Binocular vision-based device and method for detecting segment assembly quality in shield construction period

Abstract

The invention relates to the technical field of shield tunnel segment assembly quality detection, in particular to a shield construction period segment assembly quality detection device and method based on binocular vision. The shield construction period segment assembling quality detection device based on binocular vision comprises a main structure assembly, an automatic walking assembly and a binocular vision imaging assembly, the main structure assembly is provided with the automatic walking assembly, and the automatic walking assembly is further provided with the binocular vision imaging assembly. According to the device and method provided by the invention, the segment assembly quality in the shield tunnel construction period can be intelligently detected in real time, so that the efficiency and precision of segment assembly quality detection are improved.

Description

Technical Field

[0001] The present application relates to the technical field of shield tunnel segment assembly quality detection, and in particular to a device and method for detecting segment assembly quality during shield tunnel construction based on binocular vision. Background Art

[0002] Shield tunnels have gradually become the preferred method for urban tunnel construction due to their high efficiency, high quality, environmental protection, safety, and low impact on the surrounding environment. The assembly quality of shield tunnel segments is directly related to the safety and durability of shield tunnels. The assembly quality inspection during the construction period is an important part of the shield tunnel construction quality inspection.

[0003] The existing shield tunnel segment assembly quality inspection methods mainly include manual inspection and automated inspection. Among them, manual inspection is mainly through visual inspection and manual measurement of segment misalignment, opening amount and other indicators, while automated inspection uses automated equipment for high-precision inspection. Manual inspection is inefficient, time-consuming and labor-intensive, and it is difficult to accurately and comprehensively detect the actual maximum misalignment value of the segment; although automated inspection can improve the inspection accuracy, due to the high technical threshold and expensive equipment cost, it is difficult to widely popularize because it cannot monitor the segment assembly quality in real time during the shield tunnel construction period.

[0004] The above-mentioned existing technical means have the problem of low detection efficiency when detecting the quality of segment assembly during shield tunnel construction, especially after the segment assembly is completed and the shield tail is pulled out, it is difficult to achieve real-time detection of the segment assembly quality.

[0005] Therefore, based on the above problems, the prior art needs to be improved. Summary of the invention

[0006] The first purpose of this application is to provide a device for detecting the quality of segment assembly during shield construction based on binocular vision.

[0007] The above-mentioned technical purpose of the present application is achieved through the following technical solutions: a device for detecting the quality of segment assembly during shield construction based on binocular vision, comprising a main structure component, an automatic running component and a binocular vision imaging component; an automatic running component is arranged on the main structure component, and the main structure component provides guidance and positioning for the movement of the automatic running component so that the automatic running component moves along the trajectory set by the main structure component; a binocular vision imaging component is arranged on the automatic running component.

[0008] By adopting the above technical solution, the main structure component provides a stable movement foundation and precise guidance for the automatic running component, ensuring that the automatic running component can accurately move along the preset trajectory. The binocular vision imaging component set on the automatic running component can perform comprehensive and accurate image acquisition of the shield tunnel segments as the automatic running component moves, thereby realizing real-time detection of the segment assembly quality. This design has a compact structure, and the components work together to improve the reliability and detection accuracy of the detection device.

[0009] Optionally, the main structure component includes an arc-shaped beam, which is arranged at the connecting bridge between the shield tail of the shield machine and the rear supporting trolley, and a movable track is arranged on the arc-shaped beam; the automatic traveling component includes a walking trolley, which is provided with a first walking wheel and a second walking wheel, and the first walking wheel and the second walking wheel are spaced apart to form an installation gap, and the walking trolley is installed and engaged on the movable track through the installation gap, and the first walking wheel cooperates with the second walking wheel to clamp the movable track.

[0010] By adopting the above technical solution, the moving track on the arc beam provides a clear movement path for the trolley, ensuring that the trolley can move stably on the arc beam. The first and second running wheels on the trolley are engaged with the moving track through the installation gap and clamp the moving track. This design enables the trolley to maintain a stable posture during movement and is not easy to deviate from the track, thereby ensuring that the automatic running component can accurately move along the track of the arc beam, providing a reliable moving foundation for the normal operation of the binocular vision imaging component.

[0011] Optionally, a mounting groove is provided on the arc-shaped beam, a chain is provided in the mounting groove, and a tensioning member is provided on the arc-shaped beam for fixing both ends of the chain; a driving member is provided on the walking trolley, and the driving member is transmission-connected with a sprocket, the sprocket and the chain are meshed, and the sprocket is driven to rotate by the driving member to drive the walking trolley to move.

[0012] By adopting the above technical solution, the chain set in the installation groove on the arc beam and the tensioning parts used to fix the two ends of the chain provide the basis for power transmission for the movement of the trolley. The driving part on the trolley is connected to the sprocket transmission, and the sprocket is meshed with the chain. When the driving part drives the sprocket to rotate, the meshing action of the sprocket and the chain enables the trolley to move along the direction of the chain. This transmission method is stable and reliable, and can effectively drive the trolley to move on the arc beam, ensuring that the binocular vision imaging component can accurately reach the designated image acquisition position to detect the assembly quality of the pipe segment.

[0013] Optionally, a mounting part is provided on the walking trolley, and the binocular vision imaging component includes a baseline calibration ruler arranged on the mounting part, and a slide groove is arranged inwardly on the baseline calibration ruler, and a first mounting seat and a second mounting seat are arranged in the slide groove, and the first mounting seat and the second mounting seat can both slide along the slide groove, and a first camera is arranged on the first mounting seat, and a second camera is arranged on the second mounting seat.

[0014] By adopting the above technical solution, the mounting parts on the trolley provide an installation basis for the binocular vision imaging component. The baseline calibration ruler is set on the mounting parts, and the first mounting seat and the second mounting seat in the slide slot are slidable, so that the first camera and the second camera can adjust their positions in the slide slot. This design enables the binocular vision imaging component to flexibly adjust the spacing and angle of the cameras according to actual needs, thereby obtaining more accurate segment image information, and improving the accuracy and reliability of shield tunnel segment assembly quality inspection.

