Precise measurement method for shield tunnel segment misalignment based on binocular vision

Abstract

The application relates to a kind of shield tunnel segment misalignment precision measurement methods based on binocular vision.The method comprises the following steps: firstly, a large number of similar photos are collected by a binocular camera at the segment misalignment to establish an image dataset;secondly, the Yolo algorithm and the convolutional neural network algorithm are used to automatically select target measurement points at the misalignment;thirdly, the binocular vision ranging technology is used to obtain the world three-dimensional coordinates of the target measurement points;then, the world three-dimensional coordinates of the target measurement points are related to the deflection angle θ due to the fact that the XY plane of the camera is not completely parallel to the lining ring tangent plane, and the deflection angle θ is obtained;finally, the influence of the deflection angle θ on the depth measurement is eliminated through coordinate system conversion, so as to accurately calculate the misalignment.

Description

Technical Field

[0001] This invention relates to the field of shield tunnel construction and operation and maintenance, and in particular to a method for accurate measurement of shield tunnel segment misalignment based on binocular vision. Background Technology

[0002] Segment misalignment is a common phenomenon during the construction and long-term operation of shield tunnels. Severe misalignment can cause structural defects such as segment cracking and tunnel leakage, which in turn adversely affect structural durability and train operation safety. Therefore, accurate and efficient measurement of lining segment misalignment is an important prerequisite for ensuring high-quality construction and efficient operation and maintenance of shield tunnels. Currently, devices for measuring shield tunnel segment misalignment are broadly classified into two types based on their working principles: mechanical ranging and laser (structured light) ranging.

[0003] Mechanical distance measuring methods typically place the measuring device at the misalignment point of the tunnel lining segments and then use mechanical devices such as a measuring ruler to measure the height difference (i.e., misalignment) between the lining segments. For example, in patents 202210890656.X and 202221219846.0, the measuring device is fixed at the misalignment point, and the extension length of the measuring ruler inside the device is the misalignment amount of the tunnel lining segments.

[0004] Laser ranging can be further divided into two main categories based on its principle: direct measurement and indirect measurement. The former directly calculates distance by recording the time difference (e.g., patents 201720582610.6 and 202121640476.3) or phase difference (e.g., patents 201911208332.8, 202111626349.2, and 201811089324.1) of the laser beam's round trip, obtaining point cloud data of the lining segments, and then calculating the misalignment between the segments. The latter projects a line laser near the segment joints (e.g., patent 201910524743.1), and then indirectly calculates the three-dimensional coordinates of each pixel in the light stripe using computer vision methods, thus determining the misalignment.

[0005] Binocular vision, based on the principle of precise positioning or calculation of three-dimensional coordinates of target objects, has been widely applied in various industries (especially in the field of autonomous driving). In the field of civil engineering, for example, patents 202021321273.3 and 202110597490.8 utilize binocular vision technology to monitor slopes in real time and measure three-dimensional displacement values, analyzing the displacement data to issue timely warning signals. Patent 202020343602.8 installs strip-shaped target objects with measuring points at the edge of bridge expansion joints, achieving real-time monitoring of the expansion joint width through binocular vision. Patent 202210348866.6 uses a binocular camera to acquire the spatial coordinates of monitoring points on the tunnel invert arch, achieving real-time monitoring of tunnel invert arch deformation. Although binocular vision technology has been initially applied in several fields of civil engineering, its application in measuring the misalignment of shield tunnel segments has not yet been reported.

[0006] For segment misalignment measurement, mechanical ranging is simple in principle and reliable in structure, but manual measurement and reading are inconvenient, have large errors, and are time-consuming and labor-intensive. Laser ranging requires high observation accuracy (measured time difference at the nanosecond level and phase difference at the nanoarc level), thus its technical threshold and cost are relatively high. Binocular ranging has the advantages of wide applicability and relatively inexpensive hardware. Compared with traditional ranging instruments, it only requires two cameras and related image processing algorithms. However, to measure the misalignment with millimeter-level precision, the requirements for image quality and ranging algorithm are very high. In terms of image quality, the positional relationship between the two cameras or the tilt angle between the camera's field of view and the target object can easily lead to perspective effects where near objects appear larger than distant objects, resulting in larger measurement errors (i.e., the measured "apparent height difference" is taken as the "true height difference"). In terms of ranging algorithm, due to the unavoidable camera placement errors and the perspective effects that may occur during photography, in order to reduce the impact of these errors and ensure that the measurement results meet expectations, the algorithm must be developed and optimized to effectively improve the ranging accuracy and robustness. Summary of the Invention

