Shield tunnel segment dislocation measurement method based on three-dimensional laser scanning

Abstract

The present invention relates to a shield tunnel segment dislocation measurement method based on three-dimensional laser scanning, mainly solving the technical problems of low efficiency, high cost, unsafe high-altitude operation, etc., of existing manual shield tunnel segment dislocation amount measurement technology. The method comprises the steps of: S1, generating a grayscale image on the basis of tunnel laser point cloud data, and performing edge detection on the grayscale image; S2, on the basis of the edge-detected image, collecting statistics about column grayscale cumulative values of the edge-detected image, and realizing circumferential joint position identification by means of threshold determination; S3, selecting a certain angle range on the left and right of a circumferential joint to divide different intervals, collecting statistics about an interval densest radial vector, and calculating a circumferential joint dislocation amount; S4, designing a feature convolution kernel on the basis of shape features of a longitudinal joint, and using the feature convolution kernel to perform convolution calculation on a segment grayscale image in a sliding window manner to identify the position of the longitudinal joint; and S5, selecting point cloud data within a certain range on the left and right of the longitudinal joint, screening for local valid points, and calculating a longitudinal joint dislocation amount. The present invention achieves automatic identification of a longitudinal joint of shield tunnel segments and rapid measurement of a dislocation amount.

Description

A shield tunnel segment misalignment measurement method based on 3D laser scanning Technical Field

[0001] The invention relates to a method for measuring the misalignment of tunnel segments, and in particular discloses a method for measuring the misalignment of shield tunnel segments based on three-dimensional laser scanning. Background Art

[0002] With the rapid development of urban rail transit, the number of tunnels constructed using the shield method has increased significantly. Shield tunnels offer the advantages of fast construction and minimal impact on urban traffic. However, over long-term operation, shield tunnel segments are susceptible to misalignment due to environmental fluctuations and operational vibrations. To ensure safe tunnel operation, regular or irregular inspections of shield tunnel segment misalignment are necessary.

[0003] Traditional shield tunnel segment misalignment measurement relies primarily on manual labor, with inspection technicians using a ruler to directly measure. As rail transit operating hours increase, maintenance window time shrinks, and the workload for shield tunnel inspection increases, manual inspection methods are no longer sufficient to meet the needs of misalignment detection. Chinese patent application number 201811089324.1, "Method for Extracting Shield Tunnel Lining Misalignment Based on Point Cloud Data," uses mutations in point clouds between and within rings to identify and calculate inter- and intra-ring misalignment. However, this method requires a station-type scanner combined with a total station for data acquisition, resulting in low data acquisition efficiency and failing to meet engineering requirements. Chinese patent application number 201510662198.4, "Method for Detecting Segment Misalignment in Subway Tunnel," utilizes tunnel depth images to detect misalignment. However, this method is subject to limitations in pixel resolution and depth accuracy, which in turn affects the accuracy of misalignment calculations. Current manual detection techniques for shield tunnel segment misalignment suffer from low efficiency, high cost, and unsafe working at height.

[0004] The development of a non-contact, efficient and accurate method for detecting shield tunnel segment misalignment has important theoretical significance and practical application value for the detection of shield tunnel structural defects.

[0005] Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a shield tunnel segment misalignment measurement method based on 3D laser scanning, which mainly solves the technical problems of the current manual detection technology of shield tunnel segment misalignment, such as low efficiency, high cost, and unsafe high-altitude operation. To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A shield tunnel segment misalignment measurement method based on three-dimensional laser scanning, characterized by comprising the following steps:

[0008] S1. Generate a grayscale image based on the tunnel laser point cloud data and perform edge detection on the grayscale image;

[0009] S2. Count the grayscale cumulative values ​​of the image after edge detection and identify the annular seam position through threshold judgment;

[0010] S3. Select a certain angle range around the annular seam to divide it into different intervals, count the most dense radial vectors in the intervals, and calculate the annular seam stagger amount;

[0011] S4. Design a feature convolution kernel based on the morphological characteristics of the longitudinal seam, and perform convolution calculation on the grayscale image of the pipe segment using the feature convolution kernel in a sliding window manner to identify the location of the longitudinal seam;

[0012] S5. Select the point cloud data within a certain range on the left and right of the longitudinal seam, filter out the local effective points, and calculate the longitudinal seam misalignment.