[0015] Optionally, both the first camera and the second camera are configured with a shadowless light source.

[0016] By adopting the above technical solution and configuring a shadowless light source on the first camera and the second camera, the problem of insufficient light inside the tunnel can be effectively solved. The shadowless light source can provide uniform and bright light, reduce the influence of shadows and reflections, and enable the camera to clearly capture the details of the pipe segment, thereby improving the quality and clarity of the image. This is very important for accurately detecting the assembly quality of shield tunnel segments, helping to discover defects, cracks and other problems on the surface of the pipe segment, and providing a reliable basis for subsequent quality evaluation and improvement.

[0017] Optionally, a blue light scanner is provided on the baseline calibration ruler.

[0018] By adopting the above technical solution, the blue light scanner set on the baseline calibration ruler can generate grid calibration lines for accurately calibrating the position and angle relationship between the first camera and the second camera. This helps to improve the accuracy and precision of binocular visual imaging, making the detection of pipe segment assembly quality more reliable and accurate. At the same time, the use of the blue light scanner can also provide important reference information for subsequent image processing and analysis, which helps to better evaluate the assembly quality of the pipe segment and discover potential problems.

[0019] Optionally, a bearing platform is provided on the arc-shaped beam.

[0020] By adopting the above technical solution, the bearing platform set on the arc beam can provide a stable docking and starting position for the walking trolley. When the walking trolley completes a detection task, it can be docked on the bearing platform to facilitate data transmission, equipment maintenance or waiting for the next detection task. In addition, the bearing platform can also play a role of buffering and positioning to ensure the safety and accuracy of the walking trolley during movement. In summary, the setting of the bearing platform helps to improve the operating stability of the entire detection device.

[0021] The second purpose of this application is to provide a method for detecting the quality of segment assembly during shield construction based on binocular vision.

[0022] The above technical objectives of the present application are achieved through the following technical solutions: a method for detecting the quality of segment assembly during shield construction based on binocular vision, using a device for detecting the quality of segment assembly during shield construction based on binocular vision as described in the first objective, comprising: S1: Start the segment assembly quality inspection device during shield construction based on binocular vision, turn on the blue light scanner, generate grid calibration lines, and adjust the center distance between the first camera and the second camera; S2: After a ring of assembled segments is dragged out of the shield tail, the walking trolley drives the binocular vision imaging component to automatically move along the arc beam from the starting position of the arc beam; S3: At the first shooting point, a first camera and a second camera are used to collect segment photos at the segment annular seam; S4: The walking car moves at a constant speed, and shoots the next point at a preset interval until the end position of the arc beam, completing a forward shooting task; S5: Taking the end position of the arc beam as the starting point, the trolley drives the binocular vision imaging component to automatically move in the reverse direction of the arc beam and take photos of the segments again. The trolley moves at a constant speed and takes photos of the next point at a preset interval until the trolley returns to the starting position of the arc beam, completing a reverse shooting task. The segments dragged out of the shield tail are photographed for assembly quality, and a sufficient number of segment assembly quality image data are obtained.

[0023] By adopting the above technical scheme, the shield construction period segment assembly quality detection device based on binocular vision can accurately photograph and collect data of the segments. After starting the device, the grid calibration lines generated by the blue light scanner and the adjustment of the camera center distance help to improve the precision and accuracy of the shooting. The walking trolley drives the binocular vision imaging component to move from the starting position of the arc beam, and collects photos of the segments at each shooting point to ensure full coverage of the segments. Through uniform motion and shooting at preset time intervals, as well as one forward or one reverse shooting task, a sufficient amount of high-quality image data can be obtained, providing a reliable basis for subsequent segment assembly quality analysis and evaluation. This detection method is efficient and accurate, and can effectively improve the detection level of shield tunnel segment assembly quality.

[0024] Optionally, the method further includes: S6: Collect segment photos at the segment annular joints through binocular vision imaging components to establish segment annular joint image data sets; S7: performing a preprocessing operation on the established segment annular seam image data set; S8: 3D reconstruction of the inner surface of the shield tunnel lining is performed using a binocular vision stereo algorithm that matches the lining molding quality during the shield tunnel construction period; S9: Preliminary evaluation of the shield tunnel lining inner surface forming quality indicators based on 3D reconstruction data; S10: using a deep learning algorithm to conduct an in-depth analysis of the lining quality data to obtain a test result, wherein the test result includes a segment damage area, crack width, length, and water leakage area; S11: Transmit the detection results to the host computer and the remote monitoring center.

[0025] By adopting the above technical solutions, the inspection process of shield tunnel segment assembly quality is further improved. The binocular visual imaging component is used to collect and establish the segment annular seam image data set, and the preprocessing operation is performed on it, which can improve the quality and availability of the data. The binocular visual stereo algorithm is used to reconstruct the inner surface of the shield tunnel segment in three dimensions, which provides a more intuitive and accurate basis for subsequent quality evaluation. The initial evaluation of the inner surface forming quality index of the shield tunnel segment based on the three-dimensional reconstruction data can quickly understand the basic quality of the segment. The deep learning algorithm is used to conduct in-depth analysis of the lining quality data, which can obtain more detailed inspection results, including key information such as the segment damage area, crack width, length and water leakage area. Finally, the inspection results are transmitted to the host computer and the remote monitoring center, which is convenient for relevant construction technicians to understand the assembly quality of the segment in a timely manner, so as to make corresponding decisions and adjustments, thereby effectively improving the inspection efficiency and accuracy of the assembly quality of the shield tunnel segment and ensuring the quality of the shield tunnel construction.

[0026] In summary, this application has at least the following beneficial effects: 1. Through binocular vision imaging device and shadowless light source device, it can provide clear images of the inner surface of the tunnel lining in the environment of insufficient light inside the tunnel, improving the reliability and quality of image data acquisition; 2. The combination of the automatic travel device and the main structure device enables the binocular visual imaging device to move accurately and efficiently on the inner surface of the shield tunnel lining, solving the problems of slow speed, difficult positioning and incompleteness in manual inspection; 3. The use of three-dimensional reconstruction based on binocular vision stereo algorithm can accurately evaluate the assembly quality of the shield tunnel segments during construction and improve the detection efficiency, solving the problems of insufficient accuracy and low efficiency in the detection of segment assembly quality during construction.

