Tunnel surrounding rock structure deformation measurement method based on machine vision and structured light

Abstract

The present application relates to the construction period tunnel monitoring technical field, specifically at tunnel surrounding rock structure deformation measurement method based on machine vision and structure light: S10, in the tunnel ground pressure area's side wall sets up multiple structure light light source, installs observation terminal on the side wall, the observation terminal with multiple structure light light source communication;S20, the observation terminal is provided with image processing unit, the image processing unit is according to the structure light displacement change in the region to be measured, judges whether the surrounding rock structure is deformed;The image processing unit includes image preprocessing module and surrounding rock structure deformation judgment module;The image preprocessing module is used for detecting the first image data pixel points of the region to be measured, judges and processes according to whether the structure light line in the first image data is in clear state, solves the technical problems that the measurement equipment installation and integration of the prior art measurement method are inconvenient and the measurement cost is too high.

Description

Technical Field

[0001] This invention relates to the field of tunnel monitoring technology during construction, specifically to a method for measuring the deformation of tunnel surrounding rock structures based on machine vision and structured light. Background Technology

[0002] During tunnel construction, the stability and safety of the tunnel structure and support system are crucial. Similarly, the quality and safety of railway tunnels are closely related to the construction period. Therefore, it is necessary to monitor the tunnel construction process to ensure construction quality and safety, prevent accidents, and take timely measures to deal with them.

[0003] The methods or equipment for tunnel deformation monitoring mainly include total station measurement, laser scanning, and displacement sensor methods. In addition, machine vision structured light measurement technology is widely used in contour detection, 3D reconstruction, and deformation analysis. The literature ("Research on Deformation Measurement Method Based on Structured Light Point Cloud", Journal of Northwestern Polytechnical University, Xiao Weizhong et al., 2023, 41(01)) records the 3D reconstruction of the surface of the object to be measured to obtain the point cloud of the structural surface. The literature ("Measurement of Shape and Deformation Strain Based on Machine Vision", Nanjing University of Aeronautics and Astronautics, Tao Xuejiao, 2017) takes wheel tread and blade as measurement objects and studies the shape and deformation strain measurement method based on machine vision. Structured light measurement technology is currently mostly used in environments with relatively stable light, such as laboratories or factories. However, in environments with strong light or insufficient light, or when the surface of the object being measured has high reflectivity, transparency, or gloss, the accuracy and precision of the measurement may be affected.

[0004] In existing technologies, high-resolution cameras are often used to meet measurement requirements and improve the resolution of visual measurement images to ensure measurement accuracy. However, configuring high-resolution cameras will lead to larger imaging module size, increased power, and higher costs, which is not conducive to equipment installation and integration.

[0005] To address the aforementioned problems, this invention provides a method for measuring the deformation of tunnel surrounding rock structures based on machine vision and structured light. Summary of the Invention

[0006] The present invention aims to solve the technical problem that existing measurement methods suffer from inconvenient installation and integration of measuring equipment and excessively high measurement costs.

[0007] To solve the aforementioned technical problem, the technical solution adopted by the present invention is: a method for measuring the deformation of tunnel surrounding rock structure based on machine vision and structured light, comprising the following steps:

[0008] S10. Multiple structured light sources are installed on the sidewalls of the underground pressure zone in the tunnel, and observation terminals are installed on the sidewalls. The observation terminals communicate with the multiple structured light sources.

[0009] S20. The observation terminal is equipped with an image processing unit, which determines whether the surrounding rock structure has deformed based on the structural light displacement change in the area to be measured; the image processing unit includes an image preprocessing module and a surrounding rock structure deformation judgment module.

[0010] The image preprocessing module is used to detect the first image data of the area to be tested and to process it according to whether the structure light in the first image data is clear.

[0011] When the structured light is unclear, the image preprocessing module activates an early warning, causing the first image data to enter the segmentation module. The segmentation module is used to re-segment the real-time first image data to obtain clear second image data.

[0012] Preferably, processing based on whether the structured rays in the first image data are clear includes:

[0013] When the structure is clearly illuminated, the first or second image data is transmitted to the surrounding rock structure deformation judgment module. The surrounding rock structure deformation judgment module is used to calculate the change data of the segmented image in the area to be tested, and to determine in real time whether deformation or movement has occurred based on the initial position of the arch S and / or sidewall h of the surrounding rock structure, so as to understand the stability and safety of the tunnel structure and support system in real time.

