Tunnel detection robot deformation identification method based on multi-dimensional data conversion

Abstract

The invention discloses a tunnel detection robot deformation identification method based on multi-dimensional data conversion, and the method comprises the steps: obtaining tunnel three-dimensional point cloud data through employing a coordinate correction algorithm and a three-dimensional point cloud data recovery algorithm, and obtaining the tunnel three-dimensional point cloud data through employing a point cloud data splicing technology and a statistical filtering method; the method comprises the following steps: acquiring a two-dimensional central axis of a tunnel by adopting a double-projection method, further converting the two-dimensional central axis into a three-dimensional central axis, extracting and refining the cross section of the tunnel by utilizing a direction vector of the three-dimensional central axis and an improved RANSAC algorithm, calculating basic parameters of the cross section based on a tunnel cross section ellipse fitting method, and constructing a deformation index to realize recognition of tunnel structure deformation. The method is suitable for identification, detection and early warning of tunnel structure deformation diseases. The problems that when the tunnel detection robot is used for recognizing deformation, the recognition precision is low when two-dimensional point cloud data is directly used, the data size is large when three-dimensional point cloud data is used, data processing is difficult, and related equipment is high in cost and poor in economical efficiency are solved.

Description

Technical Field

[0001] The present invention belongs to the field of tunnel structure disease detection and relates to a tunnel structure deformation identification method, and specifically to a tunnel detection robot deformation identification method based on multi-dimensional data conversion. Background Art

[0002] Tunnel structures, as crucial infrastructure that traverse obstacles and connect different regions, play an irreplaceable role in transportation, public transportation, and economic development. By the end of 2023, China had 27,297 highway tunnels totaling over 30.2318 million meters, and 18,573 railway tunnels totaling over 23.508 million meters. Furthermore, with the continuous development of my country's economy and the advancement of urbanization, the construction, operation, and maintenance of tunnel structures will intensify. During their design and operation, tunnel structures are subject to the combined effects of environmental, traffic, and sudden loads. Long-term, repetitive loads can lead to reduced resistance and structural damage, resulting in deformation, settlement, and other defects. These can impact transportation efficiency at best, and even threaten life and property at worst. Therefore, regular inspections of tunnel structures to promptly detect defects and damage are crucial to ensuring their safety.

[0003] Deformation is a relatively common defect in tunnel structures, and its identification, diagnosis, and early warning are gradually becoming an important part of tunnel structure inspection. To compensate for the shortcomings of manual deformation detection, such as low efficiency, large errors, and safety hazards, tunnel inspection robots equipped with scanning equipment such as lidar and laser rangefinders have emerged. They use point cloud data to identify deformation. However, the point cloud data obtained by two-dimensional laser scanning equipment cannot fully describe the structural state, and the identification error is large; the point cloud data obtained by three-dimensional laser scanning equipment is large in volume and difficult to process. In addition, three-dimensional laser scanning equipment is expensive and less economical. Therefore, by equipping the tunnel inspection robot with a two-dimensional laser scanning device and using some technical means to convert two- and three-dimensional point cloud data, it is possible to complete a relatively accurate identification of tunnel structure deformation. Summary of the Invention

[0004] In order to solve the problems of low recognition accuracy when using two-dimensional point cloud data to identify deformation using a tunnel inspection robot, large data volume and difficulty in data processing when using three-dimensional point cloud data, as well as high cost and poor economy of related equipment, the present invention provides a tunnel inspection robot deformation recognition method based on multi-dimensional data conversion, which can achieve relatively accurate and effective recognition of tunnel structure deformation.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A tunnel inspection robot deformation recognition method based on multi-dimensional data conversion includes the following steps:

[0007] Step 1: For a double-deck tunnel used for both road and rail, the 2D point cloud data of each layer obtained by the inspection robot is introduced. A coordinate correction algorithm based on 2D inclination monitoring data and a 3D point cloud data recovery algorithm based on 1D mileage data are used to convert the original 2D tunnel point cloud data into high-precision 3D point cloud data.

[0008] Step 2: Use control point-based point cloud data stitching technology to stitch the point cloud data of each layer of the tunnel to obtain complete 3D point cloud data of the tunnel. Statistical filtering methods are then used to remove outlier noise points in the 3D point cloud data.

