Highway tunnel inner contour deformation analysis method

Abstract

The invention discloses a highway tunnel inner contour deformation analysis method, which comprises the steps of 1, acquiring point cloud data of a tunnel inner contour, and preprocessing the point cloud data to obtain n groups of preprocessed point cloud data of the tunnel inner contour; step 2, performing tunnel section ellipse fitting on one group of preprocessed point cloud data of the tunnel inner contour obtained in the step 1 to obtain a single tunnel section ellipse; by means of data preprocessing, normal distribution statistics and an improved least square method ellipse fitting mode, highway tunnel overall deformation monitoring based on tunnel inner wall point cloud slice coordinates is achieved, and the problem that a three-dimensional laser point cloud monitoring method is prone to being interfered by external factors (vibration of collection equipment and wrong collection of optical equipment) is solved. And interference point clouds (attachments of connecting bolt holes and nuts, grouting holes, cables, lighting equipment and the like on corresponding facilities in the tunnel, such as shield ring pieces, are difficult to remove are solved, and accurate and efficient monitoring of highway tunnel inner contour deformation analysis is realized.

Description

Technical Field

[0001] The invention belongs to the technical field of highway tunnel civil construction structure detection, and in particular is a method for analyzing contour deformation in a highway tunnel. Background Art

[0002] As important transportation channels, highway tunnels are prone to lateral and longitudinal deformation as they age. If not addressed promptly, this can seriously impact traffic safety. Therefore, deformation monitoring is a key component of regular highway tunnel inspections.

[0003] Current methods for monitoring deformation in highway tunnels primarily include traditional manual monitoring using total stations, GPS, distributed fiber optic sensors, and 3D laser point cloud monitoring. However, traditional manual monitoring using total stations and GPS is time-consuming, labor-intensive, and requires a harsh environment. Deploying distributed fiber optic sensors is costly and difficult, making it unsuitable for large-scale tunnel deformation monitoring. Conventional 3D laser point cloud-based monitoring methods are also susceptible to interference from external factors, as actual tunnel data often contains data from attached objects that need to be removed.

[0004] Therefore, it is urgent to propose a highway tunnel overall deformation monitoring algorithm based on the coordinates of tunnel inner wall point cloud slices to analyze the internal deformation of highway tunnels. This method can eliminate noise that occurs during the fitting process, reduce errors, enhance tunnel analysis, determine the overall deformation of the tunnel, and improve the accuracy of monitoring data. It is not affected by the density and number of point clouds. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the object of the present invention is to provide a method for analyzing the deformation of the inner contour of a highway tunnel, so as to solve the technical problem that the deformation analysis method in the prior art is not accurate enough.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A method for analyzing profile deformation in a highway tunnel, the method comprising the following steps:

[0008] Step 1: collecting point cloud data of the tunnel inner contour and preprocessing it to obtain n groups of preprocessed point cloud data of the tunnel inner contour;

[0009] The preprocessing includes sequentially performing gross error elimination and interference point removal;

[0010] The gross error elimination is used to eliminate point cloud data that exceeds the tunnel contour range and point cloud data collected under the wrong vehicle posture;

[0011] Step 2: performing tunnel cross-sectional ellipse fitting on one set of pre-processed point cloud data of the tunnel inner contour obtained in step 1 to obtain a single tunnel cross-sectional ellipse;

[0012]

[0013] in:

[0014] a represents the major semi-axis parameter of the tunnel cross-section ellipse;

[0015] b represents the minor semi-axis parameter of the tunnel section ellipse;

[0016] Step 3: Repeat steps 1 to 2 to obtain n tunnel cross-sectional ellipses, and connect the centers of the n tunnel cross-sectional ellipses to obtain the tunnel central axis;

[0017] Step 4: Compare the tunnel cross-section ellipse obtained in step 3 with the standard tunnel ellipse to determine whether deformation occurs and the magnitude of the deformation.

