divided into a plurality of rectangular sub-regions along the major axis and the minor axis, and in each rectangular sub-region, the distribution characteristics of the point cloud are calculated; then, the distribution characteristics of the point cloud in the same rectangular sub-region at different times are compared, and if the distribution characteristics of the point cloud exceed a set deviation threshold, it is determined that the rectangular sub-region has local deformation;

[0235] Among them, for each elliptical section, the standard deviation of the coordinates of the points in the major and minor axes is calculated as the distribution feature of the point cloud, indicating the degree of discreteness in the major and minor axes. By comparing the changes in the standard deviation of the major and minor axes in the same rectangular sub-area at different times, if the change in the major axis standard deviation or the minor axis standard deviation exceeds the set deviation threshold, it is determined that there is local deformation in the rectangular sub-area.

[0236] On this basis, the number and distribution of areas with local deformation can be further counted. If the local deformation areas are concentrated in a certain direction or in a continuous area, it is necessary to further analyze the cause, whether it is caused by local uneven force or changes in geological conditions. At the same time, the local deformation information is combined with the global deformation information to comprehensively evaluate the deformation state of the tunnel section. For example, if the global parameters change little, but there are multiple areas with obvious deformation locally, it means that the tunnel structure has shown early signs of damage locally and needs to be paid special attention.

[0237] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. A tunnel section recognition and deformation detection method based on laser scanning point cloud data, characterized in that: include: Step 1: Obtain point cloud data of any two adjacent sites in the tunnel, complete the registration of point cloud data of all laser scanning sites through point cloud registration, and obtain tunnel point cloud data; Step 2: De-noise the point cloud data of the entire tunnel and perform spatial simplification of the point cloud data to obtain voxelized point cloud data; Step 3: Extract the local central axis points of the tunnel based on polar coordinate transformation, and obtain the central axis of the tunnel after global central axis fitting optimization; Step 4: Based on the obtained tunnel centerline as a reference, determine a plane perpendicular to the centerline according to the spacing as a section intercepting plane, and calculate the section intercepting plane equation; Step 5: Segment the points in the point cloud data that are within a certain distance range on both sides of each cross-section interception plane as the original point cloud data of the cross-section interception plane, and sort and number them after removing isolated points to obtain cross-section point cloud data; Step 6: Perform adaptive shape analysis on the cross-section point cloud data and perform adaptive shape fitting to obtain the shape and size parameters of the fitted tunnel cross section; Step 7: Based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation of the tunnel section is evaluated to determine the magnitude and direction of the deformation change; Step 8: Locally divide the tunnel section and monitor the local deformation according to the changes of the divided sub-areas.

2. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 1 is characterized in that: In step 1, the registration of point cloud data of all laser scanning sites is completed by point cloud registration to obtain tunnel point cloud data, including: Step 1.1, based on the feature points, solve the rigid transformation rotation matrix and translation vector between the stations, transform the point cloud data, and complete the rough registration; Step 1.2: Based on the rough registration, optimization is performed based on the ICP algorithm, using a point-to-point distance measurement method, and using the LM algorithm to iteratively optimize and solve the transformation matrix to minimize the error function.

3. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 1 is characterized in that: In step 2, the entire tunnel point cloud data is denoised and the point cloud data space is simplified to obtain voxelized point cloud data, including: Step 2.1, statistical filtering and radius filtering are performed on the tunnel point cloud data to remove the marked noise points; Step 2.2, divide the point cloud data space into voxel grids, retain only one representative point for each voxel grid, and take the point closest to the center of the voxel grid as the representative point to obtain voxelized point cloud data.

4. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 1 is characterized in that: In step 3, the local central axis point of the tunnel is extracted based on polar coordinate transformation, and the central axis of the tunnel is obtained after global central axis fitting optimization, including: Step 3.1, convert the voxelized point cloud data from the Cartesian coordinate system to the polar coordinate system, and determine the polar diameter and polar angle of any point in the point cloud data; Step 3.2, in each polar angle interval, select points within a certain range with the polar angle as the center and expand to the left and right, and calculate the local central axis point according to the polar diameter of the selected points, and use weighted average to calculate the polar diameter of the local central axis point; Step 3.3, convert the local central axis points calculated in each polar angle interval back to the Cartesian coordinate system to obtain a set of discrete local central axis point sets; Step 3.4, using the spline curve fitting method to perform curve fitting on the local central axis point set, and obtain a smooth curve as the preliminary central axis of the tunnel; Step 3.5: Optimize the preliminary central axis to obtain the tunnel central axis.

5. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 4 is characterized in that: In each polar angle interval, points within a certain range are selected from the left and right sides of the polar angle as the center, and the local central axis point is calculated according to the polar diameter of the selected point, and the polar diameter of the local central axis point is calculated by weighted average, including: In each polar angle interval, with the polar angle as the center, the left and right expansions are Δθ ext , choose Δθ ext For points within the extended range, the polar diameter of the local central axis point is calculated using weighted average: Assume that Δθ ext Any point P within the extended range i , whose polar diameter is ρ i , whose weight ω i According to point P i The distance to the polar angle center ray is determined by: Where, d i Indicates point P i The distance to the polar angle center ray, ∈ represents a minimum constant to prevent the denominator from being zero; Polar diameter of the local midline point The calculation formula is:

6. The method for tunnel section recognition and deformation detection based on laser scanning point cloud data according to claim 4, characterized in that: The step of optimizing the preliminary central axis to obtain the tunnel central axis includes: For each point in the point cloud data within the set boundary range around the preliminary central axis, the distance deviation between each point and the preliminary central axis is calculated; For points whose distance deviation exceeds the set distance threshold, they are marked as deviation points; Divide the preliminary central axis and its surrounding point cloud data into local areas at equal length intervals; Count the number of deviation points marked in each local area, and mark the local area where the number of deviation points exceeds the proportion as a problem segment; For the central axis segment of the marked problem segment, take the two end points of the problem segment as the center, expand to both sides by a certain distance, and select point cloud data within a certain width in the direction perpendicular to the central axis as the candidate point cloud for recalculating the local central axis; Within the selected candidate point cloud range, the weighted average is used to recalculate the local central axis segment and replace the central axis segment of the problem segment to update the preliminary central axis; The central axis segments of all problem segments are updated, and the endpoints of the positions are smoothly replaced to obtain the central axis of the tunnel.

7. The method for tunnel section recognition and deformation detection based on laser scanning point cloud data according to claim 1, characterized in that: In step 6, adaptive shape analysis is performed on the cross-section point cloud data, and adaptive shape fitting is performed to obtain the shape and size parameters of the fitted tunnel cross section, including: Step 6.1, using the weighted least squares method to fit the circular section based on the cross-section point cloud data, and determining the circular radius and coordinates through iterative optimization; Step 6.2: Estimate the directions of the major axis and minor axis of the ellipse, as well as the initial length, based on the covariance matrix of the cross-section point cloud data, and determine the major axis and minor axis of the ellipse through iterative optimization.

8. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 1 is characterized in that: In step 7, based on the shape and size parameters of the tunnel section at the same position obtained by scanning at different times, the deformation assessment of the tunnel section is performed to determine the magnitude and direction of the deformation change, including: For the tunnel sections at the same location scanned at different times, the section parameters are extracted in the same way: for circular sections, the coordinates of the center of the circle and the radius at different times are recorded; for elliptical sections, the length of the major axis, the length of the minor axis, the center coordinates, and the angle between the major axis and the coordinate axis at different times are recorded; For circular sections, the radius change differences and circular coordinate change differences at different periods are compared with the preset circular section difference threshold. If the circular section difference threshold is exceeded, it is determined that the circular tunnel section has global deformation. For the elliptical section, the difference in the change of the major axis or minor axis length and the difference in the change of the major axis angle at different periods are compared with the preset elliptical section difference threshold. If the elliptical section difference threshold is exceeded, it is determined that the elliptical tunnel section has global deformation.

9. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 1 is characterized in that: In the step 8, the tunnel section is locally divided, and the local deformation is monitored according to the changes of the divided sub-areas, including: For a circular section, the circular section is divided into several sectors with equal angles. For each sector, the local parameters of the sector are calculated: point cloud density and average radius. Then, the point cloud density changes and average radius changes of the same sector in different periods are compared. If the point cloud density change rate exceeds the set density threshold or the average radius change exceeds the radius deviation threshold, it is judged that the sector has local deformation. For each circular section, the number of points in the sector area is counted as the point cloud density, and the average distance from the point to the center of the circle is calculated as the average radius.

10. The tunnel section recognition and deformation detection method based on laser scanning point cloud data according to claim 1, characterized in that: For an elliptical cross section, the elliptical cross section is divided into multiple rectangular sub-regions along the major and minor axes. In each rectangular sub-region, the distribution characteristics of the point cloud are calculated. Then, the distribution characteristics of the point cloud in the same rectangular sub-region at different times are compared. If the distribution characteristics of the point cloud exceed the set deviation threshold, it is determined that the rectangular sub-region has local deformation. Among them, for each elliptical section, the standard deviation of the coordinates of the points in the major and minor axes is calculated as the distribution feature of the point cloud, indicating the degree of discreteness in the major and minor axes. By comparing the changes in the standard deviation of the major and minor axes in the same rectangular sub-area at different times, if the change in the major axis standard deviation or the minor axis standard deviation exceeds the set deviation threshold, it is determined that there is local deformation in the rectangular sub-area.

Images