[0027] 4. The segment assembly quality inspection data obtained during the shield tunnel construction period can provide a reliable implementation basis for the shield tunneling operation parameter control and assembly operation of the next ring of segments, thereby improving the segment assembly quality during the shield tunnel lining construction period. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the structure of a device for detecting the quality of segment assembly during shield construction based on binocular vision; Figure 2 It is a status diagram of the segment assembly quality detection device during shield construction based on binocular vision; Figure 3 It is a structural diagram of the main structural components; Figure 4 It is a schematic diagram of the structure of the automatic travel component; Figure 5 It is a cross-sectional view of the automatic travel assembly; Figure 6 It is a schematic diagram of the structure of binocular vision imaging components; Figure 7 It is a flow chart of the steps S1-S5 of the method for detecting the quality of segment assembly during shield construction based on binocular vision; Figure 8 It is a flow chart of the steps S6-S11 of the method for detecting the quality of segment assembly during shield construction based on binocular vision; Fig. 9 It is a schematic diagram of the image transformation coordinate system; Fig.10 It is a schematic diagram of the conversion between the world coordinate system and the camera coordinate system; Fig.11 It is a schematic diagram of the perspective projection principle.

[0029] Reference numerals 1. Main structure assembly; 101. Arc beam; 102. Moving track; 103. Chain; 104. Tensioner; 105. Carrying platform; 2. Automatic traveling assembly; 201. Traveling trolley; 202. First traveling wheel; 203. Second traveling wheel; 204. Driving member; 205. Sprocket; 206. Mounting member; 3. Binocular vision imaging assembly; 31. Baseline calibration ruler; 32. First mounting seat; 33. Second mounting seat; 34. First camera; 35. Second camera; 36. Shadowless light source; 37. Blue light scanner; 4. Installation gap; 5. Installation slot; 6. Slide slot. DETAILED DESCRIPTION

[0030] The present application is further described in detail below in conjunction with the accompanying drawings.

[0031] In this embodiment, refer to Figure 1-6 The device for detecting the quality of segment assembly during shield construction period based on binocular vision comprises a main structure component 1, an automatic running component 2 and a binocular vision imaging component 3. The main structure component 1 is provided with the automatic running component 2, and the main structure component 1 provides guidance and positioning for the movement of the automatic running component 2, so that the automatic running component 2 moves along the track set by the main structure component 1; the automatic running component 2 is provided with the binocular vision imaging component 3, which realizes the effect of intelligent detection of the quality of segment assembly during shield tunnel construction period and solves the problem of low detection efficiency.

[0032] In this embodiment, refer to Figure 3The main structure component 1 includes an arc-shaped beam 101, and the arc-shaped beam 101 is arranged at the connection bridge between the shield tail of the shield machine and the rear supporting trolley. The arc-shaped beam 101 is made of high-strength materials, such as high-strength steel or aluminum alloy, has excellent mechanical properties, and can withstand large mechanical stress. The mounting groove 5 is arranged on the arc-shaped beam 101, which is used to fix the chain 103 to ensure the smooth operation of the automatic running component 2. The chain 103 is arranged in the mounting groove 5 and is made of high-strength materials to ensure stability during long-term high-intensity operation. The tensioning member 104 is used to fix the two ends of the chain 103 to prevent the chain 103 from loosening, so that the chain 103 is always arranged in close contact with the arc-shaped beam 101. Preferably, a winding device can be connected to both ends of the chain 103 to adjust the stability of the chain 103. A bearing platform 105 is provided on the arc beam 101, and preferably, a bearing platform 105 is provided at both ends of the arc, and the bearing platform 105 can provide a stable docking and starting position for the walking trolley 201. After the walking trolley 201 completes a detection task, it can be docked on the bearing platform 105, which is convenient for data transmission, equipment maintenance or waiting for the start of the next detection task. In addition, the bearing platform 105 can also play a role of buffering and positioning, ensuring the safety and accuracy of the walking trolley 201 during movement. The setting of the bearing platform 105 helps to improve the operating efficiency and stability of the entire detection device.

[0033] In this embodiment, refer to Figure 4-5 The automatic travel component 2 includes a walking trolley 201, and the walking trolley 201 is provided with a first walking wheel 202 and a second walking wheel 203, and preferably, a plurality of the first walking wheel 202 and the second walking wheel 203 are provided respectively. An installation gap 4 is formed between the first walking wheel 202 and the second walking wheel 203, and the walking trolley 201 is installed and engaged on the moving track 102 through the installation gap 4, and the first walking wheel 202 and the second walking wheel 203 clamp the moving track 102. The first walking wheel 202 and the second walking wheel 203 on the walking trolley 201 both include a wheel body and a bearing. The wheel body material is usually made of polyurethane, which has good wear resistance and pressure resistance; the bearing is made of high-precision ball bearings to ensure good rolling performance. For example, a wheel body made of polyurethane can be affixed with a wear-resistant layer to extend the service life, or a wheel body made of carbon steel can be used to improve the load-bearing capacity. In addition, the stability of the walking trolley 201 can be further improved by increasing the number of bearings. For example, the first running wheel 202 and the second running wheel 203 may adopt a double-row ball bearing structure to improve strength and ensure stability under long-term high-load working conditions.

[0034] The automatic travel assembly 2 further includes a driving member 204 disposed on the traveling trolley 201, preferably the driving member 204 is a motor, a sprocket 205 is connected to the driving member 204, the sprocket 205 is meshed with the chain 103, and the sprocket 205 is driven to rotate by the driving member 204 to ensure the stable movement of the traveling trolley 201 on the arc beam 101. For example, the chain 103 may be made of stainless steel to enhance corrosion resistance and strength, or the sprocket 205 may be made of high-strength iron to enhance the carrying capacity, thereby further improving the stability and durability of the overall device.