[0013] (1) The specific operations of step S1 are as follows:

[0014] Grayscale image generation and edge detection contrast enhancement processing. The grayscale image is obtained by interpolating the grayscale values ​​of the tunnel 3D laser scanning points. Edge detection retains the original grayscale values ​​of the identified edge points, and the grayscale values ​​of the remaining background are unified to 255;

[0015] 1) Establish a rectangular coordinate system with the laser scanner lens as the origin O, the horizontal right direction as the positive X-axis, and the vertical upward direction as the positive Y-axis;

[0016] 2) Calculate the angle of each scanning point in each ring relative to the origin O Vertically downward is the starting angle, counterclockwise is positive, where i is the ring number and j is the sequence number of the ring point cloud;

[0017] 3) Let the height of the grayscale image I0 be H and the width be W, then the new index of the scan point on the grayscale image is N ij , where i∈[1,W],j∈[1,H], the calculation formula is:

[0018] 4) Traverse the index of grayscale image I0, if there is N ij , then N ij The gray value at the position is set to the point cloud gray value G(N ij ), if there is no N ij , then N i(j-1) The gray value at the position is set to the point cloud gray value G(N i(j-1) );

[0019] 5) The generated grayscale image I0 is processed by the Soble edge detection operator. The original grayscale values ​​of the identified edge points are retained, and the grayscale values ​​of the remaining background are uniformly set to 255. The grayscale image obtained after edge detection is recorded as I1.

[0020] (2) The specific operations of step S2 are as follows:

[0021] The annular seam position is identified based on the column grayscale accumulation value. The calculated image is a grayscale image processed by edge detection, and the annular seam position is identified based on the threshold value.

[0022] 1) Statistical grayscale image I1 each column grayscale cumulative value G j , where i∈[1, W], the calculation formula is:

[0023] Where V start ∈[1,W],V End ∈[1,W], is the starting and ending positions of eliminating pipeline influence;

[0024] 2) Divide the grayscale image into different intervals N every N columns i , i∈[1, W / N], calculation interval N i The average grayscale accumulation value The calculation formula is:

[0025] 3) Basis Determine the segment annular gap judgment threshold in each interval, where P c is the constraint coefficient;

[0026] 4) Comparison interval N i The column grayscale accumulation value and the segment annular gap judgment threshold Size relationship, if Then column j is the annular crack anomaly location and is included in the set S;

[0027] 5) Traverse the set S and introduce the ring gap width control threshold W d , based on distance clustering, cluster the columns in set S into set S i , i is the number of cluster sets;

[0028] 6) Calculate the set S i The average value of the middle annular gap column is used as the annular gap position.

[0029] (3) The specific operations of step S3 are as follows:

[0030] The calculation of annular joint misalignment is based on the interval's densest radius vector. A certain angle range is taken on the left and right rings of the annular joint to divide the point cloud into different intervals. The average radius vector of the interval is calculated. The interval radius vector with the smallest difference in interval radius vector is counted as the interval's densest radius vector, and the annular joint misalignment is calculated.

[0031] 1) Using the least squares method of algebraic distance to fit the scanning ring to an ellipse;

[0032] 2) Let the ring gap ID number be i, intercept n scanning rings on the left and right sides of the ring gap, plus the i-th ring, a total of 2n+1 rings: in≤I L ≤i+n Formula (4)

[0033] 3) Establish two coordinate systems with the ellipse fitting center point O1 and the scanner center point O2 as the centers, with the horizontal right direction as the positive direction of the X axis, the vertical upward direction as the positive direction of the Y axis, and counterclockwise as positive;

[0034] 4) Select 5 calculation points on the circumference of O2, and the angles between them and the x-axis are θ j =0°, 45°, 90°, 135°, 180°;

[0035] 5) Set the point angle θ in the O2 coordinate system j Convert to angle in O1 coordinate system The conversion formula is as follows:

