Abstract
Long-distance tunnels are essential hydraulic structures that guarantee the safe operation and optimal performance of engineering projects. Monitoring the operational status of tunnels through regular inspections is crucial to promptly detect defects such as seepage, collapses, and damages. This proactive approach prevents structural instability, damages, and blockages, ensuring the safe and stable operation of engineering projects. Water conveyance tunnels are typically long-distance, enclosed spaces with no illumination and no satellite navigation signals. Traditional manual inspection methods suffer from low efficiency, poor accuracy, and high safety risks. The tunnel defect detection method based on image automatic collection and 3D reconstruction addresses these challenges. It enables rapid inspections during flood periods and comprehensive examinations during dry seasons, adapting to the tunnel's water conditions. This method facilitates unmanned intelligent inspections of long-distance tunnels, three-dimensional visualization, and multi-period comparative analysis of defects, offering vast prospects for widespread application. Finally, the proposed method and device were experimentally applied in a practical engineering drainage tunnel to demonstrate their effectiveness and reliability.
You have full access to this open access chapter, Download chapter PDF
Keywords
1 Introduction
In recent years, China has undertaken the construction of numerous long-distance water conveyance and drainage tunnels, including significant projects such as the South-to-North Water Diversion Project, the Yunnan Central Diversion Project, the Sichuan Yalong River Jinping II Hydropower Station, and various diversion and drainage tunnels for high dams and reservoirs. The operational condition of these tunnels directly influences the safe and stable operation of crucial engineering projects, impacting areas such as people’s livelihoods, water supply, flood control, and electricity generation. Consequently, conducting regular inspections is vital to understanding their actual operational status, promptly detecting potential issues such as collapses, damages, blockages, cracks, and seepages, and preventing structural instability, damages, and blockages that could arise from undetected or worsening defects. Ensuring the safety of engineering projects is paramount [1, 2].
Traditional methods of tunnel defect detection primarily rely on manual inspection, combining human efforts with various instruments to inspect defects such as cracks, falling blocks, and seepage. Long-distance water conveyance and drainage tunnels, being enclosed spaces, lack satellite navigation signals, illumination, and have poor conditions for human entry and inspection. This results in low efficiency during manual inspections, inaccurate positioning of defects like cracks, collapses, and damages, and difficulties in timely identification of significant hidden defects [3, 4]. Additionally, the tunnel floor may have sedimentation, pits, steel bars, tree branches, and other obstacles. Inadequate lighting, unclear inspection conditions, and the possibility of falling rocks and blocks pose safety risks during manual inspections.
During manual inspections, various equipment can be employed, such as ultrasonic testing, acoustic emission testing, radar detection, and fiber optic sensing, in addition to visual observation, to detect tunnel defects [5, 6]. Acoustic emission testing utilizes the phenomenon of material acoustic emissions to detect defects in tunnel lining concrete such as cracks. Radar detection determines the existence of cracks on the tunnel lining surface by analyzing the difference in reflection wave strength when radar waves pass through different mediums. Fiber optic sensing detects cracks in tunnel lining concrete by identifying changes in optical waveguides. However, these crack detection methods mentioned above face challenges such as low detection efficiency, poor detection results, difficulties in accurately and efficiently locating crack positions and their orientations within a limited time frame, and the complexity of simultaneously detecting various defects like seepage and damages [7, 8].
In addition to traditional methods, new equipment and technologies have been introduced for tunnel lining concrete crack detection, such as laser scanning detection technology and digital image processing technology. Laser scanning technology primarily utilizes laser scanners to obtain point cloud data of the tunnel lining, determining the presence of cracks based on this point cloud. Alternatively, it constructs images of the tunnel lining based on the reflectivity data of scanned points, further extracting crack information [9, 10]. While laser scanning technology can achieve the goal of crack detection, its precision is low when detecting small cracks. Additionally, the operational procedures of laser scanners are complex, and their detection efficiency is low, making it challenging to rapidly detect cracks in tunnel lining concrete within a limited time frame. Digital image processing technology utilizes cameras to capture images of the tunnel lining and then employs deep learning algorithms such as neural networks to identify cracks [11,12,13]. Digital image processing technology replaces human visual recognition with machine vision, allowing for more accurate crack identification and extraction of crack characteristics compared to other methods [14]. However, tunnels are enclosed spaces with long distances and no illumination, and the satellite navigation signals inside the tunnels are weak. As a result, there is a lack of mature and reliable image acquisition equipment that meets the requirements of regular tunnel inspections. Moreover, existing algorithms for identifying and classifying various tunnel defects lack universality, and their accuracy is still insufficient to meet practical inspection requirements. Therefore, this paper proposes a tunnel defect detection method and device based on image automatic collection and 3D reconstruction. This method aims to enhance detection efficiency and accuracy, enabling unmanned intelligent inspections of long-distance underwater tunnels. Experimental applications have been conducted in practical engineering projects to validate its effectiveness and reliability.