[0014] Preferably, the method for distortion correction of the observation terminal includes the following steps:

[0015] S11. Obtain the calibrated intrinsic parameters and distortion parameters of the observation terminal, and obtain the relationship between the coordinates of the spatial 3D reference point M and the pixel coordinates of the observation terminal based on the pinhole principle, and obtain the pixel coordinates.

[0016] S12. Measure the camera focal length f from the parameters within the observation terminal. x / f y Optical center coordinates c x / c y Substitute the pixel coordinates into the calculation, and transform the pixel coordinates into the observation terminal coordinates:

[0017] S13. Correct the coordinate displacement of the measuring camera by applying the radial distortion parameters k1, k2, and k3, and the tangential distortion parameters p1 and p2 from the calibrated distortion parameters, and obtain the corrected observation terminal coordinates corresponding to the pixels:

[0018]

[0019] S14. Finally, the corrected observation terminal coordinates are converted into image pixel coordinates:

[0020] Preferably, obtaining the second image data includes the following steps:

[0021] S21. The image preprocessing module receives the first image data of the area to be measured, which is recorded in real time by the observation terminal after correction.

[0022] S22. The region segmentation module uses the PSPNet image segmentation network to extract the structured light region and adds a convolutional layer to the last layer of the PSPNet architecture. The first image is downsampled by 8 times to form a feature map. The feature map is input into the PPM module and added to its output. Finally, the result is obtained by convolution and upsampling by 8 times bilinear interpolation, which is to obtain the super-resolution segmented image.

[0023] S23. Dice Loss analysis is used to obtain the pixel labels of the real segmented images and the pixel categories of the segmented images predicted by the model.

[0024] Based on the pixel labels of the obtained ground truth segmented image and the pixel categories of the model-predicted segmented image, the overlap between the predicted and ground truth results is calculated: Obtain clear second image data;

[0025] S24. Obtain the pixel coordinates of the super-resolution structured light image based on the clear image region:

[0026] S25. Transmit the obtained structured light image pixel coordinates to the surrounding rock structure deformation judgment module to analyze the changes of the structured light image in the tunnel surrounding rock structure area of ​​the real-time super-resolution segmentation image.

[0027] Preferably, step S23 further includes the following steps:

[0028] S231. Compare the super-resolution structured light image with the tunnel cross-section to obtain the left and right rotation angle α between them. c =α 1c +α 2c Pitch angle β c =β 1c +β 2c ;

[0029] S232, Based on the left and right rotation angle α of the observation terminal c Pitch angle β c The installation angles of the measuring camera and the main station were adjusted to correct the errors, and the pixel coordinates of the structured light image after angle correction were obtained. Used to ensure the accuracy of image data.

[0030] Preferably, the calculation when the structured light is clear includes the following steps:

[0031] S31. Extract the structured rays from the first image data or the second image data in real time, and take the highest point of the center line of the structured ray as the arch vertex, and divide multiple sidewall convergence regions on both sides of the arch vertex.

[0032] S32. Match the real-time structural rays with the reference structural rays stored in the surrounding rock structure deformation judgment module, and analyze the positional deformation of the pixel crown and / or side; the positional deformation is a variable in both vertical and horizontal directions.

[0033] Based on real-time assessment of whether the surrounding rock structure and support system of the tunnel are in a stable and safe state, support is provided for the prediction of tunnel structure deformation;

[0034] S33. By comparing the arch height and sidewall convergence point deformation obtained from the analysis with the deformation measured by the total station, the accuracy of the image preprocessing module is verified.

[0035] Preferably, a relatively stable secondary lining area is provided behind the observation terminal, and a reference structure light beam is erected on the road surface within the secondary lining area. The position of the reference structure light beam remains unchanged and is used to monitor the position of the observation terminal in real time.

[0036] The positional changes obtained by the observation terminal in real time are transmitted to the surrounding rock structure deformation judgment module to compensate for the error in the deformation of the arch and sidewalls in the image caused by the changes of the observation terminal itself, and to obtain the actual deformation of the arch and sidewalls.

[0037] The beneficial effects of this invention are as follows:

[0038] 1. Compared with the existing tunnel structure deformation measurement method based on machine vision structured light, the present invention uses PSPNet network for image super-resolution and foreground and background segmentation, which improves the image resolution while ensuring the measurement accuracy requirements. It also calculates the deformation of the tunnel arch and sidewalls in real time based on the segmented structured light area. This avoids the problems of the original method, which required replacing the camera with a high-resolution one, resulting in larger imaging module size, increased power, increased cost, and difficulties in equipment installation and integration.