[0018] This method also includes the following technical features:

[0019] The specific process of removing gross errors includes the following steps:

[0020] S1, perform mean filtering on the point cloud data of the tunnel contour, and use a sliding window to divide it into multiple regions. Calculate the average value of the point cloud data of the tunnel contour in each region and use it as the sampling point of the corresponding region;

[0021] S2, creating multiple 3D voxel grids of the same size, and calculating the distance between each sampling point in the 3D voxel grid and the center of the 3D voxel grid, and selecting the sampling point corresponding to the minimum value of the distance value as the final sampling point;

[0022] S3, cuts the final sampling points obtained in S2 in three dimensions to obtain an ordered kd-tree;

[0023] S4, obtain the K nearest neighbors of each node on the ordered kd-tree through binary search;

[0024] S5, calculate the distance from each node to the K nearest neighboring points on the ordered kd-tree, and calculate the average R of the distances between each node and its K nearest neighboring points;

[0025] S6, count the number Q of final sampling points within the range of 2R from each node on the ordered kd-tree, and determine whether the number Q satisfies the following formula. If so, the node is an isolated point and is removed; if not, the process ends;

[0026] Q<2K.

[0027] S3 specifically includes the following steps:

[0028] S3.1, given the initial coordinate axis x, and use it as the normal vector of the root node splitting hyperplane;

[0029] S3.2, sort the final sampling points obtained in S2 on the initial coordinate axis x, select the median of the sort as the split point, and divide the space into two parts using a plane perpendicular to the coordinate axis. The split point is used as the parent node, and the final sampling points on its left and right sides are used as the left and right subtrees of the partition;

[0030] S3.3, select the next level coordinate axis and use it as the normal vector of the node splitting hyperplane;

[0031] S3.4, sort the final sampling points on the left subtree and the right subtree on the next level coordinate axis, select the median of the sorted points as the split point, and use a plane perpendicular to the coordinate axis to divide the space into two parts. The split point is used as the parent node, and the final sampling points on its left and right sides are used as the left subtree and the right subtree of the next level of division;

[0032] S3.5, repeat S3.3 to S3.4 until it cannot be split any further and an ordered kd-tree is obtained.

[0033] In step 1, the interference point removal includes the following steps:

[0034] Q1, divide the final sampling points into n groups along the tunnel direction, perform ellipse fitting on each of the n groups of final sampling points, and obtain n cross-sectional ellipses;

[0035] Q2, calculate the distance between the final sampling point of the i-th group and its corresponding cross-sectional ellipse, sort the distance from small to large, use the sequence number as the x-axis and the distance as the y-axis to obtain the corresponding fitting curve y i =g(x);

[0036] Q3, for the fitting curve y i = g(x) and take the derivative to get y i ', and then get y i 'The maximum corresponding distance value M i , and use it as the distance threshold of the corresponding cross-section ellipse;

[0037] Q4, determine whether the distance obtained in Q2 is less than the corresponding distance threshold. If so, retain the final sampling point; if not, remove the final sampling point, and obtain the i-th group of pre-processed point cloud data of the tunnel inner contour;

[0038] In Q5, let i = i + 1 and return to Q2 until i > n, and obtain n sets of pre-processed point cloud data of the tunnel inner contour.

[0039] Step 2 specifically includes the following steps:

[0040] Step 2.1, respectively calculate the residuals of one set of pre-processed point cloud data of the tunnel inner contour obtained in step 1, and determine whether the multiple residuals obtained conform to the normal distribution. If so, the fitting is completed and a single tunnel cross-sectional ellipse is obtained; if not, proceed to step 2.2;

[0041] Step 2.2, calculate the discrete value of each residual by the MAD method, and determine whether there is a residual greater than the discrete value. If so, remove the pre-processed point cloud data of the tunnel contour corresponding to the residual and return to step 2.1. Otherwise, go to step 3.

[0042] Step 4 specifically includes the following steps:

[0043] Step 4.1, determine whether f(x) satisfies f(x)<σ. If so, then in (x j -x j+1 ) tunnel deformation occurs, execute step 4.2; otherwise, no deformation occurs;

[0044]

[0045] in:

[0046] a1 represents the major semi-axis of the standard tunnel ellipse;

[0047] b1 represents the minor semi-axis of the standard tunnel ellipse;

[0048] x j Indicates the mileage in the tunnel, km;

[0049] Step 4.2: Perform weighted averaging on the tunnel cross-section ellipse with a fixed length L to obtain the parameters of the new ellipse after weighted averaging. The parameters of the new ellipse Including the major semi-axis, minor semi-axis and ellipse angle;

[0050]

[0051] in:

[0052] w i represents the weighted value of the tunnel section ellipse i;

[0053] d i are the parameters of the ellipse of the tunnel section ellipse i, including the major semi-axis, minor semi-axis and ellipse angle;

[0054] Step 4.3, the new ellipse parameters obtained according to step 4.2 The new ellipse is drawn and compared with the standard tunnel ellipse to obtain the deformation.

