Shield tunnel lining defect detection method based on ground-based LiDAR point cloud

Abstract

The invention relates to the technical field of tunnel detection, and in particular to a shield tunnel lining defect detection method based on a ground-based LiDAR point cloud; based on a ground-based LiDAR tunnel lining laser point cloud, the point cloud is preprocessed to obtain a three-dimensional laser point cloud; the three-dimensional laser point cloud is projected to obtain a two-dimensional laser point cloud image, and the two-dimensional laser point cloud image is grayed to obtain a laser point cloud gray image; based on a local pre-fitting energy-driven active contour fast model, the laser point cloud gray image is segmented to detect and analyze tunnel lining defects. The method has the advantages of high accuracy and recall rate, low false alarm rate and area error, and short calculation running time in defect detection, and can quickly and accurately identify shield tunnel lining defects, and solves the problem that traditional subway shield tunnel defect monitoring can only obtain limited discrete monitoring point changes and cannot fully reflect the tunnel lining defect status.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel detection, and in particular to a shield tunnel lining defect detection method based on ground-based LiDAR point cloud. Background Art

[0002] The subway shield tunnel is an important urban infrastructure. With the increase of operation time and the changes in the environment inside and outside the tunnel, tunnel lining surface defects will inevitably begin to appear. In order to ensure the safe operation of the subway tunnel, it is necessary to detect tunnel lining defects and take timely remedial measures.

[0003] The traditional method of monitoring defects in subway shield tunnels is to use static sensors such as joint meters and inclinometers, as well as instruments such as total stations. These devices have a long monitoring cycle and can only obtain limited changes in discrete monitoring points, and cannot fully reflect the condition of tunnel lining defects. Summary of the invention

[0004] The purpose of the present invention is to provide a shield tunnel lining disease detection method based on ground-based LiDAR point cloud, aiming to solve the problem that traditional subway shield tunnel disease monitoring can only obtain limited discrete monitoring point changes and cannot fully reflect the tunnel lining disease status.

[0005] To achieve the above object, the present invention provides a shield tunnel lining defect detection method based on ground-based LiDAR point cloud, comprising the following steps:

[0006] Based on the ground-based LiDAR tunnel lining laser point cloud, the point cloud is preprocessed to obtain a three-dimensional laser point cloud;

[0007] Projecting the three-dimensional laser point cloud to obtain a two-dimensional laser point cloud image, and graying the two-dimensional laser point cloud image to obtain a laser point cloud grayscale image;

[0008] The laser point cloud grayscale image is segmented based on a local pre-fitting energy driven active contour fast model to detect and analyze tunnel lining defects.

[0009] The preprocessing includes removing noise points of the cross section by using an iterative ellipse fitting technique.

[0010] Wherein, the laser point cloud grayscale image includes the overall and local features of the image.

[0011] The two-dimensional laser point cloud image grayscale process mainly includes laser point cloud intensity correction and laser point cloud rasterization.

[0012] The tunnel lining diseases include tunnel lining water leakage and segment falling off.

[0013] The shield tunnel lining defect detection method based on ground-based LiDAR point cloud of the present invention is based on the ground-based LiDAR tunnel lining laser point cloud, and the point cloud is preprocessed to obtain a three-dimensional laser point cloud; the three-dimensional laser point cloud is projected to obtain a two-dimensional laser point cloud image, and the two-dimensional laser point cloud image is grayed to obtain a laser point cloud grayscale image; the laser point cloud grayscale image is segmented based on a local pre-fitting energy-driven active contour fast model to detect and analyze tunnel lining defects. The method is based on the ground-based LiDAR point cloud to expand the tunnel lining three-dimensional point cloud to generate The two-dimensional point cloud of the tunnel is formed, and the two-dimensional point cloud grayscale image of the tunnel lining surface is obtained according to the point cloud reflection intensity. The tunnel lining defects are detected based on the two-dimensional grayscale image and the local pre-fitting energy-driven active contour fast model (LPF image) segmentation algorithm. In terms of defect detection, it has the advantages of high accuracy and recall rate, small false alarm rate and area error, and less calculation running time. It can quickly and accurately identify the defects of the shield tunnel lining, and solve the problem that the traditional subway shield tunnel defect monitoring can only obtain limited discrete monitoring point changes and cannot fully reflect the tunnel lining defect status. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is the principle diagram of unfolding the 3D laser point cloud into a 2D plane point cloud.

[0016] Figure 2 It is a schematic diagram of the expanded point cloud of the tunnel lining surface.

[0017] Figure 3 It is a flow chart of shield subway tunnel lining defect detection based on ground-based LiDAR point cloud and LPF method.