[0039] 2. This invention uses the PSPNet image segmentation network to extract structured light regions through a region segmentation module. A convolutional layer is added to the last layer of the PSPNet architecture to downsample the first image by 8 times to create a feature map. This feature map is input to the PPM module and its output is added. Finally, after convolution and 8-fold bilinear interpolation upsampling, the result is obtained, i.e., the super-resolution segmented image. This improves the accuracy of comparing image data with historical benchmark data, meeting the requirements for measurement accuracy. Furthermore, the difference between the measured structural deformation values ​​and those obtained using a total station in existing technologies is small, thus verifying the feasibility of using the PSPNet network for image super-resolution and foreground / background segmentation, providing support for tunnel structure deformation prediction.

[0040] 3. This invention enables real-time measurement via an observation terminal, supplementing the data between two measurements by the total station, providing abundant data for tunnel structure deformation prediction, and effectively ensuring the continuity and timeliness of tunnel monitoring.

[0041] 4. This invention avoids the measurement camera from being affected by vibration or external forces due to tunnel construction during long-term observation in the tunnel, thus preventing the measurement camera from changing. Furthermore, it compensates for the cross-sectional deformation based on the measured camera's changes obtained under monitoring, thereby ensuring the accuracy of the image data. Attached Figure Description

[0042] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below.

[0043] Figure 1 This is a flowchart illustrating the method for measuring the deformation of tunnel surrounding rock structures based on machine vision and structured light according to the present invention.

[0044] Figure 2 This is a schematic diagram of the distortion correction process for the observation terminal of the present invention;

[0045] Figure 3 This is a schematic diagram of the PSPNet-based depth segmentation-super-resolution network of the present invention;

[0046] Figure 4 This is a schematic diagram illustrating the process of determining the pixel coordinates of the super-resolution structured light image according to the present invention.

[0047] Figure 5 This is a schematic diagram of the angle between the measuring camera and the imaging plane of the present invention;

[0048] Figure 6This is a schematic diagram illustrating the process of determining pixel coordinates after angle correction according to the present invention.

[0049] Figure 7 This is a schematic diagram of tunnel image displacement measurement according to the present invention;

[0050] Figure 8 This is a schematic diagram of the tunnel deformation calculation process of the present invention;

[0051] Figure 9 This is a schematic diagram comparing the data from the DK0+440-TCMS and the total station arch monitoring data of the Yigong Tunnel left tunnel in the experimental results of this invention.

[0052] Figure 10 This is a schematic diagram comparing the experimental results of the present invention with the monitoring data of the DK315+020-TCMS and the total station in the Zhonghua Shengshan Tunnel.

[0053] Figure 11 This is a schematic diagram comparing the experimental results of the Zhonghua Shengshan Tunnel measurement using DK315+025-TCMS with the monitoring data from a total station. Detailed Implementation

[0054] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0055] like Figure 1 As shown in the figure, this embodiment proposes a method for measuring the deformation of tunnel surrounding rock structures based on machine vision and structured light, including the following steps:

[0056] S10. Multiple structured light sources are installed on the sidewalls of the underground pressure zone in the tunnel, and observation terminals are installed on the sidewalls. The observation terminals communicate with the multiple structured light sources.

[0057] S20. The observation terminal is equipped with an image processing unit, which determines whether the surrounding rock structure has deformed based on the structural light displacement change in the area to be measured; the image processing unit includes an image preprocessing module and a surrounding rock structure deformation judgment module.

[0058] The image preprocessing module is used to detect the first image data of the area to be tested and to process it according to whether the structure light in the first image data is clear.

[0059] When the structure light is unclear, the image preprocessing module activates an early warning, causing the first image data to enter the segmentation module, wherein the segmentation module is used to re-segment the real-time first image data to obtain clear second image data.

[0060] When the structure is clearly visible, the first image data or the second image data is transmitted to the surrounding rock structure deformation judgment module. The surrounding rock structure deformation judgment module is used to calculate the change data of the segmented image in the area to be tested, and to judge in real time whether deformation and movement have occurred based on the initial position of the arch S and / or sidewall h of the surrounding rock structure, so as to understand the stability and safety of the tunnel structure and support system in real time.