[0007] The purpose of this invention is to overcome the problems and shortcomings of existing technologies and provide a method for accurate measurement of misalignment of shield tunnel segments based on binocular vision. This method involves selecting several measuring points on the segments (such as...) Figures 1-2 As shown), by utilizing the principle of binocular vision ranging, the deflection angle between the camera's XY plane and the lining circumferential plane is corrected, converting the "apparent height difference" measured under perspective into the "true height difference," thereby achieving accurate measurement of the misalignment (as shown). Figures 3-5 (As shown).

[0008] To achieve the above objectives, the technical solution of the present invention is: a method for accurate measurement of misalignment of shield tunnel segments based on binocular vision, comprising the following steps:

[0009] S1. Collect a large number of similar photos at the misalignment of the tunnel segments using a binocular camera to establish a dataset of tunnel segment misalignment images.

[0010] S2. Automatically select target measurement points at the misalignment of the tunnel segments using the YOLO algorithm and convolutional neural network algorithm;

[0011] S3. Obtain the world three-dimensional coordinates of the target measurement point using binocular vision ranging technology;

[0012] S4. Establish a relationship between the world three-dimensional coordinates of the target measurement point and the deflection angle θ due to the fact that the camera XY plane is not completely parallel to the lining circumferential tangent plane, and obtain the deflection angle θ.

[0013] S5. By transforming the coordinate system, the influence of the deflection angle θ on the apparent depth measurement is eliminated, thereby accurately calculating the misalignment.

[0014] Of the five steps described above, steps S2 and S3 employ the YOLO algorithm for automatic target recognition, convolutional neural network algorithm, and binocular vision ranging, all of which have been widely used in the field of computer vision and possess high reliability and accuracy. Steps S4 and S5 are the core of this invention. Based on binocular vision ranging technology, they introduce the concept of deflection angle θ to correct the depth measurement of the target point, eliminating the influence of deflection angle θ on the accuracy of misalignment measurement. This also provides a reference for applications in other related fields.

[0015] In one embodiment of the present invention, step S1 is specifically implemented as follows:

[0016] Images were acquired at the misalignment of the tunnel segments using a binocular camera, and the images were distinguished by the labels "left_00.png" and "right_00.png" when saved. A dataset of images of the misalignment of the tunnel segments was established by acquiring a large number of similar photos. When acquiring images, the camera's XY plane was kept parallel to the lining circumferential cutting plane as much as possible, and the camera's X-axis direction was kept perpendicular to the joint to reduce the influence of perspective effect on the measurement results.

[0017] In one embodiment of the present invention, step S2 is specifically implemented as follows:

[0018] In the image dataset of misaligned pipe segments, the joints are treated as "cracks" and the rest as "background" for image annotation. The misaligned pipe segment areas are automatically identified by the YOLO algorithm, and then the convolutional neural network algorithm is used to detect the joints in the misaligned areas and accurately locate the two joint edge lines.

[0019] Using a convolutional neural network algorithm, target measurement point A is automatically selected on one seam edge line and target measurement point B is automatically selected on the other seam edge line, so that the distance between target measurement point A and target measurement point B is minimized; target measurement point M and target measurement point N are selected at equal pixel intervals on both sides of the line AB, resulting in 4 target measurement points for subsequent 3D coordinate calculation.

[0020] In one embodiment of the present invention, step S3 is specifically implemented as follows:

[0021] Using binocular vision ranging technology, in the world coordinate system O w (X w Y w Z w Camera coordinate system O c (X c Y c Z c ), focal plane coordinate system O f (U f V f ) and pixel coordinate system O p (U p V p By establishing a coordinate transformation relationship between the left and right images, a disparity map is obtained through stereo matching. The distance Zc from the target measurement point to the camera's XY plane, i.e., the apparent depth, is then calculated. Finally, the world three-dimensional coordinates of the target measurement point are obtained based on the coordinate transformation relationship.

[0022] In one embodiment of the present invention, step S4 is specifically implemented as follows:

[0023] The world three-dimensional coordinates of the target measurement points A(x1, y1, z1), B(x2, y2, z2), M(x3, y3, z3) and N(x4, y4, z4) are obtained through step S3. Since it is difficult to ensure that the camera XY plane is completely parallel to the lining ring cutting plane during actual image acquisition, it is assumed that there is a deflection angle θ between the two planes.