[0055] Compared with the prior art, the present invention has the following beneficial technical effects:

[0056] (I) The present invention realizes the overall deformation monitoring of highway tunnels based on the coordinates of tunnel inner wall point cloud slices through data preprocessing, normal distribution statistics, and improved least squares ellipse fitting. It can effectively remove noise and solve the problem that the monitoring method of three-dimensional laser point cloud is easily interfered by external factors (vibration of acquisition equipment and erroneous acquisition of optical equipment) and difficult to eliminate interfering point clouds (corresponding facilities in the tunnel, such as connecting bolt holes and nuts on shield ring segments, grouting holes, cables, lighting equipment and other attachments). It realizes accurate and efficient monitoring of contour deformation analysis in highway tunnels.

[0057] (II) The present invention is based on the point cloud data of the tunnel contour. It analyzes tunnel deformation by constructing a kd-tree and selecting tunnel section data of a fixed length L for weighted averaging. It can judge the tunnel deformation situation from the overall trend, reduce the error caused by single section data analysis, and improve the accuracy of monitoring data. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flow chart of the method of the present invention;

[0059] Figure 2 This is a comparison diagram before and after preprocessing in the present invention, (a) is the point cloud data before preprocessing, and (b) is the point cloud data after preprocessing.

[0060] The specific contents of the present invention are further explained in detail below with reference to the embodiments. DETAILED DESCRIPTION

[0061] It should be noted that, unless otherwise specified, all components in the present invention are components known in the art.

[0062] Specific embodiments of the present invention are given below. It should be noted that the present invention is not limited to the following specific embodiments, and all equivalent modifications made on the basis of the technical solution of this application fall within the protection scope of the present invention.

[0063] The present invention provides a method for analyzing the deformation of the inner contour of a highway tunnel, which comprises the following steps:

[0064] Step 1: collecting point cloud data of the tunnel inner contour and preprocessing it to obtain n groups of preprocessed point cloud data of the tunnel inner contour;

[0065] Preprocessing includes sequentially removing gross errors and removing interference points;

[0066] Gross error elimination is used to eliminate point cloud data that exceeds the tunnel contour range and point cloud data collected under the wrong vehicle posture;

[0067] Step 2: performing tunnel cross-sectional ellipse fitting on one set of pre-processed point cloud data of the tunnel inner contour obtained in step 1 to obtain a single tunnel cross-sectional ellipse;

[0068]

[0069] in:

[0070] a represents the major semi-axis parameter of the tunnel cross-section ellipse;

[0071] b represents the minor semi-axis parameter of the tunnel section ellipse;

[0072] Step 3: Repeat steps 1 to 2 to obtain n tunnel cross-sectional ellipses, and connect the centers of the n tunnel cross-sectional ellipses to obtain the tunnel central axis;

[0073] Step 4: Compare the tunnel cross-section ellipse obtained in step 3 with the standard tunnel ellipse to determine whether deformation occurs and the magnitude of the deformation.

[0074] In the above technical solution, the overall deformation monitoring of highway tunnels based on the coordinates of tunnel inner wall point cloud slices is realized through data preprocessing, normal distribution statistics, and improved least squares ellipse fitting. This can effectively remove noise, solve the problem that the three-dimensional laser point cloud monitoring method is easily interfered by external factors (vibration of acquisition equipment and erroneous acquisition of optical equipment) and is difficult to eliminate interfering point clouds (corresponding facilities in the tunnel, such as connecting bolt holes and nuts on shield ring segments, grouting holes, cables, lighting equipment and other attachments), and realize accurate and efficient monitoring of contour deformation analysis in highway tunnels.