[0018] Figure 4 It is a flow chart of a shield tunnel lining defect detection method based on ground-based LiDAR point cloud provided by the present invention. DETAILED DESCRIPTION

[0019] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0020] See also Figures 1 to 4 The present invention provides a shield tunnel lining defect detection method based on ground-based LiDAR point cloud, comprising the following steps:

[0021] S1 is based on the ground-based LiDAR tunnel lining laser point cloud, and pre-processes the point cloud to obtain a three-dimensional laser point cloud;

[0022] Specifically, due to the obstruction of obstacles or the limitation of the tunnel space scanning angle, noise data that does not belong to the tunnel lining often appears in the shield tunnel lining laser point cloud. If these noises are not eliminated, it will have a great impact on the detection and identification of lining defects. Since the tunnel cross section can be regarded as an ellipse close to a standard circle, the iterative ellipse fitting technology is used to remove the noise points of the cross section. The parameters during the ellipse fitting use the random sampling consistency algorithm, and the specific steps are as follows:

[0023] The least squares fitting algorithm of the ellipse is used to fit the noisy cross-sectional point cloud, so as to calculate the initial parameters of the ellipse (including the coefficients of the ellipse equation, the coordinates of the center of the circle, and the major and minor semi-axes).

[0024] Find the shortest distance d from the section point to the ellipse i , forming a distance point set d mean =d{d1,d2,…,d mean}, the mean of the point set d mean and standard deviation are calculated by equations (1) and (2) respectively.

[0025]

[0026]

[0027] Provisions when |d i -d mean |>2σ is a noise point. After a series of iterations, the points that are not on the tunnel wall are finally removed.

[0028] S2 projects the three-dimensional laser point cloud to obtain a two-dimensional laser point cloud image, and grayscales the two-dimensional laser point cloud image to obtain a laser point cloud grayscale image;

[0029] Specifically, the shield tunnel lining 3D laser point cloud is first transformed into a 2D plane point cloud by projection transformation, and then the 2D plane point cloud is transformed into a grayscale image according to the reflection intensity of the point cloud. The point cloud is projected onto the center O(X0, Z0) of the cylindrical section fitting after denoising. Figure 2 The small dashed line shows the fitting surface. is the positive unit vector of the X-axis, P=(x,z) is an arbitrary point in the 3D laser point cloud, P'=(x',z') is the point after P=(x,z) is projected onto the fitting surface, and the point where the fitting surface where P'=(x',z') is unfolded along the top is represented by P"=(x",z"). ∠POA=θ, which is and ∠P'OB=α, which is and The angle is assumed to be positive when it rotates counterclockwise around the positive direction of the X-axis, R is the inner diameter of the tunnel, and the Y coordinate of point P remains unchanged before and after the transformation, that is, y=y'=y". The principle of expanding a 3D laser point cloud into a 2D plane point cloud is as follows: Figure 1 As shown in Figure 2, the three-dimensional laser point cloud of the tunnel lining surface is unfolded according to this method, as shown in Figure 2. Figure 2 shown.

[0030] According to equations (3) and (4), the coordinates (x', z') of point P' after projection of point P can be obtained:

[0031]

[0032]

[0033] From equations (5) and (6), we can get the coordinates (x, z") of P' expansion P":

[0034]

[0035]

[0036] The information contained in the expanded laser point cloud grayscale image can fully reflect the overall and local features of the image. In order to more intuitively represent the tunnel lining image and detect and analyze the tunnel lining defects based on the LPF image segmentation algorithm, it is necessary to convert the two-dimensional laser point cloud into a grayscale value image. The grayscale conversion process of the two-dimensional laser point cloud image mainly includes laser point cloud intensity correction and laser point cloud rasterization. The laser point cloud intensity is a function of the laser reflection intensity and the measurement distance, and its mathematical model is shown in formula (7).

[0037]

[0038] Where I is the measured laser point cloud intensity, ρ is the target reflectivity, f(ρ) is the function of the target reflectivity, and α i is the polynomial for model estimation based on actual laser point cloud data, and D is the actual measured distance. The corrected laser point cloud intensity value is expressed as I c Its calculation model is shown in formula (8).

[0039]

[0040] Where Ds is the reference distance. Corrected laser point cloud intensity I c It is no longer related to the measured distance D, but only affected by the reflectivity ρ of the scanned target. Since the reflectivity of the tunnel lining defect (such as water leakage, shedding) area is different from that of other tunnel linings, it is easier to detect tunnel lining defects based on the intensity-corrected laser point cloud.

[0041] S3 segments the laser point cloud grayscale image based on a local pre-fitting energy-driven active contour fast model to detect and analyze tunnel lining defects.