[0009] Step 3: Use the double projection method to obtain the 2D central axis of the tunnel, and use the two 2D central axes to complete the conversion of the 3D central axis of the tunnel;

[0010] Step 4: For the railway layer, separate the tunnel and the trackbed;

[0011] Step 5: Use the 3D central axis direction vector to extract the tunnel cross section, and use the improved RANSAC algorithm to perform fine filtering on the tunnel cross section;

[0012] Step 6: Based on the tunnel cross-section ellipse fitting method, the basic cross-section parameters are calculated and the deformation index is constructed to identify the deformation of the tunnel structure.

[0013] Compared with the prior art, the present invention has the following advantages:

[0014] The present invention utilizes a coordinate correction algorithm and a 3D point cloud data recovery algorithm to obtain high-precision 3D point cloud data of the tunnel, adopts point cloud data splicing technology and a statistical filtering method to obtain 3D point cloud data of the tunnel that is more in line with the actual situation, adopts a double projection method to obtain the 2D central axis of the tunnel, and further converts it into a 3D central axis. The 3D central axis direction vector and an improved RANSAC algorithm are used to extract and refine the tunnel cross section, and based on the tunnel cross section ellipse fitting method, the basic parameters of the cross section are calculated, and deformation indicators are constructed to realize the identification of tunnel structure deformation. The present invention is suitable for the identification, detection and early warning of tunnel structure deformation diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of the deformation recognition method of the tunnel inspection robot based on multi-dimensional data conversion.

[0016] Figure 2 Schematic diagram of the relevant coordinate system during coordinate correction.

[0017] Figure 3 Schematic diagram of the coordinate system used for tunnel centerline, cross-section extraction, fitting, and separation of the tunnel and trackbed.

[0018] Figure 4 Tunnel cross section and tunnel 3D point cloud data before roadbed separation, filtering and denoising.

[0019] Figure 5 The tunnel cross section, tunnel 3D point cloud data and its central axis after separating the roadbed and filtering and denoising.

[0020] Figure 6 This is the tunnel structure deformation identification result obtained using the method of the present invention. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is further described below with reference to the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention that does not depart from the spirit and scope of the technical solution of the present invention should be included in the scope of protection of the present invention.

[0022] The present invention provides a deformation identification method for a tunnel inspection robot based on multi-dimensional data conversion. The method obtains two-dimensional point cloud data of the tunnel structure through a two-dimensional laser scanning device mounted on the tunnel inspection robot, and uses a series of technical means to convert it into three-dimensional point cloud data, and extracts information such as the tunnel centerline, cross section, and its geometric parameters therefrom, and then realizes the identification of tunnel structure deformation through changes in related information. Existing tunnel structure deformation identification methods usually follow this idea: directly use the two-dimensional or three-dimensional point cloud data obtained by the tunnel inspection robot for deformation identification. However, two-dimensional point cloud data cannot fully describe the structural state, the identification error is large, the three-dimensional point cloud data has a large amount of data and is difficult to process, and the related equipment cost is high and the economic efficiency is poor. Therefore, in the present invention, a three-dimensional point cloud data recovery algorithm is adopted to realize the conversion of two- and three-dimensional point cloud data, and the double projection method and ellipse fitting method are used to extract information and parameters such as the tunnel centerline, cross section, etc. By constructing deformation indicators, the deformation of the tunnel structure is identified to ensure the safety of the tunnel structure. Figure 1 As shown, the specific steps include:

[0023] Step 1: For a double-deck tunnel used for both road and rail, the 2D point cloud data of each layer obtained by the inspection robot is introduced. A coordinate correction algorithm based on 2D inclination monitoring data and a 3D point cloud data recovery algorithm based on 1D mileage data are used to convert the original 2D tunnel point cloud data into high-precision 3D point cloud data. The specific steps are as follows:

[0024] Step 1: Use the coordinate correction algorithm based on the 2D tilt monitoring data to correct the coordinates of the 2D point cloud data collected by the detection robot:

[0025]

[0026] Where α is the coordinate conversion angle; (x′, z′) is the measured coordinate value; (x, z) is the corrected coordinate value; γ is the tilt sensor monitoring data; and D is the distance from the center of the detection robot to the center of the laser scanner.

[0027] Step 1 and 2: Use a 3D point cloud data recovery algorithm based on 1D mileage data to convert the 2D point cloud data into 3D point cloud data:

[0028] y i =M i i=1,2,···,N (2)

[0029] Where y i is the Y-axis coordinate in the 3D point cloud data; M i is the mileage data; i=1,2,···, N is the number of mileage data.