[0035] In this embodiment, refer to Figure 6 The binocular vision imaging assembly 3 includes a baseline calibration ruler 31 arranged on the walking trolley 201, and a slide groove 6 is arranged inwardly on the baseline calibration ruler 31. A first mounting seat 32 and a second mounting seat 33 are arranged in the slide groove 6. The first mounting seat 32 and the second mounting seat 33 can slide in the slide groove 6. A first camera 34 is arranged on the first mounting seat 32, and a second camera 35 is arranged on the second mounting seat 33. The first camera 34 and the second camera 35 preferably use high-performance sensors with the characteristics of high precision and high speed. The shadowless light source 36 device uses an LED light source or other types of light sources to emit uniform and shadowless light to ensure the clarity of the image. Specifically, the shadowless light source 36 can use a ring-shaped shadowless light source 36 or a rectangular shadowless light source 36, which can be selected according to the specific application scenario. The first mounting seat 32 and the second mounting seat 33 slide in the slide groove 6, and the slide groove 6 is made of high-strength material, and wear-resistant layers are installed on both sides to ensure stability during translation. The mounting seat can be fixed by adjusting bolts to ensure adjustability in position. The automatic running component 2 also includes a device for controlling the overall system. For example, the device can use a high-performance motor and controller to ensure real-time control of speed and stability. The data communication module uses wireless or wired communication technology to ensure a stable communication connection with the host computer or remote monitoring center. Specifically, a high-performance Wi-Fi module or 4G / 5G communication module can be used to ensure the real-time and stability of data transmission. The central control device is responsible for scheduling the entire system and coordinating the collaborative work between the components.

[0036] In general, the design of the baseline calibration ruler 31 ensures accurate calibration between the first camera 34 and the second camera 35, thereby obtaining high-quality image data, and finally realizing intelligent detection of the assembly quality of the pipe segments during the shield tunnel construction period. A blue light scanner 37 is also provided on the baseline calibration ruler 31. Preferably, the blue light scanner 37 is provided at the center of the baseline calibration ruler 31, between the first camera 34 and the second camera 35. The blue light scanner 37 provided on the baseline calibration ruler 31 can generate a grid calibration line for accurately calibrating the position and angle relationship between the first camera 34 and the second camera 35. This helps to improve the accuracy and precision of binocular visual imaging, making the detection of pipe segment assembly quality more reliable and accurate. At the same time, the use of the blue light scanner 37 can also provide important reference information for subsequent image processing and analysis, which helps to better evaluate the assembly quality of the pipe segments and discover potential problems.

[0037] The implementation principle of this embodiment is as follows: This embodiment realizes the precise alignment and stable operation of the binocular vision imaging component 3 through reasonable structural design and excellent material selection. The main structure component 1 provides reliable guidance and positioning for the automatic running component 2, ensuring the smoothness and stability of the entire detection device during operation. The automatic running component 2 can achieve full coverage detection of the surface of the pipe segment, improving the comprehensiveness and accuracy of the detection. The images collected by the binocular vision imaging component 3, combined with subsequent image processing and three-dimensional reconstruction procedures, can realize efficient and accurate detection of the assembly quality of the pipe segment during the shield tunnel construction period, providing a strong quality assurance for the shield tunnel construction. Compared with the prior art, this embodiment has the advantages of high detection efficiency and high degree of automation.

[0038] In this embodiment, refer to Figure 7 , a method for detecting the quality of segment assembly during shield construction based on binocular vision, using a device for detecting the quality of segment assembly during shield construction based on binocular vision as described in the first object, comprising: S1: Start the shield construction period segment assembly quality inspection device based on binocular vision, turn on the blue light scanner 37, generate a grid calibration line, and adjust the center distance between the first camera 34 and the second camera 35; S2: After a ring of assembled segments is dragged out of the shield tail, the walking trolley 201 drives the binocular vision imaging component 3 to automatically move along the curved beam 101 from the starting position of the curved beam 101; S3: At the first shooting point, the first camera 34 and the second camera 35 are used to collect the segment photos at the segment annular seam; S4: The walking trolley 201 moves at a constant speed, and shoots the next point at a preset interval until the end position of the curved beam 101, completing a forward shooting task; S5: Taking the end position of the curved beam 101 as the starting point, the walking trolley 201 drives the binocular vision imaging component 3 to automatically move in the reverse direction along the curved beam 101, and takes photos of the pipe segments again. The walking trolley 201 moves at a constant speed, and takes photos of the next point at a preset interval until the walking trolley 201 returns to the starting position of the curved beam 101, completing a reverse shooting task, thereby taking assembly quality photos of the pipe segments dragged out of the shield tail, and obtaining a sufficient number of pipe segment assembly quality image data.

[0039] Specifically, during the equipment preparation stage, ensure that all components of the binocular vision-based shield construction period segment assembly quality inspection device are firmly installed and in normal working condition. Check whether the curved beam 101 is firmly set at the shield tail and the rear supporting trolley connection bridge of the shield machine, the moving track 102 is not deformed or damaged, and the chain 103 is installed in the installation groove 5 of the curved beam 101 and the two ends are well fixed by the tensioning member 104. Check the first walking wheel 202 and the second walking wheel 203 on the walking trolley 201 to ensure that the installation gap 4 between them is appropriate, can be stably embedded in the moving track 102, and can clamp the moving track 102 to achieve smooth movement. At the same time, confirm that the transmission components such as the drive member 204 and the sprocket 205 on the walking trolley 201 are connected normally and can drive the walking trolley 201 to move normally. Check the binocular vision imaging assembly 3 to ensure that the first camera 34 and the second camera 35 are firmly and reliably mounted on the first mounting seat 32 and the second mounting seat 33 of the baseline calibration ruler 31, and that the first mounting seat 32 and the second mounting seat 33 can slide smoothly in the slide groove 6 to adjust the position of the camera. Turn on the blue light scanner 37 to check whether it can generate grid calibration lines normally. Confirm that the shadowless light source 36 on the first camera 34 and the second camera 35 can work normally and provide sufficient lighting.