[0036] Where: R is the design radius of the tunnel, d is the vertical distance between O1 and O2;

[0037] 6) Take The point cloud within the range calculates the most dense vector radius of the left segment interval and the most dense radial vector between the right segment The specific calculation steps of the interval's densest radius vector are:

[0038] (a) The point cloud is divided into M intervals and the average radius vector value r in the interval is calculated. i , where i∈[1,M],r i The calculation formula is:

[0039] Where: N is the number of points in the interval, r j The calculation formula is:

[0040] Where: x o1 and y o1 are the x-axis and y-axis coordinates of the ellipse fitting center point O1, x j and y j are the x-axis and y-axis coordinates of point j respectively;

[0041] (b) Set the interval average radius vector threshold γ, cluster the average radius vector values ​​in the interval according to the distance clustering, and cluster them into set D i , i is the number of cluster sets;

[0042] (c) Select the set with the largest number of clusters as the set of the densest radius vector interval, and calculate its average value as the densest radius vector of the interval;

[0043] 7) That is the amount of segment annular seam misalignment.

[0044] (4) The specific operations of step S4 are as follows:

[0045] Based on feature convolution combined with sliding window for longitudinal seam identification and location, a feature convolution kernel is designed according to the morphological characteristics of oblique seams and straight seams. The feature convolution kernel is used to convolve the grayscale image of the pipe segment in a sliding window manner to identify the longitudinal seam location.

[0046] 1) Construct a feature convolution kernel based on the morphological features of the oblique seams and straight seams. There are two types of oblique seam feature convolution kernels: the main diagonal feature convolution kernel and the sub-diagonal feature convolution kernel. The vertical seam convolution kernel is specifically:

[0047] (a) Main diagonal convolution kernel K Pri :

[0048] Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H1, and V mn Let V be the value of the mth row and nth column of the convolution kernel. jk =1, where and Convolution kernel K Pir The rest of the values ​​are 0;

[0049] (b) Subdiagonal convolution kernel K Sub :

[0050] Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H1, and V mn Let V be the value of the mth row and nth column of the convolution kernel. jk =1, where and Convolution kernel K pirThe rest of the values ​​are 0;

[0051] (c) Horizontal convolution kernel K Str :

[0052] Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H2, and all values ​​in the horizontal convolution kernel are set to 1;

[0053] 2) Use the sliding window method to convolve the main diagonal convolution kernel K Pir , sub-diagonal convolution kernel K Sub , horizontal convolution kernel K Str The corresponding column interval in the grayscale image [ID i , ID i+1 ] to perform convolution operation, and the convolution results are Where j1, j2∈[1, H-H1+1], j3∈[1, H-H2+1], H is the width of the grayscale image;

[0054] 3) Set T c =255*W i *R c is the horizontal longitudinal seam judgment threshold within the segment row interval, where R c is the constraint coefficient;

[0055] 4) Traverse the horizontal convolution results and compare them with the horizontal longitudinal seam judgment threshold. If Then record the number of rows j3 to the set S i middle;

[0056] 5) From the convolution result Eliminate set S i The number of rows and convolution results of [j3, j3+H2] in the matrix are recorded as

[0057] 6) Traverse and eliminate the results Select the number of rows j1 with the minimum value in the elimination result min and j2 min , which is the location of the two longitudinal seams;

[0058] 7) Set S i Remove j3 from the adjacent segments, and the result is set S′ i , set S′ i The j3 value in is the location of the remaining longitudinal seams.

[0059] (5) The specific operations of step S5 are as follows:

[0060] The longitudinal seam stagger amount is calculated based on the local effective point vector radius. The point cloud data within a certain range on the left and right of the longitudinal seam are selected, the local noise points are eliminated, and the local effective point vector radius is calculated to realize the calculation of the longitudinal seam stagger amount.