[0030] Therefore, the coordinates of the tunnel 3D point cloud data can be expressed as (x, y, z).

[0031] Step 2: Use control point-based point cloud data stitching technology to stitch the point cloud data of each layer of the tunnel to obtain complete 3D point cloud data of the tunnel. Then use statistical filtering to remove outlier noise points in the 3D point cloud data. The specific steps are as follows:

[0032] Step 21: Use the control point-based point cloud data stitching technology to stitch the point cloud data of each layer of the tunnel obtained in step 1:

[0033]

[0034] Where, the translation parameter [ΔXΔYΔZ] T , rotation parameter [ε X ε Y ε Z ] T and scale parameter m are obtained through target coordinates; [X′Y′Z′] T is the point cloud data before stitching; [XYZ] T The point cloud data after stitching.

[0035] Calculate the scale parameter m:

[0036]

[0037] Where, P upper,i and P lower,i are the i-th point cloud data coordinates of the upper and lower layers respectively; and are the centroids of the upper and lower targets respectively, and n is the total number of point cloud data.

[0038] Calculate the rotation parameter [ε X ε Y ε Z ] T :

[0039]

[0040] Where Q upper,i and Q lower,i are the coordinates of the upper and lower layer point cloud data after removing the centroid, H is the covariance matrix, and n is the total number of point cloud data.

[0041] Perform SVD decomposition on the covariance matrix H = UOV T , get the rotation matrix R = VU T , the components of the rotation matrix are the rotation parameters ε X , ε Y , ε Z .

[0042] Calculate translation parameters [ΔXΔYΔZ] T :

[0043]

[0044] Where T is the translation matrix, and the components of the translation matrix are the translation parameters ΔX, ΔY, and ΔZ.

[0045] Step 22: Use statistical filtering method to remove outlier noise points in the spliced point cloud data:

[0046] Let S i For the average distance between the i-th point cloud data and its neighborhood point cloud, calculate its mean μ and standard deviation σ:

[0047]

[0048] Where n is the total number of point cloud data.

[0049] The standard deviation variation factor a is introduced to determine the valid point cloud data range (μ-aσ, μ+aσ). When the average distance between the i-th point cloud data and its neighboring point clouds is outside this range, the point cloud is regarded as a noise point and removed.

[0050] Step 3: Use the double projection method to obtain the 2D central axis of the tunnel, and use the two 2D central axes to complete the conversion of the 3D central axis of the tunnel. The specific steps are as follows:

[0051] Step 31: Project the tunnel 3D point cloud data onto the XOZ and YOZ planes respectively, and calculate the coordinates of the projected point cloud data:

[0052]

[0053] Where x a ,y a ,z a is the coordinate of the projected point cloud data; x b ,y b ,z b is the coordinate of the point cloud data before projection; A, B, C, D are the projection plane parameters.

[0054] Step 32: Divide the projected point cloud data into several sets and calculate the average value of the point cloud data coordinates in each set The two-dimensional centerline points of the tunnel are discretized and the centerline equation is fitted using the least squares method.

[0055] Taking the YOZ plane as an example, let the central axis equation be y=a1z+b1, and solve the central axis parameters:

[0056]

[0057] Similarly, we can get the two-dimensional central axis equation x=a2z+b2 on the XOZ plane and solve the central axis parameters:

[0058]

[0059] In summary, the three-dimensional central axis equation of the tunnel can be converted to:

[0060]

[0061] Step 4: For the railway layer, separate the tunnel and the trackbed. The specific steps are as follows:

[0062] Let the coordinates of the i-th point cloud data on the XOY plane be (x i ,y i ), the coordinates of the point on the tunnel's three-dimensional central axis on the XOY plane are (x m ,y m ), the origin of the XOY plane is (x o ,y o ), calculate the angle θ between each point cloud on the XOY plane and the line connecting the coordinate origin and the positive direction of the X axis:

[0063]

[0064] The angle range Θ of the roadbed is determined according to the actual situation. When θ∈Θ, the point cloud is determined to be a roadbed point and removed.

[0065] Step 5: Use the 3D central axis direction vector to extract the tunnel cross section, and use the improved RANSAC algorithm to perform fine filtering on the tunnel cross section. The specific steps are as follows:

[0066] Step 51: Convert the three-dimensional central axis of the tunnel obtained in step 3 into:

[0067]

[0068] Where (x0, y0, z0) is the point on the tunnel axis; (l, m, n) is the direction vector of the tunnel axis.