[0040] In the detection startup phase, the shield construction period segment assembly quality detection device based on binocular vision is started. The blue light scanner 37 is turned on to generate clear grid calibration lines so that the position and angle of the camera can be accurately determined later. At the same time, according to the actual detection requirements, the center distance between the first camera 34 and the second camera 35 is carefully adjusted to ensure that the best binocular vision effect can be obtained. Wait for a ring of assembled segments to be dragged out of the shield tail. After the segments are dragged out of the shield tail, the walking trolley 201 starts to drive the binocular vision imaging component 3 to automatically move along the arc beam 101 from the starting position of the arc beam 101. At this time, the walking trolley 201 should run smoothly along the moving track 102, and its speed is controlled by the driving member 204 to ensure uniform motion. In the shooting phase, when the walking trolley 201 reaches the first shooting point, the first camera 34 and the second camera 35 are used to collect segment photos at the segment annular seam. At this time, the camera should be clearly focused, and the shadowless light source 36 provides uniform lighting to ensure that high-quality segment photos are taken. The walking trolley 201 continues to move at a constant speed and shoots the next point at a preset time interval. During the movement, the stability of the walking trolley 201 must be ensured to avoid factors such as vibration affecting the shooting quality. Shoot until the end position of the curved beam 101 to complete a forward shooting task. Taking the end position of the curved beam 101 as the starting point, the walking trolley 201 drives the binocular vision imaging component 3 to automatically move in the reverse direction along the curved beam 101. Take photos of the pipe segment again, also maintaining a constant speed, and shoot the next point at a preset interval. Until the walking trolley 201 returns to the starting position of the curved beam 101, a reverse shooting task is completed. Through a forward or reverse shooting, the pipe segment dragged out of the shield tail is comprehensively shot for assembly quality, thereby obtaining a sufficient number of pipe segment assembly quality image data.

[0041] In this embodiment, refer to Figure 8 , the method further comprises: S6: collecting segment photos at the segment annular joints through the binocular vision imaging component 3 to establish a segment annular joint image data set; S7: performing a preprocessing operation on the established segment annular seam image data set; S8: 3D reconstruction of the inner surface of the shield tunnel lining is performed using a binocular vision stereo algorithm that matches the lining molding quality during the shield tunnel construction period; S9: Preliminary evaluation of the shield tunnel lining inner surface forming quality indicators based on 3D reconstruction data; S10: using a deep learning algorithm to conduct an in-depth analysis of the lining quality data to obtain a test result, wherein the test result includes a segment damage area, crack width, length, and water leakage area; S11: Transmitting the detection results to the host computer and the remote monitoring center.

[0042] Specifically, ensure that the first camera 34 and the second camera 35 in the binocular visual imaging component 3 are in good working condition, turn on the blue light scanner 37 and confirm that the grid calibration lines generated by it are clearly visible. Check whether the movement of the walking trolley 201 on the curved beam 101 is smooth so that it can accurately reach each shooting point. When a ring of assembled segments is dragged out of the shield tail, the walking trolley 201 drives the binocular visual imaging component 3 to move from the starting position of the curved beam 101. At the segment annular joint, the first camera 34 and the second camera 35 are used to collect photos according to preset parameters. Different shooting angles and distances can be set to ensure that all parts of the segment annular joint can be fully covered. During the collection process, the clarity and resolution of the photos must be guaranteed for subsequent analysis and processing. The collected segment annular joint photos are sorted and classified to establish a segment annular joint image data set. They can be classified according to different segment types, construction stages or quality characteristics to better manage and use the data set.

[0043] The segment annular joint image data set is preprocessed, and a suitable image denoising algorithm is used to remove noise interference in the segment annular joint photos. For example, median filtering, Gaussian filtering and other methods can be used to reduce random noise in the image and improve the image quality. The features of the segment annular joint are made more obvious by adjusting the image contrast. Histogram equalization, contrast stretching and other methods can be used to enhance the image contrast and improve the recognizability of the segment annular joint. The segment annular joint photos are cropped to remove unnecessary background information and only retain the segment annular joint part. At the same time, the cropped images are normalized to unify the image size, resolution and other parameters to facilitate subsequent algorithm processing. The preprocessed segment annular joint images are annotated to mark the quality problems such as segment damage, cracks, and annular joint misalignment. Annotation can be done manually or automatically to provide training data for the deep learning algorithm.

[0044] To reconstruct the inner surface of the shield tunnel lining in three dimensions, select a binocular vision stereo algorithm that matches the lining molding quality during the shield tunnel construction period, and set the algorithm parameters according to the actual situation. For example, the camera's internal and external parameters can be adjusted, and the parameters of the stereo matching can be set to ensure the accuracy and precision of the three-dimensional reconstruction. The pipe segment photos collected by the binocular vision imaging component 3 are used for image matching and depth calculation. By finding the corresponding points in the left and right images, the depth information of each pixel is calculated to construct the three-dimensional point cloud data of the pipe segment. The three-dimensional point cloud data is surface reconstructed to generate a three-dimensional model of the inner surface of the shield tunnel lining. Methods such as triangulation and surface fitting can be used to convert the point cloud data into a continuous surface model so that the quality of the pipe segment can be analyzed more intuitively. It is preferred to use the beam adjustment based on OpenCV to complete the three-dimensional reconstruction and splicing of the inner surface of a ring of shield tunnel segments to obtain a complete three-dimensional reconstruction image of the inner surface of the shield tunnel segment.

[0045] Initially evaluate the inner surface forming quality indicators of the shield tunnel lining, and select appropriate inner surface forming quality indicators of the shield tunnel lining according to actual needs. For example, you can select indicators such as the segment, the amount of misalignment of the annular joint, the amount of opening, the area and number of surface defects. Clarify the definition and calculation method of each indicator for accurate evaluation. Use the segment model obtained by three-dimensional reconstruction to analyze each quality indicator. For example, you can evaluate the ellipticity by calculating the change in curvature of the inner surface of the segment, evaluate the segment alignment by measuring the misalignment and opening of the annular joint, and evaluate the area and number of surface defects by detecting the defective area on the surface. Output the results of the initial evaluation as a visual chart or report so that relevant personnel can understand the forming quality of the inner surface of the shield tunnel lining. Color coding, charts, etc. can be used to intuitively display the distribution and changes of various quality indicators.