[0075] The specific process of gross error elimination includes the following steps:

[0076] S1, perform mean filtering on the point cloud data of the tunnel contour, and use a sliding window to divide it into multiple regions. Calculate the average value of the point cloud data of the tunnel contour in each region and use it as the sampling point of the corresponding region;

[0077] S2, creating multiple 3D voxel grids of the same size, and calculating the distance between each sampling point in the 3D voxel grid and the center of the 3D voxel grid, and selecting the sampling point corresponding to the minimum value of the distance value as the final sampling point;

[0078] S3, cuts the final sampling points obtained in S2 in three dimensions to obtain an ordered kd-tree;

[0079] S4, obtain the K nearest neighbors of each node on the ordered kd-tree through binary search;

[0080] S5, calculate the distance from each node to the K nearest neighboring points on the ordered kd-tree, and calculate the average R of the distances between each node and its K nearest neighboring points;

[0081] S6, count the number Q of final sampling points within the range of 2R from each node on the ordered kd-tree, and determine whether the number Q satisfies the following formula. If so, the node is an isolated point and is removed; if not, the process ends;

[0082] Q<2K.

[0083] In the above technical solution, due to the interference of emergency shelters and other moving vehicles, the collected data becomes abnormal, which seriously affects the accuracy of the preliminary ellipse fitting and makes the adaptive threshold for eliminating subsequent interference points inappropriate, so gross errors are eliminated.

[0084] Among them, the purpose of S1 is to remove noise points close to the point group.

[0085] S3 specifically includes the following steps:

[0086] S3.1, given the initial coordinate axis x, and use it as the normal vector of the root node splitting hyperplane;

[0087] S3.2, sort the final sampling points obtained in S2 on the initial coordinate axis x, select the median of the sort as the split point, and divide the space into two parts using a plane perpendicular to the coordinate axis. The split point is used as the parent node, and the final sampling points on its left and right sides are used as the left and right subtrees of the partition;

[0088] S3.3, select the next level coordinate axis and use it as the normal vector of the node splitting hyperplane;

[0089] S3.4, sort the final sampling points on the left subtree and the right subtree on the next level coordinate axis, select the median of the sorted points as the split point, and use a plane perpendicular to the coordinate axis to divide the space into two parts. The split point is used as the parent node, and the final sampling points on its left and right sides are used as the left subtree and the right subtree of the next level of division;

[0090] S3.5, repeat S3.3 to S3.4 until it cannot be split any further and an ordered kd-tree is obtained.

[0091] In step 1, interference point removal includes the following steps:

[0092] Q1, divide the final sampling points into n groups along the tunnel direction, perform ellipse fitting on each of the n groups of final sampling points, and obtain n cross-sectional ellipses;

[0093] Q2, calculate the distance between the final sampling point of the i-th group and its corresponding cross-sectional ellipse, sort the distance from small to large, use the sequence number as the x-axis and the distance as the y-axis to obtain the corresponding fitting curve y i =g(x);

[0094] Q3, for the fitting curve y i = g(x) and take the derivative to get y i ', and then get y i 'The maximum corresponding distance value M i , and use it as the distance threshold of the corresponding cross-section ellipse;

[0095] Q4, determine whether the distance obtained in Q2 is less than the corresponding distance threshold. If so, retain the final sampling point; if not, remove the final sampling point, and obtain the i-th group of pre-processed point cloud data of the tunnel inner contour;

[0096] In Q5, let i = i + 1 and return to Q2 until i > n, and obtain n sets of pre-processed point cloud data of the tunnel inner contour.

[0097] In the above technical solution, removing interference points specifically means removing points that do not belong to the tunnel section, such as connecting bolt holes and nuts on the shield ring segments, grouting holes, cables, lighting equipment and other attachments.

[0098] According to the above technical solution, the tunnel wall point cloud data is processed and the following is obtained: Figure 2 As shown in the comparison chart, we can see that after preprocessing, the point cloud inside the tunnel is evenly distributed, and bolt holes, cables and other facilities and other mixed points can be effectively removed, with good denoising effect and high analysis accuracy.