[0069] Transform the coordinate system so that the tunnel centerline coincides with the plumb bob axis and calculate the coordinate system transformation matrix:

[0070]

[0071] After coordinate conversion, the point cloud data is sliced according to the width of the tunnel segments and pipe segments and the actual detection accuracy requirements to extract the tunnel cross section.

[0072] Step 52: Use the improved RANSAC algorithm to perform cross-section refinement filtering:

[0073] Calculate the extreme values x of the two coordinate axes in each cross-section point cloud data max ,x min ,y max and y min ,According to the extreme difference of the two coordinates, the cross-section point cloud is evenly divided into k segments.,After RANSAC filtering is performed on each segment of the point cloud, the segment point clouds are fused as the final result of the tunnel cross-section refinement filtering.

[0074] Step 6: Based on the tunnel cross-section ellipse fitting method, calculate the basic cross-section parameters and construct deformation indicators to identify the tunnel structure deformation. The specific steps are as follows:

[0075] Step 61: Calculate the sum of the algebraic distances between each point on the cross section and the fitted ellipse:

[0076]

[0077] In the formula, let F(P,Q) be the elliptic equation F(P,Q)=PQ=Ax 2 +Bxy+Cy 2 +Dx+Ey+F=0, satisfying B 2 -4AC<0, for the convenience of calculation, take B 2 -4AC = -1; Q=[ABCDEF] T ;(x hi ,y hi ) are the coordinates of the point on the cross section.

[0078] Step 62: Solve the ellipse equation with the minimum sum of algebraic distances mentioned in step 61, which is the fitting ellipse of the tunnel cross section:

[0079]

[0080] Where,

[0081] Introducing the Lagrangian operator λ, auxiliary variable s = PQ and penalty coefficient μ, transform the above optimization problem into:

[0082]

[0083] The Bregman method is used, and the Bregman iteration with the Bregman variable t is introduced to convert the above optimization problem into two sub-problems for solution:

[0084]

[0085] Where r is the number of iterations.

[0086] Set the initial parameters according to the actual situation. When the number of iterations reaches the set value or D k+1 >D k When , the iteration stops and the geometric parameters of the fitted ellipse are output. The center coordinates (X0, Y0) of the ellipse and the lengths of the major and minor axes l are calculated according to the geometric parameters. x ,l y and the deflection angle θ:

[0087]

[0088] Step 63: Construct deformation index based on ellipticity:

[0089]

[0090] Where R is the design radius.

[0091] Based on the change in ellipticity, the degree of tunnel structure deformation can be identified; if relevant thresholds are set, diagnosis and early warning of tunnel structure deformation can also be achieved.

[0092] The following test is used to verify the effect of the present invention:

[0093] This experiment verifies the effectiveness of the method based on the tunnel two-dimensional laser point cloud data, inclination monitoring data and mileage data obtained during the inspection of the detection robot.

[0094] The details of this test are as follows:

[0095] According to the two-dimensional laser point cloud data of the selected tunnel structure, a coordinate correction algorithm based on two-dimensional inclination monitoring data is used to correct the point cloud data coordinates; for the corrected coordinates, a three-dimensional point cloud data recovery algorithm based on one-dimensional mileage data is used to obtain high-precision three-dimensional tunnel point cloud data.

[0096] For the separated point cloud data of the road layer and the railway layer, the point cloud data splicing technology based on control points is used to splice them into complete point cloud data. The statistical filtering method is used on the complete point cloud data to remove outlier noise points.

[0097] Based on the complete three-dimensional point cloud data, the double projection method is used to obtain the two-dimensional central axis of the tunnel; the three-dimensional central axis of the tunnel is obtained using the two two-dimensional central axes. Figure 5 The red straight line is the three-dimensional central axis of the tunnel obtained using this method.

[0098] Separate the tunnel and the trackbed from the point cloud data of the railway layer.

[0099] Based on the tunnel's 3D central axis, the tunnel cross section is extracted using the central axis direction vector. The extracted cross section is filtered using an improved RANSAC algorithm to obtain a more accurate tunnel cross section. Figure 4 To separate the roadbed and filter and remove noise from the tunnel cross section and tunnel 3D point cloud data, Figure 5 The tunnel cross section and 3D point cloud data are obtained after using the aforementioned method to separate the roadbed, filter, and denoise the roadbed. Comparing the two images reveals that the roadbed and outlier noise points significantly impact the extraction of tunnel point cloud data, while the proposed method allows for more accurate extraction of tunnel point cloud data.