[0024] The coordinate axes of a stereo camera are defined as follows: the X-axis points from the left camera to the right camera; the Y-axis is perpendicular to the camera's optical axis, i.e., the line of sight; the Z-axis is the direction of the camera's optical axis, i.e., from the camera towards the scene, forming a right-handed coordinate system with the X-axis and Y-axis; the origin of the world coordinate system is defined at the left camera.

[0025] Assume that plane FG is the initial camera XY plane, plane CD is the camera XY plane after rotation θ, the lining ring cutting plane is parallel to plane CD, AF and BG are the line of sight depth of target measurement points A and B respectively, and AC⊥CD, BD⊥CD, draw AE⊥BD intersecting at point E, draw the extensions of AC and BD to plane FG intersecting at points H and J, draw HK⊥BJ intersecting at point K;

[0026] Similarly, assume that plane PS is the initial camera XY plane, plane QU is the camera XY plane after rotation θ, the lining ring cutting plane is parallel to plane QU, MP and NS are the depth lines of the target measurement points M and N respectively, and MQ⊥QU, NU⊥QU, construct AM′ / / BE such that MM′⊥AM′, construct the extension of EB to N′ such that NN′⊥BN′, construct the extension of NU to plane PS and intersect at point T;

[0027] Based on trigonometric relationships, the following system of equations is established for the world three-dimensional coordinates of the four target measuring points and the deflection angle θ; where AE, EB, MM′, NN′ and θ are unknowns, and the others are knowns, and ρ is the radius of the lining segment;

[0028]

[0029] Solving the above system of equations simultaneously, eliminating AE and EB, the formula for calculating the deflection angle θ is:

[0030]

[0031] In one embodiment of the present invention, step S5 is specifically implemented as follows:

[0032] According to the definition of the deflection angle θ, the original world three-dimensional coordinates (x, y, θ) of the target measurement point are... w y w , z w ), transformed into new world three-dimensional coordinates (x′) w y′ w , z′ w As shown in the following formula:

[0033]

[0034] After coordinate transformation, the new world three-dimensional coordinates of the four target measurement points are (x′1, y′1, z′1), (x′2, y′2, z′2), (x′3, y′3, z′3) and (x′4, y′4, z′4); then the apparent depth difference between the A and B target measurement points is z′1-z′2, which is the misalignment amount EB.

[0035] Compared with the prior art, the present invention has the following beneficial effects: The present invention selects several measuring points on the tube segment and uses the principle of binocular vision ranging to correct the deflection angle between the XY plane of the camera and the lining circumferential plane, thereby converting the "apparent height difference" measured under perspective into the "true height difference", thus realizing the accurate measurement of the misalignment. Attached Figure Description

[0036] Figure 1 A plan view of the measurement points for actual lining segments;

[0037] Figure 2 A 3D diagram for using a stereo camera;

[0038] Figure 3 A schematic diagram for calculating the "viewpoint difference" for a binocular camera;

[0039] Figure 4 This is a schematic diagram (AB measuring points) for calculating the "true height difference" in this invention.

[0040] Figure 5 This is a schematic diagram (MN measuring point) for calculating the "true height difference" in this invention;

[0041] In the diagram: 1-Segment A; 2-Segment B; 3-Rinse circumferential plane; 4-Binocular camera; 41-Left camera; 42-Right camera; 5-Joint; 51-Joint edge line a; 52-Joint edge line b. Detailed Implementation

[0042] The following is in conjunction with the appendix Figure 1 ~Attached Figure 5 The present invention will be described in conjunction with specific embodiments.

[0043] The coordinate axes of the camera coordinate system for the stereo camera 4 are defined as follows: the X-axis points from the left camera 41 to the right camera 42; the Y-axis is perpendicular to the camera's optical axis (viewing direction); and the Z-axis is the direction of the camera's optical axis, i.e., from the camera towards the scene, forming a right-handed coordinate system with the X-axis and Y-axis. The origin of the world coordinate system is defined at the left camera 41.

[0044] Based on this, the following five steps are used to accurately measure the misalignment of the tunnel segments.