[0046] In-depth analysis of lining quality data, select deep learning algorithms suitable for segment assembly quality detection, such as convolutional neural network (CNN), recurrent neural network (RNN), etc. Use the preprocessed segment annular seam image data set to train the deep learning algorithm, adjust the algorithm parameters, and improve the accuracy and computational efficiency of the algorithm. Apply the trained deep learning algorithm to new segment photos to extract segment assembly quality data such as damaged area, crack width, length, annular seam misalignment, opening, and water leakage area. Accurate assembly quality data can be obtained by segmenting, detecting, and measuring images. Analyze and verify the quality data extracted by the deep learning algorithm. It can be compared with the results of manual detection to evaluate the accuracy and reliability of the algorithm. At the same time, statistical analysis of the results is performed to understand the distribution and changes of segment assembly quality, providing a reference for subsequent segment assembly quality control.

[0047] Transmit the test results, select the appropriate data transmission method, and transmit the test results to the host computer and remote monitoring center. Transmission can be carried out using wired networks, wireless networks or data storage devices. Ensure the stability and security of data transmission to prevent data loss or tampering. Display and analyze the transmitted test results on the host computer and remote monitoring center. A visual interface can be used to display the quality data and three-dimensional model of the pipe segment, which is convenient for relevant personnel to view and analyze. At the same time, the test results can be statistically analyzed, and reports and charts can be generated to provide a basis for pipe segment assembly quality control and decision-making. Integrate the above modules into a system, optimize the interfaces and data flows between modules, and ensure the efficient operation of the system; conduct system tests in the actual tunnel construction environment to verify the performance and reliability of the system; adjust and optimize system parameters according to the test results to improve the accuracy and stability of the system.

[0048] Specifically, on this basis, the following nine steps are used to realize the intelligent detection of the assembly quality of shield tunnel segments during the construction period: Step 1: Stereo camera calibration Binocular correction is to eliminate distortion and align the left and right images respectively after the internal parameters (focal length, imaging origin, radial distortion coefficient, etc.) and the position relationship of the two cameras (relative rotation matrix and relative translation vector) obtained by camera calibration, so that the imaging origin coordinates of the two images are consistent, the optical axes of the two cameras are parallel, the two imaging planes are coplanar, and the epipolar lines are aligned. It includes the following processes: 1. Design and build a binocular vision experimental platform; 2. Change the position of the chessboard and take multiple pairs of photos of the chessboard calibration projection; 3. Detect and extract corner points in the chessboard; 4. Calculate the internal and external parameters of the camera without distortion, and use maximum likelihood estimation to improve accuracy; 5. Use the least squares method to find the actual radial distortion coefficient; 6. Comprehensive internal and external parameters and radial distortion coefficients are optimized using the maximum likelihood method to improve calibration accuracy; 7. Obtain the internal and external parameters and distortion coefficient of the camera.

[0049] Step 2: Start the system and capture images 1. Choose a suitable location inside the tunnel and install two high-resolution, high-frequency cameras. The cameras should have autofocus and optical image stabilization functions to adapt to the light conditions and possible vibrations in the tunnel.

[0050] 2. The camera installation angle must be accurately calculated to ensure that the stereoscopic vision covers the entire tunnel lining surface.

[0051] 3. Configure the image acquisition module to ensure that at least two cameras can synchronously acquire images, and the synchronization accuracy must reach the millisecond level.

[0052] 4. Start the image acquisition module and synchronously trigger the binocular camera to acquire images.

[0053] 5. Preprocess the collected original images, including denoising, enhancement, correction and other operations, to improve image quality and meet the needs of subsequent processing.

[0054] Step 3: Coordinate system transformation Rigid Body Transformation: Objects in three-dimensional space are represented by the world coordinate system, and captured images are represented by the image coordinate system. The world coordinates of objects can be transformed into camera coordinates through rigid body transformation. Fig. 9 , rigid body transformation refers to the rotation and translation of geometric objects in three-dimensional space without any deformation. Fig.10 , let X c , Y c , Z c Represents the camera coordinate system, X w , Y w , Zw represents the world coordinate system, R represents the rotation matrix, T represents the translation vector, and the coordinates of point P in the camera coordinate system are (X c , Y c , Z c ), the rigid body transformation relationship is: in, Where, T x , T y , T z —Translation of the object being measured; R x , R y , R z —The rotation matrix of the object on the X, Y, and Z axes; θ x ,θ y ,θ z —The rotation angle of the object on the X, Y, and Z axes.

[0055] Perspective Projection: The world coordinates are converted to camera coordinates after rigid body transformation. The objects under the camera coordinates need to be transformed into pixels on the image through perspective projection. The principle of perspective projection is as follows Fig.11 As shown, the three-dimensional point Q(X c ,Y c ,Z c ) is transformed into a point q(x,y) in the image coordinate system through perspective projection.

[0056] exist Fig.11 The optical center of the camera is Q c , the optical axis of the camera is Z c , the focal length of the camera is f. The focal length is a fixed parameter set at the factory and is one of the internal parameters of the camera.

[0057] According to the principle of triangular similarity: In the formula, z c —Scale factor.

[0058] According to the triangle similarity principle, q(x,y) on the image and the camera coordinates Q(X c ,Y c ,Z c ) is as follows: The transformation relationship between image coordinates and camera coordinates is expressed in matrix form, as shown in formula (9): The point Q(X,Y,Z) in the three-dimensional space world coordinate system is converted to point q(x,y) through rigid body transformation and perspective projection, and the point q(x,y) is discretized and represented by pixel coordinates q(u,v).