[0061] Tunnel workers can flexibly adjust the position of the structured light source on the sidewalls of areas such as tunnel bias, landslides, and high ground stress zones. Once the position of the structured light source is determined, it is turned on, and it is observed whether structured light rays can be formed on the tunnel cross-section. After confirming that complete structured light rays can be formed, an observation terminal is set up near the structured light equipment and calibrated. The calibrated structured light area image is acquired in real time. Then, the first image pixel of the area to be measured is detected by the image preprocessing module to determine whether the structured light rays in the image are clear. If the structured light rays in the first image are not clear, the first image data with insufficient resolution will enter the segmentation module to obtain a high-resolution second image, in which the structured light rays are clear. This ensures the accuracy of the measurement and avoids the problems that originally required replacing the camera with a high-resolution one to ensure measurement accuracy, which would have led to larger imaging module size, increased power, increased cost, and difficulties in equipment installation and integration.

[0062] If the structure light rays in the first image are clear, the resolution of the first image meets the requirements, and it directly enters the surrounding rock structure deformation judgment module. Based on the deformation caused by the modulation of the structure light illuminating the surface of the object under test, the image three-dimensional shape data is used to reflect the surface morphology information of the object under test in real time. The real-time three-dimensional shape data is transmitted to the surrounding rock structure deformation judgment module. The surrounding rock structure deformation judgment module compares the size and shape of the structure light ray area segmented in real time with the size and shape obtained from the reference structure light ray, and obtains the real-time accurate measurement results of the changes in arch deformation and sidewall convergence. It monitors the deformation and movement of the tunnel surrounding rock structure arch and sidewall, and understands the stability and safety of the tunnel structure and support system in real time.

[0063] like Figure 2 As shown, the method for distortion correction of the observation terminal includes the following steps:

[0064] S11. Obtain the calibrated intrinsic parameters and distortion parameters of the observation terminal, and obtain the relationship between the coordinates of the spatial 3D reference point M and the pixel coordinates of the observation terminal based on the pinhole principle, and obtain the pixel coordinates.

[0065] S12. Measure the camera focal length f from the parameters within the observation terminal. x / f y Optical center coordinates c x / c y Substitute the pixel coordinates into the calculation, and transform the pixel coordinates into the observation terminal coordinates:

[0066] S13. Correct the coordinate displacement of the measuring camera by applying the radial distortion parameters k1, k2, and k3, and the tangential distortion parameters p1 and p2 from the calibrated distortion parameters, and obtain the corrected observation terminal coordinates corresponding to the pixels:

[0067]

[0068] S14. Finally, the corrected observation terminal coordinates are converted into image pixel coordinates:

[0069] After the observation terminal location is set, the first step is to calibrate the measuring camera or video camera within the observation terminal and obtain the parameters (camera focal length f) of the measuring camera or video camera. x / f y Optical center coordinates c x / c y And distortion parameters (radial distortion parameters k1, k2, and k3, tangential distortion parameters p1 and p2); according to the camera pinhole principle, the relationship between the coordinates of the spatial 3D reference point M and its pixel coordinates is as follows:

[0070]

[0071] The relationship between the coordinates of a 3D spatial reference point M and its pixel coordinates is given by the formula. Represents coordinates in the pixel coordinate system. Let represent the coordinates in the world coordinate system, and s be the scale factor. The relationship between a point M in the world coordinate system (Z=0) plane and its image coordinates in the image plane can be linked using the following homography matrix:

[0072] H = A[r1,r2,t]

[0073] The camera intrinsic parameter matrix A has 5 parameters. The common Zhang Zhengyou calibration method requires taking multiple images of the checkerboard calibration board to solve for the parameters. By changing the relative position between the camera and the measurement surface, N different images are obtained, forming 2N equations to solve for the camera intrinsic parameters, distortion parameters, and extrinsic parameters. Using the camera intrinsic parameters, the pixel coordinates (u,v) are first converted into camera coordinates:

[0074]

[0075] The calibrated distortion parameters are used to correct the camera coordinate displacement, resulting in the corrected camera coordinates for each pixel:

[0076]

[0077] Where k1, k2 and k3 are mirror distortion parameters, and p1 and p2 are tangential distortion parameters;

[0078] Then use camera coordinates Transform the corrected camera coordinates to image pixel coordinates:

[0079]

[0080] Then, the measuring camera corrects the image according to the calibrated parameters.