[0045] Step 1: Acquire Images

[0046] Images were captured at the segment misalignment using a binocular camera 4, and the images were distinguished by the labels "left_00.png" and "right_00.png" when saved. A dataset of images of the segment misalignment was created by collecting a large number of similar photographs. During image acquisition, the camera's XY plane was kept as parallel as possible to the lining circumferential cutting plane 3, and the camera's X-axis direction was kept perpendicular to the joint 5 (e.g., ...). Figure 2 (as shown in the figure) to reduce the impact of perspective effect on the measurement results.

[0047] Step 2: Automatically select target measurement points

[0048] In the segment misalignment image dataset, seam 5 is treated as a "crack," while the others are labeled as "background." The misalignment area is automatically identified using the YOLO correlation algorithm, and then the convolutional neural network algorithm is used to detect seam 5 in the misalignment area and accurately locate the two seam edge lines 51 and 52.

[0049] First, target measurement point A is determined at half the length of the seam edge line 51. Then, using a convolutional neural network algorithm, the optimal target measurement point B is found on the seam edge line 52, minimizing the distance between target measurement points A and B. Further, target measurement points M and N are selected at equal pixel intervals on both sides of the AB line. This results in four target measurement points (target measurement points A and M are located on segment 1, and B and N are located on segment 2) used for subsequent 3D coordinate calculations.

[0050] Step 3: Binocular visual ranging Zc and world 3D coordinates

[0051] Using the target measurement point as a feature point, a disparity map is obtained by stereo matching based on the left and right images of the misaligned platform obtained in step one, thereby calculating the distance Z of the target measurement point from the camera's XY plane. c (i.e., apparent depth), and then obtain the three-dimensional coordinates of the target measurement point according to the coordinate system transformation relationship. The specific process is as follows:

[0052] Binocular vision ranging technology involves the following four coordinate systems: world coordinate system O w (X w Y w Z w (Unit: mm), Camera coordinate system O c (X c Y c Z c (Unit: mm), Focal plane coordinate system O f (U f V f (Unit: mm) and pixel coordinate system O p (U p V p (Unit: pixel). Let the world coordinates of target point A be (x... w y w , z w ), camera coordinates are (x c y c , z c ), the focal plane coordinates are (u f v f ) and pixel coordinates are (u p v p ).

[0053] The transformation relationship between the world coordinate system and the camera coordinate system is as follows:

[0054]

[0055] In the formula, R is a 3x3 orthogonal matrix describing the rotation direction from the world coordinate system to the camera coordinate system, and T is a 3x1 translation matrix describing the translation relationship between the world coordinate system and the camera coordinate system. R and T are extrinsic parameters of the stereo camera, provided by the manufacturer or obtained through experimental calibration.

[0056] The relationship between the camera coordinate system and the focal plane coordinate system is as follows:

[0057]

[0058] In the formula, f is the camera focal length, an intrinsic parameter of the binocular camera, provided by the manufacturer or obtained through experimental calibration.

[0059] The relationship between the focal plane coordinate system and the pixel coordinate system is as follows:

[0060]

[0061] In the formula, Δx and Δy represent the number of pixels per unit length (unit: pixel / mm), which can be obtained through experimental calibration by combining the intrinsic parameters of the stereo camera. Coordinates (u o v o ) represents the position of the origin of the focal plane coordinate system in the pixel coordinate system, which is usually taken as the center point of the pixel plane.

[0062] Combining the transformation relationships (1), (2), and (3) above, we can obtain the mapping relationship between the world coordinate system and the pixel coordinate system:

[0063]

[0064] The target automatic recognition YOLO algorithm, convolutional neural network algorithm, and binocular vision ranging used in the above steps have all been widely applied in the field of computer vision and have high reliability and accuracy. The core content of this invention mainly consists of the following two steps.

[0065] Step 4: Deflection angle θ between the camera's XY plane and the lining circumferential plane

[0066] Step three yields the world 3D coordinates of the target measurement points A(x1, y1, z1), B(x2, y2, z2), M(x3, y3, z3), and N(x4, y4, z4). However, since it is difficult to achieve perfect parallelism between the camera's XY plane and the lining ring-cutting plane 3 during actual image acquisition, it is assumed that there is a deflection angle θ between the two planes (here only the case of the camera rotating θ degrees around the Y-axis is considered; deflection angles in other axial directions are similar), such as... Figure 3 As shown. The apparent depth difference between target measuring point A (x1, y1, z1) and target measuring point B (x2, y2, z2) is AF-BG = z1-z2 ("apparent height difference"). Due to the existence of the deflection angle θ, the "apparent height difference" is not equal to the misalignment amount EB.