[0099] Step 2 specifically includes the following steps:

[0100] Step 2.1, respectively calculate the residuals of one set of pre-processed point cloud data of the tunnel inner contour obtained in step 1, and determine whether the multiple residuals obtained conform to the normal distribution. If so, the fitting is completed and a single tunnel cross-sectional ellipse is obtained; if not, proceed to step 2.2;

[0101] Step 2.2, calculate the discrete value of each residual by the MAD method, and determine whether there is a residual greater than the discrete value. If so, remove the pre-processed point cloud data of the tunnel contour corresponding to the residual and return to step 2.1. Otherwise, go to step 3.

[0102] Step 4 specifically includes the following steps:

[0103] Step 4.1, determine whether f(x) satisfies f(x)<σ. If so, then in (x j -x j+1) tunnel deformation occurs, execute step 4.2; otherwise, no deformation occurs;

[0104]

[0105] in:

[0106] a1 represents the major semi-axis of the standard tunnel ellipse;

[0107] b1 represents the minor semi-axis of the standard tunnel ellipse;

[0108] x j Indicates the mileage in the tunnel, km;

[0109] Step 4.2: Perform weighted averaging on the tunnel cross-section ellipse with a fixed length L to obtain the parameters of the new ellipse after weighted averaging. Parameters of the new ellipse Including the major semi-axis, minor semi-axis and ellipse angle;

[0110]

[0111] in:

[0112] w i represents the weighted value of the tunnel section ellipse i;

[0113] d i are the parameters of the ellipse of the tunnel section ellipse i, including the major semi-axis, minor semi-axis and ellipse angle;

[0114] Step 4.3, the new ellipse parameters obtained according to step 4.2 The new ellipse is drawn and compared with the standard tunnel ellipse to obtain the deformation.

[0115] In the above technical solution, the long semi-axis a-a1 of the single tunnel section ellipse obtained by step 2 is used as The short semi-axis b-b1 of the tunnel section ellipse is As well as the mileage position x in the tunnel where sections a and b are located, we can put them into the above two curve equations to calculate the coefficients m0, m1, m2, m3, m4 and m5, and get and expression.

[0116] Based on the point cloud data of the tunnel contour, the tunnel deformation is analyzed by constructing a kd-tree and selecting tunnel section data of a fixed length L for weighted averaging. This can determine the tunnel deformation situation from the overall trend, reduce the error caused by single section data analysis, and improve the accuracy of monitoring data.

Claims

1. A method for analyzing profile deformation in a highway tunnel, characterized in that: The method comprises the following steps: Step 1: collecting point cloud data of the tunnel inner contour and preprocessing it to obtain n groups of preprocessed point cloud data of the tunnel inner contour; The preprocessing includes sequentially performing gross error elimination and interference point removal; The gross error elimination is used to eliminate point cloud data that exceeds the tunnel contour range and point cloud data collected under the wrong vehicle posture; Step 2: performing tunnel cross-sectional ellipse fitting on one set of pre-processed point cloud data of the tunnel inner contour obtained in step 1 to obtain a single tunnel cross-sectional ellipse; in: a represents the major semi-axis parameter of the tunnel cross-section ellipse; b represents the minor semi-axis parameter of the tunnel section ellipse; Step 3: Repeat steps 1 to 2 to obtain n tunnel cross-sectional ellipses, and connect the centers of the n tunnel cross-sectional ellipses to obtain the tunnel central axis; Step 4: Compare the tunnel cross-section ellipse obtained in step 3 with the standard tunnel ellipse to determine whether deformation occurs and the magnitude of the deformation.

2. The method for analyzing deformation of a highway tunnel profile according to claim 1, wherein: The specific process of removing gross errors includes the following steps: S1, perform mean filtering on the point cloud data of the tunnel contour, and use a sliding window to divide it into multiple regions. Calculate the average value of the point cloud data of the tunnel contour in each region and use it as the sampling point of the corresponding region; S2, creating multiple 3D voxel grids of the same size, and calculating the distance between each sampling point in the 3D voxel grid and the center of the 3D voxel grid, and selecting the sampling point corresponding to the minimum value of the distance value as the final sampling point; S3, cuts the final sampling points obtained in S2 in three dimensions to obtain an ordered kd-tree; S4, obtain the K nearest neighbors of each node on the ordered kd-tree through binary search; S5, calculate the distance from each node to the K nearest neighboring points on the ordered kd-tree, and calculate the average R of the distances between each node and its K nearest neighboring points; S6, count the number Q of final sampling points within the range of 2R from each node on the ordered kd-tree, and determine whether the number Q satisfies the following formula. If so, the node is an isolated point and is removed; If not, then end; Q<2K.