[0081] like Figures 3 to 4 As shown, obtaining the second image data includes the following steps:

[0082] S21. The image preprocessing module receives the first image data of the area to be measured, which is recorded in real time by the observation terminal after correction.

[0083] S22. The region segmentation module uses the PSPNet image segmentation network to extract the structured light region, and adds a convolutional layer to the last layer of the PSPNet architecture to downsample the first image by 8 times to form a feature map. The feature map is input into the PPM module and added to its output. Finally, the result is obtained by convolution and upsampling by 8 times bilinear interpolation, which is to obtain the super-resolution segmented image.

[0084] S23. Dice Loss analysis is used to obtain the pixel labels of the real segmented images and the pixel categories of the segmented images predicted by the model.

[0085] Based on the pixel labels of the obtained ground truth segmented image and the pixel categories of the model-predicted segmented image, the overlap between the predicted and ground truth results is calculated: Obtain clear second image data;

[0086] S24. Obtain the pixel coordinates of the super-resolution structured light image based on the clear image region:

[0087] S25. Transmit the obtained structured light image pixel coordinates to the surrounding rock structure deformation judgment module to analyze the changes in the structured light image within the tunnel surrounding rock structure area of ​​the real-time super-resolution segmentation image.

[0088] The region segmentation module uses the PSPNet image segmentation network to extract structured light regions. A convolutional layer is added to the last layer of the PSPNet architecture to downsample the first image by 8 times to create a feature map. This feature map is input to the PPM module and its output is added. Finally, after convolution and upsampling with 8 times bilinear interpolation, the result is obtained, which is the super-resolution segmented image. Then, Dice Loss is applied to calculate the overlap between the predicted and true results.

[0089]

[0090] Where |X∩Y| is the intersection of X and Y, and |X| and |Y| represent the number of elements in X and Y, respectively;

[0091] X represents the pixel label of the real segmented image, and Y represents the pixel category of the segmented image predicted by the model. Dice Loss can mitigate the negative effects of foreground-background (area) imbalance in the image. Foreground-background imbalance means that most areas of the image do not contain the target, and only a small part of the image contains the target.

[0092] The image acquired after distortion correction by the main station camera is subjected to super-resolution and segmentation. The pixel coordinates of the super-resolution structured light image are:

[0093]

[0094] Where δ is the super-resolution error;

[0095] By obtaining the pixel coordinates of the super-resolution structured light image, measurement accuracy is ensured. This method replaces the existing technology that requires replacing the high-resolution camera to obtain high-resolution images. It solves the problems of increased imaging module size, power consumption, and cost caused by replacing the camera with a high-resolution one, which are detrimental to equipment installation and integration. At the same time, in the harsh environment of tunnels under construction (the tunnel is affected by the headlights of large vehicles, dust, blasting, etc.), the structured light image processing method based on depth image segmentation and super-resolution technology can stably and reliably measure the deformation of tunnel structures.

[0096] like Figures 5 to 6 As shown, step S23 further includes the following steps:

[0097] S231. Compare the super-resolution structured light image with the tunnel cross-section to obtain the left and right rotation angle α between them. c =α 1c +α 2c Pitch angle β c =β 1c +β 2c ;

[0098] S232, Based on the left and right rotation angle α of the observation terminal c Pitch angle β c The installation angles of the measuring camera and the main station were adjusted to correct the errors, and the pixel coordinates of the structured light image after angle correction were obtained. Used to ensure the accuracy of image data;

[0099] Where the left and right rotation angles α c It consists of the internal installation angle of the main station camera and the installation angle of the main station bracket relative to the tunnel axis, with the internal installation left and right rotation angle α. 1c The left and right rotation angle α of the main station relative to the tunnel axis 2c ,but,

[0100] α c =α 1c +α 2c

[0101] Pitch angle β c It consists of the internal installation angle of the main station camera and the installation angle of the main station bracket relative to the tunnel axis, with the internal installation pitch angle β. 1c The pitch angle of the main station relative to the tunnel axis is β. 2c ,but,

[0102] β c =β 1c +β 2c

[0103] Rotate left and right by angle α c Pitch angle β c The pixel coordinates of the super-resolution structured light image are Applied to formulas:

[0104]

[0105] Obtain the angle-corrected pixel coordinates (u”, v”);

[0106] The installation angles of the main station bracket and the main station camera are adjusted based on the obtained left and right rotation angles and pitch angles, and the pixel coordinates after angle correction are obtained to ensure the accuracy of the first or second image data.