[0067] To eliminate the influence of the deflection angle θ, the following relationship is established between the world 3D coordinates of the target measurement point and the deflection angle θ. The specific method includes the following 3 steps:

[0068] 1) Figure 4 Trigonometric function relationships

[0069] Draw the line of sight depth AF and BG for target measurement points A and B; AE⊥BD; lining circumferential plane 3 and plane CD, which simultaneously satisfy lining circumferential plane 3 / / plane CD, and AC⊥CD, BD⊥CD; HK⊥BJ intersects at point K;

[0070] Assume that the angle between plane FG and plane CD is θ (plane FG is the initial camera XY plane, and plane CD is the camera XY plane after rotation by θ);

[0071] Then: AC intersects FG at point H, BD intersects FG at point J, and AE = HK.

[0072] Then: AH-BK=AH-(BJ-JK)=EB (5)

[0073] In right triangles AFH, BGJ, and HJK, ∠FAH = ∠GBJ = ∠JHK = θ

[0074] but: JK=HK tanθ=AE tanθ (6)

[0075] Substituting equation (6) into equation (5):

[0076]

[0077] 2) Figure 5 Trigonometric function relationships

[0078] and Figure 4 Similarly, in Figure 5 Construct relevant auxiliary lines in the middle;

[0079] Draw the apparent depth lines MP and NS for target measurement points M and N; AE⊥BE; lining circumferential plane 3 and plane QU, simultaneously satisfying lining circumferential plane 3 / / plane QU, and MQ⊥QU, NU⊥QU, the extension of NU intersects plane PS at point T; draw AM′⊥AE, BN′⊥AE; draw MM′⊥AM′, NN′⊥BN′;

[0080] Similarly, assume that the angle between plane PS and plane QU is θ (plane PS is the initial camera XY plane, and plane QU is the camera XY plane after rotation by θ);

[0081] Then: MQ-NU=MQ-(NT-TU)=EB (8)

[0082] In right triangles MPQ, NST, and QTU, ∠PMQ=∠SNT=∠TQU=θ

[0083] but: TU=QU tanθ=(AE+MM′+NN′)tanθ (9)

[0084] Substituting equation (9) into equation (8) yields:

[0085]

[0086] 3) Solve the system of equations to find the deflection angle θ

[0087] Combining equations (7) and (10), we obtain the following system of equations.

[0088]

[0089] By combining the above equations and eliminating the unknowns AE and EB, we can obtain the expression for the deflection angle θ:

[0090]

[0091] Wherein, since the segment radius ρ is known (e.g. Figure 3 As shown), and the target measurement points M and N are selected with equal pixel spacing:

[0092] but:

[0093] Substituting equation (12) into equation (11), we obtain the formula for calculating the deflection angle θ:

[0094]

[0095] Step 5: Calculate the misalignment amount

[0096] According to the definition of the deflection angle θ in step four, the world three-dimensional coordinate transformation of the target measurement point will satisfy the following relationship:

[0097]

[0098] Right now

[0099] In the formula, (X w Y w Z w Let (X′) be the three-dimensional coordinates of the target measurement point in the original world coordinate system. w ,Y′ w Z′ w This refers to the three-dimensional coordinates of the target measuring point in the new world coordinate system after eliminating the deflection angle θ.

[0100] Therefore, the world three-dimensional coordinates of the four target measurement points in the new world coordinate system are (x′1, y′1, z′1), (x′2, y′2, z′2), (x′3, y′3, z′3) and (x′4, y′4, z′4). The apparent depth difference between measurement points A and B is z′1-z′2 (i.e., the "true height difference"), which is equal to the misalignment value EB.

[0101] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for accurate measurement of misalignment of shield tunnel segments based on binocular vision, characterized in that, Includes the following steps: S1. Collect a large number of similar photos at the misalignment of the tunnel segments using a binocular camera to establish a dataset of tunnel segment misalignment images. S2. Automatically select target measurement points at the misalignment of the tunnel segment using the YOLO algorithm and convolutional neural network algorithm; the specific implementation is as follows: In the image dataset of misaligned pipe segments, the joints are treated as "cracks" and the rest as "background" for image annotation. The misaligned pipe segment areas are automatically identified by the YOLO algorithm, and then the convolutional neural network algorithm is used to detect the joints in the misaligned areas and accurately locate the two joint edge lines. Using a convolutional neural network algorithm, target measurement point A is automatically selected on one seam edge line, and target measurement point B is automatically selected on the other seam edge line, minimizing the distance between target measurement points A and B. Target measurement points M and N are then selected at equal pixel intervals on both sides of the line AB, resulting in four target measurement points for subsequent 3D coordinate calculations. S3. Obtain the world three-dimensional coordinates of the target measurement point using binocular vision ranging technology; S4. Establish a relationship between the world three-dimensional coordinates of the target measurement point and the deflection angle θ due to the fact that the camera XY plane is not completely parallel to the lining circumferential tangent plane, and obtain the deflection angle θ. S5. By transforming the coordinate system, the influence of the deflection angle θ on the apparent depth measurement is eliminated, thereby accurately calculating the misalignment.