[0100] For the filtered cross-section, the ellipse fitting method is used to calculate the basic parameters of the cross-section. The deformation index is constructed based on the basic parameters of the cross-section, and the deformation degree of the tunnel structure is identified based on the deformation index. Figure 6 The tunnel structure deformation identification results obtained using the present invention are shown in Figure 2. According to relevant standards, the allowable deviation of tunnel structure ellipticity is 8‰. Therefore, the tunnel structure under inspection is in a safe state, which also demonstrates the effectiveness of the method proposed in the present invention.

[0101] The present invention converts the multi-dimensional detection data of the tunnel structure, extracts information such as the tunnel's central axis, cross-sectional shape, and basic cross-sectional parameters, and obtains the degree of deformation of the tunnel structure, which can achieve the purpose of identifying deformation diseases of the tunnel structure, thereby providing effective protection for the safe operation and maintenance of the tunnel structure.

Claims

1. A tunnel inspection robot deformation recognition method based on multi-dimensional data conversion, characterized in that The method comprises the following steps: Step 1: For a double-deck tunnel used for both road and rail, the 2D point cloud data of each layer obtained by the inspection robot is introduced. A coordinate correction algorithm based on 2D inclination monitoring data and a 3D point cloud data recovery algorithm based on 1D mileage data are used to convert the original 2D tunnel point cloud data into high-precision 3D point cloud data. Step 2: Use control point-based point cloud data stitching technology to stitch the point cloud data of each layer of the tunnel to obtain complete 3D point cloud data of the tunnel. Statistical filtering methods are then used to remove outlier noise points in the 3D point cloud data. Step 3: Use the double projection method to obtain the 2D central axis of the tunnel, and use the two 2D central axes to complete the conversion of the 3D central axis of the tunnel; Step 4: For the railway layer, separate the tunnel and the trackbed; Step 5: Use the 3D central axis direction vector to extract the tunnel cross section, and use the improved RANSAC algorithm to perform fine filtering on the tunnel cross section; Step 6: Based on the tunnel cross-section ellipse fitting method, the basic cross-section parameters are calculated and the deformation index is constructed to identify the deformation of the tunnel structure.

2. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 1 is characterized in that The specific steps of step one are as follows: Step 1: Use the coordinate correction algorithm based on the 2D tilt monitoring data to correct the coordinates of the 2D point cloud data collected by the detection robot: Where α is the coordinate conversion angle; (x′, z′) is the measured coordinate value; (x, z) is the corrected coordinate value; γ is the tilt sensor monitoring data; D is the distance from the center of the detection robot to the center of the laser scanner; Step 1 and 2: Use a 3D point cloud data recovery algorithm based on 1D mileage data to convert the 2D point cloud data into 3D point cloud data: y i =M i i=1,2,···,N Where y i is the Y-axis coordinate in the 3D point cloud data; M i is the mileage data; i=1,2,···, N is the number of mileage data.

3. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 1 is characterized in that The specific steps of step 2 are as follows: Step 21: Use the control point-based point cloud data stitching technology to stitch the point cloud data of each layer of the tunnel obtained in step 1: Where, the translation parameter [ΔXΔYΔZ] T , rotation parameter [ε X ε Y ε Z ] T and scale parameter m are obtained through target coordinates; [X′Y′Z′] T is the point cloud data before stitching; [XYZ] T is the point cloud data after stitching; Step 22: Use statistical filtering method to remove outlier noise points in the spliced point cloud data.

4. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 3 is characterized in that The formula for calculating the scale parameter m is: Where, P upper,i and P lower,i are the i-th point cloud data coordinates of the upper and lower layers respectively; and are the centroids of the upper and lower targets respectively, and n is the total number of point cloud data; Calculate the rotation parameter [ε X ε Y ε Z ] T The formula is: Where Q upper,i and Q lower,i are the coordinates of the upper and lower layer point cloud data after removing the centroid, H is the covariance matrix, and n is the total number of point cloud data; Perform SVD decomposition on the covariance matrix H = UOV T , get the rotation matrix R = VU T , the components of the rotation matrix are the rotation parameters ε X , ε Y , ε Z ; Calculate translation parameters [ΔXΔYΔZ] T The formula is: Where T is the translation matrix, and the components of the translation matrix are the translation parameters ΔX, ΔY, and ΔZ.

5. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 3 is characterized in that The specific steps of step 22 are as follows: Let S i For the average distance between the i-th point cloud data and its neighborhood point cloud, calculate its mean μ and standard deviation σ: Where n is the total number of point cloud data; The standard deviation variation factor a is introduced to determine the valid point cloud data range (μ-aσ, μ+aσ). When the average distance between the i-th point cloud data and its neighboring point clouds is outside this range, the point cloud is regarded as a noise point and removed.

6. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 1 is characterized in that The specific steps of step three are as follows: Step 31: Project the tunnel 3D point cloud data onto the XOZ and YOZ planes respectively, and calculate the coordinates of the projected point cloud data: Where x a ,y a ,z a is the coordinate of the projected point cloud data; x b ,y b ,z b is the coordinate of the point cloud data before projection; A, B, C, D are the projection plane parameters; Step 32: Divide the projected point cloud data into several sets and calculate the average value of the point cloud data coordinates in each set The two-dimensional centerline points of the tunnel are discretized and the centerline equation is fitted using the least squares method.

7. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 6 is characterized in that The specific steps of step 32 are as follows: Assume that the equation of the central axis on the YOZ plane is y=a1z+b1, and solve the central axis parameters according to the following formula: Assume that the two-dimensional central axis equation on the XOZ plane is x=a2z+b2, and solve the central axis parameters according to the following formula: The three-dimensional central axis equation of the tunnel is converted to:

8. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 1 is characterized in that The specific steps of step 4 are as follows: Let the coordinates of the i-th point cloud data on the XOY plane be (x i ,y i ), the coordinates of the point on the tunnel's three-dimensional central axis on the XOY plane are (x m ,y m ), the origin of the XOY plane is (x o ,y o ), calculate the angle θ between each point cloud on the XOY plane and the line connecting the coordinate origin and the positive direction of the X axis: The angle range Θ of the roadbed is determined according to the actual situation. When θ∈Θ, the point cloud is determined to be a roadbed point and removed.

9. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 7, characterized in that The specific steps of step five are as follows: Step 51: Convert the three-dimensional central axis of the tunnel obtained in step 3 into: Where (x0, y0, z0) is the point on the tunnel axis; (l, m, n) is the direction vector of the tunnel axis; Transform the coordinate system so that the tunnel centerline coincides with the plumb bob axis and calculate the coordinate system transformation matrix: After coordinate conversion, the point cloud data is sliced according to the width of the tunnel segments and pipe segments and the actual detection accuracy requirements to extract the tunnel cross section; Step 52: Use the improved RANSAC algorithm to perform cross-section refinement filtering: Calculate the extreme values x of the two coordinate axes in each cross-section point cloud data max ,x min ,y max and y min ,According to the extreme difference of the two coordinates, the cross-section point cloud is evenly divided into k segments.,After RANSAC filtering is performed on each segment of the point cloud, the segment point clouds are fused as the final result of the tunnel cross-section refinement filtering.

10. The method for tunnel inspection robot deformation recognition based on multi-dimensional data conversion according to claim 1, characterized in that The specific steps of step six are as follows: Step 61: Calculate the sum of the algebraic distances between each point on the cross section and the fitted ellipse: In the formula, let F(P,Q) be the elliptic equation F(P,Q)=PQ=Ax 2 +Bxy+Cy 2 +Dx+Ey+F=0, satisfying B 2 -4AC<0, for the convenience of calculation, take B 2 -4AC = -1; Q=[ABCDEF] T ;(x hi ,y hi ) are the coordinates of the points on the cross section; Step 62: Solve the ellipse equation with the minimum sum of algebraic distances mentioned in step 61, which is the fitting ellipse of the tunnel cross section: Where, Introducing the Lagrangian operator λ, auxiliary variable s = PQ and penalty coefficient μ, transform the above optimization problem into: The Bregman method is used, and the Bregman iteration with the Bregman variable t is introduced to convert the above optimization problem into two sub-problems for solution: Where r is the number of iterations; Set the initial parameters according to the actual situation. When the number of iterations reaches the set value or D k+1 >D k When , the iteration stops and the geometric parameters of the fitted ellipse are output. The center coordinates (X0, Y0) of the ellipse and the lengths of the major and minor axes l are calculated according to the geometric parameters. x ,l y and the deflection angle θ: Step 63: Construct deformation index based on ellipticity: Where R is the design radius; According to the change of ellipticity, the deformation degree of the tunnel structure can be identified.

Images