2 Technology and Equipment
For water conveyance, diversion, and drainage tunnels, during flood periods, when water flow is swift due to rapid fluctuations in water levels, manual inspection becomes impractical. At this time, there is a need for quick inspections to identify potential issues such as collapses, damages, and blockages within the tunnels. During the dry season, when water depth is shallow and water flow is minimal, more detailed inspections are possible. Based on this approach, image acquisition equipment has been developed. This equipment can be mounted on a mobile platform (e.g., unmanned vehicles, inspection boats, or vehicles) and automatically moves along with the platform, autonomously capturing tunnel wall images and detecting defects. The hardware system of the image acquisition equipment (Fig. 1) consists of four main components: the acquisition system framework, a high-definition single-lens reflex camera for detailed inspections, a fast inspection 360-degree panoramic camera, and an RGB visible light camera combined with an infrared thermal imaging camera.
2.1 Rapid Inspection During Flood Periods
To achieve quick tunnel inspections, an Insta360 Pro 2 panoramic camera is used for rapid image capture. This panoramic camera is equipped with six ultra-wide-angle fisheye lenses, each with a field of view of nearly 180 degrees, enabling 360-degree panoramic photography. It provides a maximum image size of 7680 × 3840 pixels with 8 K resolution. The camera can be mounted on a power-supplied E-port quick-release interface and, depending on the collection requirements and on-site conditions, can be quickly exchanged with a high-precision single-lens reflex camera acquisition device. The extensive coverage of the panoramic camera, coupled with the rapid movement of the platform along the tunnel, enables automatic image capture of the tunnel wall during flood periods. It allows for the rapid identification of significant defects such as extensive cracks, large-scale disintegration, water leakage, and accumulation.
2.2 Comprehensive Inspection During the Dry Season
During the dry season, when water depth is shallow and water flow is minimal, conditions are conducive to slow movement and detailed inspection. For this purpose, a full-frame single-lens reflex camera is used for image capture, with a maximum image size of 7952 × 5304 pixels (8 K) and a resolution of 42 million pixels. Equipped with a 35 mm fixed-focus lens, after sub-pixel processing, it achieves an accuracy of 0.1 mm at a distance of 5 m from the shooting surface. The film resolution is approximately 0.1–0.2 mm per pixel, meeting the accuracy requirements for crack identification of 0.1 mm. It can also identify local small-scale disintegration, leakage, detachment, and other defects, meeting the precision requirements for concrete surface defect detection.
The full-frame single-lens reflex camera is installed on a self-developed 3DOF brushless stabilizing gimbal for physical stabilization to prevent blurring caused by movement platform vibrations during travel. Moreover, the stabilizing gimbal features automatic camera rotation and shooting through program control, as well as real-time communication with the mobile platform.
2.3 Supplementary Inspection with Infrared Thermal Imaging
Water infiltration areas in concrete structures tend to have lower temperatures compared to adjacent areas. Infrared thermal imaging technology is employed to supplement inspections for water leakage points. A thermal imaging camera with a resolution of 40 × 480 is used to identify defects in the tunnel wall due to water infiltration. This is combined with the RGB visible light camera to quickly pinpoint water leakage locations.
2.4 Image Acquisition Process
The displacement statistical regression model utilizes environmental factors as independent variables.
The image overlap rate includes the forward overlap (R1, along the tunnel axis) and the lateral overlap (R2, the rotation angle of the camera within the cross-section). Based on the overlap rate, the lateral photo shooting angles and the forward distances of the mobile platform during different-sized tunnel image acquisition can be calculated. If the camera’s focal length is f, the sensor length is l = the longest side pixel × pixel size, the sensor width is w = the shortest side pixel × pixel size, and the camera distance from the surface to be measured is D.
During detailed image acquisition, the mobile platform moves along the tunnel axis and stops to allow the image acquisition system to control the camera for 360° circular shooting. The intervals for stopping and the camera’s rotation angles for each shot are controlled based on an 80% image overlap requirement. The process of image acquisition along the tunnel axis and the cross-sectional direction is illustrated in Figs. 2 and 3.
As shown in Fig. 4, the axial photo interval N is calculated as follows:
As shown in Fig. 5, the circumferential photo interval M is calculated as follows:
Using the above formulas, the axial interval is calculated as 1.2 m, which means the platform stops and performs 360° circular image acquisition every 1.2 m of forward movement. When the camera captures images at each stop, the lateral pitch angle of the single-lens reflex camera is 19°, meaning the camera triggers the shutter every 19° of rotation.
2.5 Three-Dimensional Reconstruction and Localization of Tunnel Defects
The vast amount of image data obtained after image acquisition needs to be stitched together for subsequent defect localization. Image stitching can be based on two reference indicators: coordinate information and photo texture. When satellite navigation signals are strong, the captured images contain three-dimensional coordinate point information, and image stitching can be performed based on these coordinate points to locate the point cloud information. When satellite navigation signals are weak, the captured images themselves do not contain three-dimensional coordinate information and cannot be positioned using point clouds. In this case, stitching can be done based on the texture characteristics of the images. Ultimately, a complete three-dimensional realistic model of the tunnel wall is obtained, serving as the background dataset for defect localization.