[0061] 1) Select point cloud data P0 and P1 on the left and right sides of the longitudinal seam within the range of ε from the longitudinal seam;

[0062] 2) Calculate the segment point cloud radius vector value, which is recorded as set E0 and E1 respectively;

[0063] 3) Select local effective points. The specific operations are as follows:

[0064] (a) Find peaks and troughs in a finite set of points A. Let the number of elements in set A be N, and let the set be P:

[0065] (b) When the midpoint of set A satisfies one of formulas (8), (9) and (10), it is a peak or valley point: i=0 Formula (8) (r i -r i-1 )(r i -r i+1 )>0 Formula (9) i=N-1 Formula (10)

[0066] Where i is the ID number of each point, r i is the radius vector of point i.

[0067] (c) Calculate the adjacent points P i , P i+1 Radius vector difference D i , if D i ≥δ Formula (11)

[0068] Then P i , P i+1 The point cloud data is cleared and the remaining points are local valid points;

[0069] 4) Calculate the average radius vector on the left and right sides based on the radius vector of the local effective point and

[0070] 5) That is the longitudinal seam misalignment.

[0071] Compared with the existing technology, the shield tunnel segment misalignment measurement method based on three-dimensional laser scanning provided by the present invention has the following effects:

[0072] (1) The present invention uses the image features of three-dimensional laser scanning technology and the positional relationship of point cloud data to identify the joints of shield tunnel segments and quantify the misalignment, which can achieve non-contact rapid measurement with high computational efficiency;

[0073] (2) The present invention automatically identifies longitudinal seams based on their morphological characteristics by using a longitudinal seam feature convolution kernel combined with a sliding window approach. Compared with deep learning detection methods, the detection speed is faster.

[0074] (3) The present invention combines the misalignment calculation method of the interval densest radius vector and the local effective radius vector to reduce the influence of point cloud noise on the misalignment and improve the robustness of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] FIG1 is a flow chart of a shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to the present invention;

[0076] FIG2 is a schematic diagram of the point cloud angle conversion between the ellipse fitting coordinate system and the scanner coordinate system;

[0077] Figure 3 is a schematic diagram of the calculation of annular seam misalignment;

[0078] Figure 4 is a schematic diagram of longitudinal seam feature convolution combined with sliding window longitudinal seam detection;

[0079] FIG5 is a schematic diagram of selecting local effective points;

[0080] Figure 6 is a diagram showing the results of shield tunnel annular seam identification;

[0081] Figure 7 is a diagram showing the calculation results of the shield tunnel annular joint misalignment;

[0082] Figure 8 is the calculation result of the longitudinal misalignment of the shield tunnel. DETAILED DESCRIPTION

[0083] The present invention will be further described in detail below with reference to the accompanying drawings and examples. The following embodiments are intended to explain the present invention but the present invention is not limited to the following embodiments.

[0084] As shown in FIG1 , a shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to the present invention is characterized by comprising the following steps:

[0085] S1. Grayscale image generation and edge detection contrast enhancement processing. The grayscale image is obtained by interpolating the grayscale values ​​of the tunnel 3D laser scanning points. Edge detection retains the original grayscale values ​​of the identified edge points, and the grayscale values ​​of the remaining background are unified to 255.

[0086] The specific operations of step S1 are as follows:

[0087] (1) Establish a rectangular coordinate system with the laser scanner lens as the origin O, the horizontal right direction as the positive direction of the X axis, and the vertical upward direction as the positive direction of the Y axis;

[0088] (2) Calculate the angle of each scanning point in each ring relative to the origin O Vertically downward is the starting angle, counterclockwise is positive, where i is the ring number and j is the sequence number of the ring point cloud;

[0089] (3) Assume that the height of the grayscale image I0 is H and the width is W, then the new index of the scan point on the grayscale image is N ij , where i∈[1,W],j∈[1,H] is calculated as follows:

[0090] (4) Traverse the index of grayscale image I0, if there is N ij , then N ij The gray value at the position is set to the point cloud gray value G(N ij ), if there is no N ij , then N i(j-1) The gray value at the position is set to the point cloud gray value G(N i(j-1) );

[0091] (5) The generated grayscale image I0 is processed by the Soble edge detection operator. The original grayscale values ​​of the identified edge points are retained, and the grayscale values ​​of the remaining background points are uniformly set to 255. The grayscale image obtained after edge detection is recorded as I1;

[0092] The kernel size of the Soble edge detection operator is 3, and the detection threshold is 25.