3. The method for analyzing deformation of a highway tunnel profile according to claim 2, wherein: S3 specifically includes the following steps: S3.1, given the initial coordinate axis x, and use it as the normal vector of the root node splitting hyperplane; S3.2, sort the final sampling points obtained in S2 on the initial coordinate axis x, select the median of the sort as the split point, and divide the space into two parts using a plane perpendicular to the coordinate axis. The split point is used as the parent node, and the final sampling points on its left and right sides are used as the left and right subtrees of the partition; S3.3, select the next level coordinate axis and use it as the normal vector of the node splitting hyperplane; S3.4, sort the final sampling points on the left subtree and the right subtree on the next level coordinate axis, select the median of the sorted points as the split point, and use a plane perpendicular to the coordinate axis to divide the space into two parts. The split point is used as the parent node, and the final sampling points on its left and right sides are used as the left subtree and the right subtree of the next level of division; S3.5, repeat S3.3 to S3.4 until it cannot be split any further and an ordered kd-tree is obtained.

4. The method for analyzing deformation of a highway tunnel profile according to claim 2, wherein: In step 1, the interference point removal includes the following steps: Q1, divide the final sampling points into n groups along the tunnel direction, perform ellipse fitting on each of the n groups of final sampling points, and obtain n cross-sectional ellipses; Q2, calculate the distance between the final sampling point of the i-th group and its corresponding cross-sectional ellipse, sort the distance from small to large, use the sequence number as the x-axis and the distance as the y-axis to obtain the corresponding fitting curve y i =g(x); Q3, for the fitting curve y i = g(x) and take the derivative to get y i ', and then get y i 'The maximum corresponding distance value M i , and use it as the distance threshold of the corresponding cross-section ellipse; Q4, determine whether the distance obtained in Q2 is less than the corresponding distance threshold. If so, retain the final sampling point; if not, remove the final sampling point, and obtain the i-th group of pre-processed point cloud data of the tunnel inner contour; In Q5, let i = i + 1 and return to Q2 until i > n, and obtain n sets of pre-processed point cloud data of the tunnel inner contour.

5. The method for analyzing deformation of the inner contour of a highway tunnel according to claim 1, wherein: Step 2 specifically includes the following steps: Step 2.1, respectively calculate the residuals of one set of pre-processed point cloud data of the tunnel inner contour obtained in step 1, and determine whether the multiple residuals obtained conform to the normal distribution. If so, the fitting is completed and a single tunnel cross-sectional ellipse is obtained; if not, proceed to step 2.2; Step 2.2, calculate the discrete value of each residual by the MAD method, and determine whether there is a residual greater than the discrete value. If so, remove the pre-processed point cloud data of the tunnel contour corresponding to the residual and return to step 2.

1. Otherwise, go to step 3.

6. The method for analyzing deformation of the inner contour of a highway tunnel according to claim 1, wherein: Step 4 specifically includes the following steps: Step 4.1, determine whether f(x) satisfies f(x)<σ. If so, then in (x j -x j+1 ) tunnel deformation occurs, execute step 4.2; otherwise, no deformation occurs; in: a1 represents the semi-major axis of the standard tunnel ellipse; b1 represents the minor semi-axis of the standard tunnel ellipse; x j Indicates the mileage in the tunnel, km; Step 4.2: Perform weighted averaging on the tunnel cross-section ellipse with a fixed length L to obtain the parameters of the new ellipse after weighted averaging. The parameters of the new ellipse Including the major semi-axis, minor semi-axis and ellipse angle; in: w i represents the weighted value of the tunnel section ellipse i; d i are the parameters of the ellipse of the tunnel section ellipse i, including the major semi-axis, minor semi-axis and ellipse angle; Step 4.3, the new ellipse parameters obtained according to step 4.2 The new ellipse is drawn and compared with the standard tunnel ellipse to obtain the deformation.

Images