[0059] The image coordinate system and the pixel coordinate system are converted using the following formula (10): Where, d x ,d y —width in pixels; u, v—number of rows and columns of the pixel matrix; u0, v0—pixel coordinates of the camera optical center.

[0060] By combining equation (9) and equation (10), a point can be converted from the world coordinate system to the pixel coordinate system, as shown in equation (11): Among them, R—rotation matrix; T—Translation vector.

[0061] Using the principle of triangulation, the object (X c ,Y c ,Z c )’s coordinates: Where, d is the difference in position of the same point in the left and right images, i.e., disparity; u0, v0—pixel coordinates of the optical center; u1, v1—pixel coordinates of the same point on the left image; u2, v2—pixel coordinates of the same point on the right image; b—horizontal distance between the optical centers of the cameras; z c —The scale factor is the inverse of the camera pixel unit size Step 4: Image Correction Image distortion in binocular vision mainly includes lens non-coplanarity and distortion, which require the use of computers for image correction.

[0062] Distortion correction: During the camera imaging process, the image will be distorted due to the manufacturing error of the optical lens and the assembly process. This distortion includes radial distortion and tangential distortion. The radial distortion has a greater impact on the measurement results of binocular vision. Radial distortion mainly includes pincushion distortion and barrel distortion.

[0063] Assume that p(u,v) is the pixel coordinate of a point in the undistorted case, p(u d ,v d ) is the pixel coordinate of the point after distortion, (u0,v0) is the pixel coordinate of the camera optical center in the image coordinate system, (u d ,v d ) is the continuous image coordinate after distortion, the distortion coefficients are k1 and k2, (x, y) is the continuous image coordinate without distortion, and k1 and k2 are used to describe the pixel coordinate after distortion: Use camera calibration to solve (u0,v0), (x,y), (u,v), (u d ,v d ): Stereo correction: For binocular vision measurement, it is necessary to know the relative position relationship of the two cameras in order to calculate the distance of the object through parallax. Stereo correction is to correct two non-parallel and non-coplanar cameras to be parallel and coplanar, which is convenient for subsequent measurement and calculation.

[0064] During calibration, two cameras are used to simultaneously shoot the chessboard, and the internal and external parameters of the cameras, the relative rotation relationship between the cameras, and the relative translation relationship are calculated: In the formula, R l , R r —The rotation matrix of the left and right cameras for the same object; T l , T r —The translation matrix of the left and right cameras for the same object; R0, T0—relative rotation matrix and relative translation matrix.

[0065] After obtaining the relative position relationship of the two cameras, it is necessary to correct the imaging planes of the two cameras to be parallel and coplanar. The relative rotation matrix between the left and right cameras is decomposed into the composite rotation matrix r of the left and right cameras. l 、r r After obtaining the synthetic rotation matrix, calculate the rotation matrix R of the left and right images lt , R rt ,in: Step 5: Stereo matching Stereo matching is a key step in binocular vision 3D reconstruction. It is a process of obtaining left and right images by shooting the same scene with two calibrated cameras, calculating the corresponding points and disparity in the left and right images, and finally obtaining the depth information. The global stereo matching algorithm first finds the feature points or feature lines in the image for matching to obtain the initial disparity value, then uses the global energy function for constraints, continuously iterates and optimizes the disparity value, and finally obtains a continuous disparity map. Global stereo matching algorithms include dynamic programming algorithms, graph cut methods, belief propagation algorithms, etc., with high calculation accuracy.

[0066] Step 6: 3D reconstruction Using the disparity map and perspective transformation matrix Q obtained by stereo matching, a mapping map of the same size as the disparity map is obtained through formula (26). Each pixel in the map has three channels that store the values ​​of the pixel point on the x, y, and z axes in the camera coordinate system, that is, the three-dimensional coordinates (X / W, Y / W, Z / W) of each pixel in the camera coordinate system. Among them, the perspective transformation matrix Q is as shown in formulas (27) and (28). After obtaining the camera coordinates of each pixel point, the three-dimensional point cloud information of the inner surface of the measured segment is generated, and the three-dimensional model of the inner wall of the shield tunnel can be reconstructed using the three-dimensional point cloud information.

[0067] Where, cx, cy—coordinates of the left camera pixel in the image; cx'—coordinates of the right camera pixel in the image; f—camera focal length; Tx—distance of the binocular camera optical center.

[0068] Step 7: Calculate relevant parameters Using the above three-dimensional point cloud data, the slope segmentation optimization algorithm is used to calculate the segment gap misalignment D. d , Annular gap opening D o .

[0069] Circumferential seam misalignment D d Calculation: Suppose the coordinates of a certain point cloud after segmentation are p. After the initial data is processed by coordinate transformation, the central axis of the tunnel coincides with the Z axis. Therefore, for any point p in p(x,y), i Distance d to the origin of the XY plane i : According to the above formula, the maximum distance, i.e. the data corresponding to the peak value, can be obtained, and the z value at this time can be extracted: Where, d maxis the maximum distance from p to the origin of the XY plane, and z is the coordinate value of the corresponding point cloud on the Z axis.

[0070] Extract the maximum value z in the point cloud p max , minimum value z min , among which there must be z max >z>z min , and create a collection O: Extract the data in the set O in pairs to segment the point cloud p on the Z axis. The final data is segmented into three segments. The front and back segments are the point clouds A and B. Then, in the XY plane, the average distance d from the point cloud data of points A and B to the origin is: Where, d A d B are the average distances of parts A and B from the origin in the XY plane, and the misalignment value is d A d B The absolute value of the difference: Annular gap opening D o Calculation: Suppose the coordinates of a certain point cloud after segmentation are p(x, y, z). After the initial data is processed by coordinate projection transformation, the central axis of the tunnel coincides with the Z axis. Therefore, for any point p i To the coordinate on the Z axis is z i .

[0071] It can be seen that for any two points on the Z axis, the corresponding coordinate data of the two points can be obtained, that is, the annular gap opening D at this time o value: In the formula, z i and z j are the coordinate values ​​of any two point clouds on the Z axis.