[0093] S2. Identify the annular seam position based on the column grayscale cumulative value. The calculated image is a grayscale image after edge detection processing. The annular seam position is identified based on the threshold value.

[0094] The specific operations of step S2 are as follows:

[0095] (1) Statistical grayscale image I1 grayscale cumulative value G of each column j , where j∈[1, W], the calculation formula is:

[0096] Where V start ∈[1,W],V End ∈[1, W], is the starting and ending positions of eliminating pipeline influence;

[0097] (2) Divide the grayscale image into different intervals N every N columns i , i∈[1, W / N], calculation interval N i The average grayscale accumulation value The calculation formula is:

[0098] (3) Basis Determine the segment annular gap judgment threshold in each interval, where P c is the constraint coefficient, which is set to 0.6 based on practical experience;

[0099] (4) Comparison interval N i The column grayscale accumulation value and the segment annular gap judgment threshold Size relationship, if Then column j is the annular crack anomaly location and is included in the set S;

[0100] (5) Traverse the set S and introduce the ring gap width control threshold W d , W d According to actual experience, it is set to 5, and the columns in set S are clustered into set S based on distance clustering. i , i is the number of cluster sets;

[0101] (6) Calculate the set S i The average value of the annular gap column is used as the annular gap position. The identification results are shown in Figure 6.

[0102] S3. Calculate the annular joint misalignment based on the interval's densest radius vector. Divide the point cloud into different intervals within a certain angle range on the left and right rings of the annular joint. Calculate the average radius vector of the interval. Count the interval radius vectors with the smallest difference as the interval's densest radius vector to calculate the annular joint misalignment.

[0103] As shown in FIG3 , the specific operations of step S3 are as follows:

[0104] (1) Use the least square method of algebraic distance to fit the scanning ring to an ellipse;

[0105] (2) Let the ring gap ID be i, and intercept n scanning rings on the left and right sides of the ring gap, plus the i-th ring, for a total of n+1 rings: in≤I L ≤i+n Formula (4)

[0106] (3) Two sets of coordinate systems are established with the ellipse fitting center point O1 and the scanner center point O2 as the centers, with the horizontal right direction as the positive direction of the X axis, the vertical upward direction as the positive direction of the Y axis, and the counterclockwise direction as positive;

[0107] (4) As shown in Figure 2, 5 calculation points are selected on the circumference of O2, and the angles between them and the x-axis are s ＝ 0°, 45°, 90°, 135°, 180°;

[0108] (5) The point angle θ in the O2 coordinate system j Convert to angle in O1 coordinate system The conversion formula is as follows:

[0109] Where: R is the design radius of the tunnel, d is the vertical distance between O1 and O2;

[0110] (6) Take The point cloud within the range calculates the most dense vector radius of the left segment interval and the most dense radial vector between the right segment The specific calculation steps of the interval's densest radius vector are:

[0111] (a) The point cloud is divided into M intervals and the average radius vector value r in the interval is calculated. i , where i∈[1,M],r i The calculation formula is:

[0112] Where: N is the number of points in the interval, r j The calculation formula is:

[0113] Where: x o1 and y o1 are the x-axis and y-axis coordinates of the ellipse fitting center point O1, x j and y j are the x-axis and y-axis coordinates of point j respectively.

[0114] (b) Set the interval average radius vector threshold γ, cluster the average radius vector values ​​in the interval according to the distance clustering, and cluster them into set D i , i is the number of cluster sets;

[0115] (c) Select the set with the largest number of clusters as the set of the densest radius vector interval, and calculate its average value as the densest radius vector of the interval;

[0116] According to experience, the preferred values ​​of δ, M, and γ are 1°, 10, and 0.005, respectively.

[0117] (7) This is the segment annular seam misalignment; see the table below for measurement and calculation:

[0118] The calculation results are shown in Figure 7.

[0119] S4. Based on feature convolution combined with sliding window, longitudinal seam identification and positioning, a feature convolution kernel is designed according to the morphological characteristics of oblique seams and straight seams. The feature convolution kernel is used to convolve the grayscale image of the pipe segment in a sliding window manner to identify the longitudinal seam position.