[0107] like Figures 7 to 8 As shown, the calculation when the structured light is clear includes the following steps:

[0108] S31. Extract the structured rays from the first image data or the second image data in real time, and take the highest point of the center line of the structured ray as the arch vertex, and divide multiple sidewall convergence regions on both sides of the arch vertex.

[0109] S32. Match the real-time structural rays with the reference structural rays stored in the surrounding rock structure deformation judgment module, and analyze the positional deformation of the pixel crown and / or side; the positional deformation is a variable in both vertical and horizontal directions.

[0110] Based on real-time assessment of whether the tunnel surrounding rock structure and support system are in a stable and safe state, support is provided for tunnel structure deformation prediction;

[0111] S33. By comparing the arch height and sidewall convergence point deformation obtained from the analysis with the deformation measured by the total station, the accuracy of the image preprocessing module is detected.

[0112] The observation terminal is provided with a relatively stable secondary lining area behind it, and a reference structure light beam is erected on the road surface within the secondary lining area. The position of the reference structure light beam remains unchanged and is used to monitor the position of the observation terminal in real time.

[0113] The positional change data obtained by the observation terminal in real time is transmitted to the surrounding rock structure deformation judgment module to compensate for the error in the deformation of the arch and sidewall in the image caused by the changes of the observation terminal itself, and to obtain the actual deformation of the arch and sidewall.

[0114] After the above steps, the corrected camera coordinates are converted to image pixel coordinates. The super-resolution structured light image pixel coordinates are determined, along with the camera installation angle and the main station installation angle error. The pixel coordinates are then corrected. Based on the image data, the arch and sidewall feature elements are extracted. The highest point of the structured light curve centerline is selected as the arch apex. The arch height and the height of point S2 are known. Based on the height standard of the peripheral convergence measurement points, the peripheral convergence measurement points can be calculated. Then, the initial observation point coordinates (u'1', v'1') are set. The camera is used to record and measure the coordinates of the observation point at a certain moment, which is (u'2', v'2'). The vertical and horizontal deformations s and h at this moment are:

[0115]

[0116] Simultaneously, a reference structural ray is installed at the secondary lining behind the main station camera, and the position of the reference structural ray remains unchanged. This is used to detect and compensate for the variations that the measuring camera may experience due to vibration or external forces caused by long-term observation inside the tunnel. The compensation is based on the observed deformation of the reference terminal section. The vertical deformation at the apex of the reference terminal arch is S. r The horizontal deformation of the sidewall is h. r Then the compensated deformations s and h are:

[0117]

[0118] Sr and h r The measurement mechanism is moved by vibration. The values ​​of s and h change according to different times. The change can be obtained by the difference between the observation point after the change and the initial observation point. Thus, the real-time vertical and horizontal deformation at a certain moment can be obtained. Then, the smoothed coordinates of the window can be calculated based on the deformation.

[0119] Window smoothing:

[0120]

[0121] Finally, the deformation of the tunnel's surrounding rock structure, including the crown height and the deformation at the sidewall convergence point, is obtained through real-time comparative analysis using the surrounding rock structure deformation judgment module. The obtained deformation is then compared with the deformation measured by the total station.

[0122] Experimental Results: By developing a prototype, the method for measuring tunnel structural deformation based on machine vision and structured light was verified. It was tested in the construction of projects such as the left entrance of the Yigong Tunnel on the CZ Railway and the Huayingshan Tunnel on the Xi'an-Chongqing Railway. The Huayingshan Tunnel is a gas tunnel, and the equipment was designed with explosion-proof features and explosion-proof networking in mind. Based on the test results, the equipment was improved and upgraded, achieving good application results.

[0123] Analysis of measurement results for the left tunnel of Yigong Tunnel:

[0124] The system was deployed at the left entrance of the Yigongdao tunnel on March 1, 2023, and monitored a total of 44 cross-sections by May 31, 2023, with mileage markers ranging from DKO+130 to DKO+590. Measurement data from the DKO+440 cross-section was compared with total station data. Table 1 shows the total station arch monitoring data for DKO+440 from May 1 to May 31, 2023.