[0072] Step 8: Calculate relevant parameters The above Mask R-CNN instance segmentation algorithm is used to calculate the segment damage area S s And leakage area S L , crack detection algorithm crack width L W , crack length L L .

[0073] Step 9: System Integration and Testing 1. Through the data transmission module, the test results and analysis reports are wirelessly transmitted to the remote monitoring center or display device.

[0074] 2. Develop a user interface that provides functions such as real-time display of test results, historical data analysis, and report generation. The user interface should have a user-friendly interface that enables construction personnel to easily monitor and analyze the quality of the tunnel lining.

[0075] 3. Integrate the above modules into a system, optimize the interfaces and data flows between modules, and ensure efficient operation of the system.

[0076] 4. Conduct system tests in actual tunnel construction environment to verify system performance and reliability.

[0077] 5. According to the test results, adjust and optimize system parameters to improve the accuracy and stability of the system.

[0078] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A device for detecting the quality of segment assembly during shield construction based on binocular vision, characterized in that: The invention comprises a main structure component (1), an automatic running component (2) and a binocular vision imaging component (3); the main structure component (1) is provided with the automatic running component (2); the main structure component (1) provides guidance and positioning for the movement of the automatic running component (2), so that the automatic running component (2) moves along a track set by the main structure component (1); the binocular vision imaging component (3) is provided on the automatic running component (2).

2. The device for detecting the quality of shield segment assembly during construction based on binocular vision according to claim 1 is characterized in that: The main structure component (1) comprises an arc-shaped beam (101), the arc-shaped beam (101) is arranged at the connection bridge between the shield tail of the shield machine and the rear supporting trolley, and a movable track (102) is arranged on the arc-shaped beam (101); the automatic running component (2) comprises a walking trolley (201), the walking trolley (201) is provided with a first walking wheel (202) and a second walking wheel (203), the first walking wheel (202) and the second walking wheel (203) are arranged to form an installation gap (4), the walking trolley (201) is installed and engaged on the movable track (102) through the installation gap (4), and the first walking wheel (202) cooperates with the second walking wheel (203) to clamp the movable track (102).

3. The device for detecting the quality of shield segment assembly during construction based on binocular vision according to claim 2 is characterized in that: The arc-shaped beam (101) is provided with a mounting groove (5), a chain (103) is arranged in the mounting groove (5), and a tensioning member (104) is provided on the arc-shaped beam (101) for fixing the two ends of the chain (103); the walking trolley (201) is provided with a driving member (204), the driving member (204) is connected to a sprocket (205) in a transmission manner, the sprocket (205) is meshed with the chain (103), and the sprocket (205) is driven to rotate by the driving member (204) to drive the walking trolley (201) to move.

4. The device for detecting the quality of segment assembly during shield construction based on binocular vision according to claim 2 is characterized in that: The walking trolley (201) is provided with a mounting member (206), and the binocular vision imaging assembly (3) comprises a baseline calibration ruler (31) arranged on the mounting member (206), a slide groove (6) is arranged inwardly on the baseline calibration ruler (31), a first mounting seat (32) and a second mounting seat (33) are arranged in the slide groove (6), the first mounting seat (32) and the second mounting seat (33) can both slide along the slide groove (6), a first camera (34) is arranged on the first mounting seat (32), and a second camera (35) is arranged on the second mounting seat (33).

5. The device for detecting the quality of segment assembly during shield construction based on binocular vision according to claim 4 is characterized in that: The first camera (34) and the second camera (35) are both equipped with a shadowless light source (36).

6. The device for detecting the quality of segment assembly during shield construction based on binocular vision according to claim 5 is characterized in that: The baseline calibration ruler (31) is provided with a blue light scanner (37).

7. The device for detecting the quality of segment assembly during shield construction based on binocular vision according to claim 2 is characterized in that: A bearing platform (105) is provided on the arc-shaped beam (101).

8. A method for detecting the quality of segment assembly during shield construction based on binocular vision, using the device for detecting the quality of segment assembly during shield construction based on binocular vision as described in claims 1 to 7 above, characterized in that: include: S1: starting a binocular vision-based shield construction period segment assembly quality detection device, turning on a blue light scanner (37), generating a grid calibration line, and adjusting the center distance between a first camera (34) and a second camera (35); S2: After a ring of assembled segments is dragged out of the shield tail, the walking trolley (201) drives the binocular vision imaging component (3) to automatically move from the starting position of the arc beam (101) along the arc beam (101); S3: At the first shooting point, a first camera (34) and a second camera (35) are used to collect a segment photo at the segment annular seam; S4: The walking trolley (201) moves at a constant speed, photographing the next point at a preset time interval until the end position of the arc beam (101) is reached, completing a forward photographing task; S5: With the end position of the curved beam (101) as the starting point, the walking trolley (201) drives the binocular vision imaging component (3) to automatically move in the reverse direction along the curved beam (101), and takes photos of the pipe segments again. The walking trolley (201) moves at a constant speed, and takes photos of the next point at a preset interval until the walking trolley (201) returns to the starting position of the curved beam (101), completing a reverse shooting task, thereby taking photos of the assembly quality of the pipe segments dragged out of the shield tail, and obtaining a sufficient number of pipe segment assembly quality image data.

9. The method for detecting the quality of segment assembly during shield construction based on binocular vision according to claim 8, characterized in that: The method further comprises: S6: collecting segment photos at the segment annular joints through the binocular vision imaging component (3) to establish a segment annular joint image data set; S7: performing a preprocessing operation on the established segment annular seam image data set; S8: The inner surface of the shield tunnel segment is reconstructed in 3D by using a binocular vision stereo algorithm to match the segment assembly quality during the shield tunnel construction period; S9: Preliminary evaluation of the inner surface forming quality indicators of shield tunnel segments based on 3D reconstruction data; S10: using a deep learning algorithm to perform in-depth analysis on the segment assembly quality data to obtain a test result, wherein the test result includes a segment damage area, crack width, length, and water leakage area; S11: Transmit the detection results to the host computer and the remote monitoring center.

Images