[0120] As shown in FIG4 , the specific operations of step S4 are as follows:

[0121] (1) Construct a feature convolution kernel based on the morphological characteristics of the oblique seam and the straight seam. There are two types of oblique seam feature convolution kernels: the main diagonal feature convolution kernel and the sub-diagonal feature convolution kernel. The vertical seam convolution kernel is specifically:

[0122] (a) Main diagonal convolution kernel K Pir:

[0123] Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H1, and V mn Let V be the value of the mth row and nth column of the convolution kernel. jk =1, where and Convolution kernel K Pir The rest of the values ​​are 0;

[0124] (b) Subdiagonal convolution kernel K Sub :

[0125] Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H1, and V mn Let V be the value of the mth row and nth column of the convolution kernel. jk =1, where and Convolution kernel K Pir The rest of the values ​​are 0;

[0126] (c) Horizontal convolution kernel K Str :

[0127] Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H2, and all values ​​in the horizontal convolution kernel are set to 1;

[0128] If the length and width of the main diagonal convolution kernel and the sub-diagonal convolution kernel are set to 12, and the length and width of the horizontal convolution kernel are set to 8, then the main diagonal convolution kernel, the sub-diagonal convolution kernel, and the horizontal convolution kernel are shown in formulas (8), (9), and (10):

[0129] (2) Use the sliding window method to convolve the main diagonal convolution kernel K Pir , sub-diagonal convolution kernel K Sub , horizontal convolution kernel K Str The corresponding column interval in the grayscale image [ID i ,IDi+1 ] to perform convolution operation, and the convolution results are Where j1, j2∈[1,H-H1+1], j3∈[1,H-H2+1], H is the width of the grayscale image;

[0130] (3) Set T c =255*W i *R c is the horizontal longitudinal seam judgment threshold within the segment row interval, where R c is the constraint coefficient, which is preferably 0.65 based on experience;

[0131] (4) Traverse the horizontal convolution results and compare them with the horizontal longitudinal seam judgment threshold. If Then record the number of rows j3 to the set S i middle;

[0132] (5) From the convolution result Eliminate set S i The number of rows and convolution results of [j3, j3+H2] in the matrix are recorded as

[0133] (6) Traverse and eliminate results Select the number of rows j1 with the minimum value in the elimination result min and j2 min , which is the location of the two longitudinal seams;

[0134] (7) Set S i Remove j3 from the adjacent segments, and the result is set S′ i , set S′ i The j3 value in is the location of the remaining longitudinal seams.

[0135] S5. Calculation of longitudinal seam misalignment based on local effective radius vector: select point cloud data within a certain range on the left and right of the longitudinal seam, remove local noise points, and calculate local effective point radius vector to achieve calculation of longitudinal seam misalignment.

[0136] The specific operations of step S5 are as follows:

[0137] (1) Select point cloud data P0 and P1 on the left and right sides of the longitudinal seam within the range of ε from the longitudinal seam;

[0138] (2) Calculate the segment point cloud radius vector value, which is recorded as set E0 and E1 respectively;

[0139] (3) As shown in Figure 5, the specific operation of selecting local effective points is:

[0140] (a) Find peaks and troughs in a finite set of points A. Let the number of elements in set A be N, and let the set be P:

[0141] (b) When the midpoint of set A satisfies one of formulas (11), (12) and (13), it is a peak or trough point: i=0 Formula (11) (r i -r i-1 )(r i -r i+1 )>0 Formula (12) i=N-1 Formula (13)

[0142] Where i is the ID number of each point, r i is the radius vector of point i;

[0143] (c) Calculate the adjacent points P i , P i+1 Radius vector difference D i , if D i ≥δ Formula (14)

[0144] Then P i , P i+1 The point cloud data is cleared and the remaining points are local valid points;

[0145] Based on actual experience, the value of δ is 0.003.

[0146] (4) Calculate the average radius vector on the left and right sides based on the radius vector of the local effective point and

[0147] (5) This is the longitudinal seam stagger; the measurement and calculation are shown in the table below:

[0148] The calculation results are shown in Figure 8.