[0125] Table 1: Summary of DK1097+440 Total Station Vault Monitoring Data

[0126]

[0127]

[0128] The settlement value was recalculated using the time when the total station and this system started monitoring as the starting point. The comparison of the arch monitoring data from this system and the total station at DK0+440 is as follows: Figure 9 As shown.

[0129] according to Figure 9It is known that at the measuring point mileage DKO+440, the overlap period between the monitoring by this system and the total station was 3 days. During this period, the system performed 182 measurements, while the total station performed 3 measurements. The number of measurements performed by this method was approximately 60 times that of the total station. Within the same measurement time, the maximum difference between the TCMS arch monitoring value and the total station arch monitoring value was 0.016 mm.

[0130] Analysis of measurement results for Huayingshan Tunnel:

[0131] Table 2 shows the monitoring data of the arches of total stations DK315+020 and DK315+025 from May 21 to May 27, 2023.

[0132] Table 2: Summary of Monitoring Data from the DK315+020 Total Station Arch

[0133]

[0134] 1) Recalculate the settlement value using the time when the total station and this method started monitoring, taking the time close to the start time. The comparison between the data from this system and the total station at section DK315+020 is as follows: Figure 10 As shown.

[0135] At the measuring point mileage DK315+020, the overlap period between TCMS and total station monitoring was one day. During this period, TCMS made 9 measurements, and the total station made 2 measurements, with the number of TCMS measurements being 4.5 times that of the total station measurements. Within the same measurement time, the maximum difference between the TCMS and total station arch top monitoring values ​​was 0.236 mm.

[0136] 2) Recalculate the settlement value using the approximate start time of the total station and TCMS monitoring as the starting point. The comparison of TCMS and total station data for DK315+025 is as follows: Figure 11 As shown.

[0137] The measuring point mileage is DK315+025. The overlap time between TCMS and total station monitoring is 1 day. During this period, the method was used to measure 14 times, the total station was used to measure 2 times, and the number of TCMS measurements was 7 times that of the total station. Within the same measurement time, the maximum difference between the arch crown monitoring value of the method and the arch crown monitoring value of the total station was 0.134 mm.

[0138] In summary, under the harsh environment of tunnels under construction, the machine vision-based structured light method for measuring tunnel structural deformation, combined with the characteristics of structured light imaging and structured light image processing based on depth image segmentation and super-resolution technology, can achieve real-time and stable measurement of tunnel structural deformation. Furthermore, comparison with total station measurement data shows minimal difference between the proposed method's deformation measurements and those obtained by the total station, providing support for tunnel structural deformation prediction. Additionally, this system can perform real-time measurements, supplementing the data between two total station measurements, providing abundant data for tunnel structural deformation prediction, and effectively ensuring the continuity and timeliness of tunnel monitoring.

[0139] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for measuring the deformation of tunnel surrounding rock structures based on machine vision and structured light, characterized in that, Includes the following steps: S10. Multiple structured light sources are installed on the sidewalls of the underground pressure zone in the tunnel, and observation terminals are installed on the sidewalls. The observation terminals communicate with the multiple structured light sources. S20. The observation terminal is equipped with an image processing unit, which determines whether the surrounding rock structure has deformed based on the structural light displacement change in the area to be measured; the image processing unit includes an image preprocessing module and a surrounding rock structure deformation judgment module. The image preprocessing module is used to detect the first image data of the area to be tested and to process it according to whether the structure light in the first image data is clear. When the structure light is unclear, the image preprocessing module activates an early warning, causing the first image data to enter the segmentation module, wherein the segmentation module is used to re-segment the real-time first image data to obtain clear second image data. Obtaining the second image data includes the following steps: S21. The image preprocessing module receives the first image data of the area to be measured, which is recorded in real time by the observation terminal after correction. S22. The region segmentation module uses the PSPNet image segmentation network to extract the structured light region, and adds a convolutional layer to the last layer of the PSPNet architecture to downsample the first image by 8 times to form a feature map. The feature map is input into the PPM module and added to its output. Finally, the result is obtained by convolution and upsampling by 8 times bilinear interpolation, which is to obtain the super-resolution segmented image. S23. Dice Loss analysis is used to obtain the pixel labels of the real segmented images and the pixel categories of the segmented images predicted by the model. Based on the pixel labels of the obtained ground truth segmented image and the pixel categories of the model-predicted segmented image, the overlap between the predicted and ground truth results is calculated: This yields clear second image data; where X represents the pixel label of the real segmented image, and Y represents the pixel category of the segmented image predicted by the model. S24. Obtain the pixel coordinates of the super-resolution structured light image based on the clear image region. : Where δ is the super-resolution error; S25. Transmit the obtained structured light image pixel coordinates to the surrounding rock structure deformation judgment module to analyze the changes of the structured light image in the tunnel surrounding rock structure area of ​​the real-time super-resolution segmentation image.