2. The method for accurate measurement of misalignment of shield tunnel segments based on binocular vision according to claim 1, characterized in that, Step S1 is implemented as follows: Images were acquired at the misalignment of the tunnel segments using a binocular camera, and the images were distinguished by the labels "left_00.png" and "right_00.png" when saved. A dataset of images of the misalignment of the tunnel segments was established by acquiring a large number of similar photos. When acquiring images, the camera's XY plane was kept as parallel as possible to the lining circumferential cutting plane, and the camera's X-axis direction was kept perpendicular to the joint.

3. The method for accurate measurement of misalignment of shield tunnel segments based on binocular vision according to claim 1, characterized in that, Step S3 is implemented as follows: Using binocular vision ranging technology, in the world coordinate system O w (X) w Y w Z w Camera coordinate system O c (X) c Y c Z c ), focal plane coordinate system O f (U) f V f ) and pixel coordinate system O p (U) p V p By establishing a coordinate transformation relationship between the two images, the disparity map is obtained from the acquired left and right images through stereo matching, and the distance Z from the target measurement point to the camera's XY plane is calculated. c Based on the apparent depth, the world three-dimensional coordinates of the target measurement point are obtained according to the coordinate transformation relationship.

4. The method for accurate measurement of misalignment of shield tunnel segments based on binocular vision according to claim 3, characterized in that, Step S4 is implemented as follows: The target measurement point A is obtained through step S3. B () ), M ( ) and N ( The world three-dimensional coordinates are assumed; since it is difficult to ensure that the camera's XY plane is completely parallel to the lining ring cutting plane during actual image acquisition, it is assumed that there is a deflection angle θ between the two planes. The coordinate axes of a stereo camera are defined as follows: the X-axis points from the left camera to the right camera; the Y-axis is perpendicular to the camera's optical axis, i.e., the line of sight. The Z-axis is the direction of the camera's optical axis, i.e., from the camera towards the scene, and together with the X-axis and Y-axis, it forms a right-handed coordinate system; the origin of the world coordinate system is defined at the left camera. Assume that plane FG is the initial camera XY plane, plane CD is the camera XY plane after rotation θ, the lining ring cutting plane is parallel to plane CD, AF and BG are the line of sight depth of target measurement points A and B respectively, and AC⊥CD, BD⊥CD, draw AE⊥BD intersecting at point E, draw the extensions of AC and BD to plane FG intersecting at points H and J, draw HK⊥BJ intersecting at point K; Similarly, assuming plane PS is the initial camera XY plane, plane QU is the camera XY plane after rotation θ, the lining circumferential plane is parallel to plane QU, MP and NS are the line of sight depth of target measurement points M and N respectively, and MQ⊥QU, NU⊥QU, construct A ∥BE makes ⊥A EB extension line to Make ⊥B Draw the extension of NU to plane PS, intersecting at point T; Based on trigonometric relationships, the following system of equations is established between the world three-dimensional coordinates of the four target measurement points and the deflection angle θ; Among them, AE, EB, , θ is an unknown quantity, and all others are known quantities. ρ is the radius of the lining segment. Solving the above system of equations simultaneously, eliminating AE and EB, the formula for calculating the deflection angle θ is: 。 5. The method for accurate measurement of misalignment of shield tunnel segments based on binocular vision according to claim 4, characterized in that, Step S5 is implemented as follows: According to the definition of the deflection angle θ, the original world three-dimensional coordinates of the target measurement point ( ), transformed into new world three-dimensional coordinates ( ), as shown in the following formula: After coordinate transformation, the new world 3D coordinates of the four target measurement points are as follows: ), ( ), ( )and( The apparent depth difference between the target measurement points A and B is then calculated as follows: This is the misalignment amount EB.

Images