[0042] Specifically, by calculating the local average image intensity Ω before the curve evolves x To define two pre-fitting functions f s (x) and f l (x). Active contour model of local pre-fitting energy: Let Ω∈R 2 is the image domain, I(x):Ω→R is a given grayscale image. Formula (10) defines the following function:

[0043]

[0044] In the formula, Ω x represents a small area with x as the center and radius w, Ω s ,Ω l Defined by formula (11):

[0045]

[0046] The energy functional based on the local pre-fitting function is integrated into the length constraint L(φ) and the distance regularization term P(φ) for smoothing and shortening the curve. For all points x in the image domain Ω, it is necessary to minimize E x The integral of the energy functional E LPF , E LPF (φ), L(φ) and P(φ) can be defined by equations (12), (13) and (14):

[0047]

[0048]

[0049]

[0050] (3) The fast gradient descent method is used to minimize the energy functional of the level set function φ. The calculation model is shown in formula (15):

[0051]

[0052] Among them, es (x) and e l (x) The calculation model is shown in formula (16):

[0053]

[0054] The specific implementation steps of the grayscale image obtained by expanding the three-dimensional laser point cloud of the subway tunnel lining surface and detecting defects such as water leakage and segment shedding based on the LPF method are as follows:

[0055] (1) Setting the initial level of the length term, setting function φ0, length term ν, time step Δt and other parameters;

[0056] (2) Calculate f according to equation (10) s (x) and f l (x);

[0057] (3) Go to the next point x=x+1, and then return to the second step until all points are calculated;

[0058] (4) Update the level set function φ according to equations (14) and (15) until the convergence criterion is reached.

[0059] In this model, the initial level set function φ0 is set to a small constant function φ0(x) = c0. The stopping criterion of the curve evolution is |(S i+5 -S i ) / S|<10 -5 , S represents the total area of ​​the image. i represents the area enclosed by the curve at time i.

[0060] What is disclosed above is only a preferred embodiment of the shield tunnel lining defect detection method based on the foundation LiDAR point cloud of the present invention. Of course, this cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A shield tunnel lining defect detection method based on ground-based LiDAR point cloud, characterized in that: The following steps are involved: Based on the ground-based LiDAR tunnel lining laser point cloud, the point cloud is preprocessed to obtain a three-dimensional laser point cloud; Projecting the three-dimensional laser point cloud to obtain a two-dimensional laser point cloud image, and graying the two-dimensional laser point cloud image to obtain a laser point cloud grayscale image; The laser point cloud grayscale image is segmented based on the local pre-fitting energy driven active contour fast model to detect and analyze tunnel lining defects, including: By calculating the local average image intensity Ω before the curve evolution x To define two pre-fitting functions f s (x) and f l (x), Active contour model with local pre-fitting energy: Let Ω∈R 2 is the image domain, I(x):Ω→R is a given grayscale image, and Equation (10) defines the following function: In the formula, Ω x represents a small area with x as the center and radius w, Ω s ,Ω l Defined by formula (11): Based on the energy functional of the local pre-fitting function, it is integrated into the length constraint term L(φ) and the distance regularization term P(φ) for smoothing and shortening the curve. For all points x in the image domain Ω, it is necessary to minimize E x The integral of the energy functional E LPF , E LPF (φ), L(φ) and P(φ) can be defined by equations (12), (13) and (14): The fast gradient descent method is used to minimize the energy functional of the level set function φ. The calculation model is shown in formula (15): Among them, e s (x) and e s (x) The calculation model is shown in formula (16): The specific implementation steps of the grayscale image obtained by expanding the three-dimensional laser point cloud of the subway tunnel lining surface and detecting defects such as water leakage and segment shedding based on the LPF method are as follows: (1) Setting the initial level of the length term, setting function φ0, length term ν, time step Δt and other parameters; (2) Calculate f according to equation (10) s (x) and f l (x); (3) Go to the next point x=x+1, and then return to the second step until all points are calculated; (4) Update the level set function φ according to equations (14) and (15) until the convergence criterion is reached. In this model, the initial level set function φ0 is set to a small constant function φ0(x) = c0, and the stopping criterion of the curve evolution is |(S i+5 -S i ) / S|<10 -5 , S represents the total area of ​​the image, S i represents the area enclosed by the curve at time i.

2. The shield tunnel lining defect detection method based on ground-based LiDAR point cloud according to claim 1, characterized in that: The preprocessing includes removing noise points of the cross section by using an iterative ellipse fitting technique.

3. The shield tunnel lining defect detection method based on ground-based LiDAR point cloud according to claim 1, characterized in that: The tunnel lining diseases include tunnel lining water leakage and segment shedding.

Images