2. The method for measuring the deformation of tunnel surrounding rock structure based on machine vision and structured light according to claim 1, characterized in that, Processing based on whether the structured light rays in the first image data are clear includes: When the structure is clearly illuminated, the first or second image data is transmitted to the surrounding rock structure deformation judgment module. The surrounding rock structure deformation judgment module is used to calculate the change data of the segmented image in the area to be tested, and to determine in real time whether deformation or movement has occurred based on the initial position of the arch S and / or sidewall h of the surrounding rock structure, so as to understand the stability and safety of the tunnel structure and support system in real time.

3. The method for measuring the deformation of tunnel surrounding rock structure based on machine vision and structured light according to claim 1, characterized in that, The method for distortion correction of the observation terminal includes the following steps: S11. Obtain the calibrated intrinsic parameters and distortion parameters of the observation terminal, and obtain the relationship between the coordinates of the spatial 3D reference point M and the pixel coordinates of the observation terminal based on the pinhole principle, and obtain the pixel coordinates. S12. Measure the camera focal length f from the parameters within the observation terminal. x and f y Optical center coordinates c x and c y Substitute the pixel coordinates into the calculation, and transform the pixel coordinates into the observation terminal coordinates: ; S13. Correct the coordinate displacement of the measuring camera by applying the radial distortion parameters k1, k2, and k3, and the tangential distortion parameters p1 and p2 from the calibrated distortion parameters, and obtain the corrected observation terminal coordinates corresponding to the pixels: ; S14. Finally, the corrected observation terminal coordinates are converted into image pixel coordinates. .

4. The method for measuring the deformation of tunnel surrounding rock structure based on machine vision and structured light according to claim 1, characterized in that, Step S23 further includes the following steps: S231. Compare the super-resolution structured light image with the tunnel cross-section to obtain the left and right rotation angle α between them. c =α 1c +α 2c Pitch angle β c =β 1c +β 2c ; where α 1c For internal installation, the left and right rotation angle is α 2c The left and right rotation angle of the main station relative to the tunnel axis; β 1c For internal installation of pitch angle, β 2c The pitch angle of the main station relative to the tunnel axis; S232, Based on the left and right rotation angle α of the observation terminal c Pitch angle β c The installation angles of the measuring camera and the main station were adjusted to correct the errors, and the pixel coordinates of the structured light image after angle correction were obtained. This is used to ensure the accuracy of image data.

5. The method for measuring the deformation of tunnel surrounding rock structure based on machine vision and structured light according to claim 2, characterized in that, The calculation, performed when the structure's light is clear, includes the following steps: S31. Extract the structured rays from the first image data or the second image data in real time, and take the highest point of the center line of the structured ray as the arch vertex, and divide multiple sidewall convergence regions on both sides of the arch vertex. S32. Match the real-time structural light with the reference structural light stored in the surrounding rock structure deformation judgment module, and analyze the positional deformation of the pixel dome and / or side. The positional deformation is a variable in both vertical and horizontal directions; Based on real-time assessment of whether the surrounding rock structure and support system of the tunnel are in a stable and safe state, support is provided for the prediction of tunnel structure deformation; S33. By comparing the arch height and sidewall convergence point deformation obtained from the analysis with the deformation measured by the total station, the accuracy of the image preprocessing module is verified.

6. The method for measuring the deformation of tunnel surrounding rock structure based on machine vision and structured light according to claim 1, characterized in that, The observation terminal is provided with a relatively stable secondary lining area behind it, and a reference structure light beam is erected on the road surface within the secondary lining area. The position of the reference structure light beam remains unchanged and is used to monitor the position of the observation terminal in real time. The positional changes obtained by the observation terminal in real time are transmitted to the surrounding rock structure deformation judgment module to compensate for the error in the deformation of the arch and sidewalls in the image caused by the changes of the observation terminal itself, and to obtain the actual deformation of the arch and sidewalls.

Images