[0149] The above describes preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection claimed by the claims of the present invention.

Claims

1. A shield tunnel segment misalignment measurement method based on 3D laser scanning, characterized by: The following steps are involved: S1. Generate a grayscale image based on the tunnel laser point cloud data, and perform edge detection on the grayscale image; S2. Count the grayscale cumulative values of the image after edge detection, and identify the position of the annular seam by threshold judgment; S3, selecting a certain angle range to the left and right of the annular seam to divide it into different intervals, counting the densest radius vectors in the intervals, and calculating the annular seam stagger; S4. Designing a feature convolution kernel based on the morphological characteristics of the longitudinal seam, and performing convolution calculation on the grayscale image of the pipe segment using the feature convolution kernel in a sliding window manner to identify the location of the longitudinal seam; S5. Select the point cloud data within a certain range on the left and right of the longitudinal seam, filter out local valid points, and calculate the longitudinal seam misalignment.

2. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 1 is characterized in that: The step S1 comprises: (1) Establish a rectangular coordinate system with the laser scanner lens as the origin O, the horizontal right direction as the positive direction of the X axis, and the vertical upward direction as the positive direction of the Y axis; (2) Calculate the angle of the scanning point in each ring relative to the origin O Vertically downward is the starting angle, counterclockwise is positive, where i is the ring number and j is the sequence number of the ring point cloud; (3) Let the height of the grayscale image I0 be H and the width be W, then the new index of the scanning point on the grayscale image is N ij , where i∈[1,W],j∈[1,H] is calculated as follows: (4) Traverse the index of the grayscale image I0, if there is N ij , then N ij The gray value at the position is set to the point cloud gray value G(N ij ), if N does not exist ij , then Ni (j-1) The gray value at the position is set to the point cloud gray value G(N i(j-1) ); (5) The generated grayscale image I0 is processed by the Soble edge detection operator, the original grayscale values of the identified edge points are retained, and the grayscale values of the remaining background are uniformly set to 255. The grayscale image obtained after edge detection is recorded as I1.

3. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 1 is characterized in that: The step S2 comprises: (1) Statistical grayscale image I1 grayscale cumulative value G of each column j , where j∈[1, W], the calculation formula is: Where V start ∈[1,W],V End ∈[1,W], is the starting and ending positions of eliminating pipeline influence; (2) Divide the grayscale image into different intervals N every N columns i , i∈[1, W / N], calculation interval N i The average grayscale accumulation value The calculation formula is: (3) Basis Determine the segment annular gap judgment threshold in each interval, where P c is the constraint coefficient; (4) Comparison interval N i The column grayscale accumulation value and the segment annular gap judgment threshold Size relationship, if Then column j is the annular crack anomaly location and is included in the set S; (5) Traverse the set S and introduce the ring gap width control threshold W d , based on distance clustering, cluster the columns in set S into set S i , i is the number of cluster sets; (6) Calculate the set S i The average value of the middle annular gap column is used as the annular gap position.

4. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 1 is characterized in that: The step S3 comprises: (1) Use the least square method of algebraic distance to fit the scanning ring to an ellipse; (2) Let the ring gap ID number be i, and intercept n scanning rings on the left and right sides of the ring gap, plus the i-th ring, for a total of 2n+1 rings: in≤I L ≤i+n Formula (4) (3) Two sets of coordinate systems are established with the ellipse fitting center point O1 and the scanner center point O2 as the centers, with the horizontal right direction as the positive direction of the X axis, the vertical upward direction as the positive direction of the Y axis, and the counterclockwise direction as positive; (4) Select 5 calculation points on the circumference of O2, and the angles between them and the X axis are θ j =0°, 45°, 90°, 135°, 180°; (5) The point angle θ in the O2 coordinate system j Convert to angle in O1 coordinate system The conversion formula is as follows: Where R is the design radius of the tunnel, d is the vertical distance between O1 and O2, (6) Take The point cloud within the range calculates the interval most dense vector radius of the left segment and the interval with the most dense radial vector of the right segment (7) That is the amount of annular seam misalignment.

5. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 4 is characterized in that: The steps for calculating the densest radius vector in the interval include: (a) The point cloud is divided into M intervals and the average radius vector value r in the interval is calculated. i , where i∈[1,M],r i The calculation formula is: In formula (6), N is the number of points in the interval, r j The calculation formula is: In formula (7), x o1 and y o1 are the x-axis and y-axis coordinates of the ellipse fitting center point O1, x j and y j are the x-axis and y-axis coordinates of point j respectively; (b) Set the interval average radius vector threshold γ, cluster the average radius vector values in the interval according to the distance clustering, and cluster them into set D i , i is the number of cluster sets; (c) Select the set with the largest number of clusters as the densest radius vector interval set, and calculate its average value as the densest radius vector of the interval.

6. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 1 is characterized in that: The step S4 comprises: (1) Construct a feature convolution kernel based on the morphological features of the oblique seam and the straight seam. There are two types of oblique seam feature convolution kernels, namely the main diagonal feature convolution kernel and the sub-diagonal feature convolution kernel. The longitudinal seam convolution kernel is specifically: (a) Main diagonal convolution kernel K Pri : Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H1, and V mn Let V be the value of the mth row and nth column of the convolution kernel. jk =1, where and Convolution kernel K Pir The rest of the values are 0; (b) Subdiagonal convolution kernel K Sub : Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the i+1th ring gap id No., height is H1, set V mn Let V be the value of the mth row and nth column of the convolution kernel. jk =1, where and Convolution kernel K Pir The rest of the values are 0; (c) Horizontal convolution kernel K Str : Width W i =ID i+1 -ID i , where ID i is the id number of the i-th ring seam, ID i+1 is the id number of the i+1th ring seam, the height is H2, and all values in the horizontal convolution kernel are set to 1; (2) Use the sliding window method to convolve the main diagonal convolution kernel K Pir , the sub-diagonal convolution kernel K Sub , the horizontal convolution kernel K Str The corresponding column interval [ID i , ID i+1 ] to perform convolution operation, and the convolution results are Where j1, j2∈[1,H-H1+1], j3∈[1,H-H2+1], H is the width of the grayscale image; (3) Set T c =255*W i *R c is the horizontal longitudinal seam judgment threshold within the segment row interval, where R c is the constraint coefficient; (4) Traverse the horizontal convolution results and compare them with the horizontal longitudinal seam judgment threshold. If Then record the number of rows j3 to the set S i middle; (5) From the convolution result Eliminate set S i The number of rows and convolution results of [j3, j3+H2] in the matrix are recorded as (6) Traverse the elimination results Select the row number j1 with the minimum value in the elimination result min and j2 min , which is the location of the two longitudinal seams; (7) Set S i Remove j3 from the adjacent segments, and the result is set S′ i , set S′ i The j3 value in is the location of the remaining longitudinal seams.

7. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 1 is characterized in that: The step S5 comprises: (1) Select point cloud data P0 and P1 on the left and right sides of the longitudinal seam within the range of ε from the longitudinal seam; (2) Calculate the segment point cloud radius vector value, which is recorded as set E0 and E1 respectively; (3) Select local effective points; (4) Calculate the average radius vector of the left and right sides based on the radius vector of the local effective point and (5) That is the longitudinal seam misalignment.

8. The shield tunnel segment misalignment measurement method based on three-dimensional laser scanning according to claim 7 is characterized in that: The step of selecting a local effective point comprises: (a) Find peaks and troughs in a finite set of points A. Let the number of elements in set A be N, and let the set be P: (b) When the midpoint of set A satisfies one of the formulas (8), (9) and (10), it is a peak or trough point: i=0 Formula (8) (r i -r i-1 )(r i -r i+1 )>0 formula (9) i=N-1 Formula (10) Where i is the ID number of each point, r i is the radius vector of point i; (c) Calculate the adjacent points P i , P i+1 Radius vector difference D i ,like D i ≥δ Formula (11) Then P i , P i+1 The point cloud data is cleared and the remaining points are local valid points.

Images