Using the fully covered tunnel wall image acquisition data, the stitched images are used to construct the original three-dimensional realistic model of the tunnel, which contains point cloud information with actual coordinates. This model serves as the base map for three-dimensional defect display and enables image feature extraction and defect localization. The defect images automatically detected are associated with the base map of the tunnel’s three-dimensional model using image retrieval methods. For defect images detected automatically, the position of the defect can be rapidly calculated based on the position of the mobile platform, camera height, and angle. With the mapping relationship between defect images and the three-dimensional realistic model, along with camera parameters and the distance between the camera and the defect area, the actual size of the defect area can be estimated. Consequently, information such as the actual position of the defect within the tunnel, the type of defect, and the defect size can be obtained. This information is annotated and displayed on the three-dimensional realistic model as the result.
3 Case Study
The image acquisition method and equipment proposed in this paper were applied in an experimental inspection of a drainage tunnel using an automatic mobile platform to validate the effectiveness and feasibility of the developed equipment. The tunnel covered a total distance of 3Â km. During on-site testing, the image transmission system was placed on the mobile platform. Data transmission information was received at both the entrance and exit of the tunnel while capturing images along the tunnel axis. This resolved the problem of no satellite navigation signals inside the tunnel, indicating that the image acquisition system is capable of capturing high-definition images and implementing transmission functions within long-distance enclosed tunnel spaces.
The hardware system of the image acquisition equipment was installed on the top of the mobile platform for automatic capture of high-definition images of the tunnel walls. Image stitching and three-dimensional modeling of the tunnel were performed. Point cloud data from the captured portions of the tunnel are shown in Fig. 6, and the three-dimensional realistic model obtained by coupling and stitching based on the coordinate positions of the point cloud is illustrated in Fig. 7.
4 Conclusion
In this paper, a tunnel defect detection method and equipment based on automatic image acquisition and three-dimensional reconstruction were proposed. Considering the tunnel water conditions, a combined inspection approach of rapid checks during flood seasons and comprehensive inspections during dry periods was introduced. The proposed method was experimentally applied in the inspection of a drainage tunnel, enabling quick identification of anomalies such as collapses, damages, and blockages within the tunnel. It enhanced the intelligence level of tunnel inspections, significantly improving inspection efficiency and safety.
-
The hardware system of the tunnel inspection image acquisition equipment was constructed, consisting of the acquisition system framework, detailed full-frame high-definition camera, general survey 360 panoramic camera, and RGB visible light combined with infrared thermal imaging camera. This system enables rapid checks during flood seasons, detailed inspections during dry periods, and supplementary inspections for water seepage points using thermal imaging. It replaced traditional manual inspection methods, greatly enhancing the efficiency, accuracy, and safety of inspections.
-
The tunnel inspection method and equipment proposed in this paper were applied in practical engineering, achieving unmanned automatic inspection of large-scale, long-distance, unilluminated, satellite navigation signal-absent, enclosed drainage tunnels. The system can create three-dimensional models of tunnels, present them in three dimensions, and compare defects over multiple periods. It possesses features of automation, intelligence, and efficiency. The application prospects of this method in the future for large-scale, deeply buried, long-distance unmanned intelligent tunnel inspections are highly promising.
References
Zhu Y, Xie M, Zhang K et al (2010) A dam deformation residual correction method for high arch dams using phase space reconstruction and an optimized long short-term memory network. Mathematics 2023:11
Xu M, Pang R, Zhou Y, et al (2023) Seepage safety evaluation of high earth-rockfill dams considering spatial variability of hydraulic parameters via subset simulation. J Hydrol 626(A):130261
Zhu Y, Zhang Z, Gu C, et al (2023) A coupled model for dam foundation seepage behavior monitoring and forecasting based on variational mode decomposition and improved temporal convolutional network. Struct Control Health Monitor 3879096
Koch C, Georgieva K, Kasireddy V et al (2015) A review on computer vision based defect detection and condition assessment of concrete and asphalt civil Infrastructure. Adv Eng Inform 29(2):196–210
Yang G, Chen T, Li P et al (2020) Application of ultrasonic and radar methods in tunnel lining crack detection. Hunan Traf Sci Technol 46(1):45–46
Wang Y (2012) Research on identification of concrete structural cracks using acoustic emission method. Construction Technology S2:127–129
Badrinarayanan V, Kendall A, Cipolla R (2017) SegNet: a deep convolutional encoder-decoder architecture for image segmentation. IEEE Trans Pattern Anal Mach Intell 39(12):2481–2495
Rowshandel H, Nicholson GL, Shen JL et al (2018) Characterisation of clustered cracks using an ACFM sensor and application of an artificial neural Network. NDT and E Int 98:80–88
Shi B, Xu X, Wang D et al (2005) Research on distributed optical fiber strain monitoring technology for tunnel health diagnosis based on BOTDR. Chin J Rock Mech Eng 24(15):2622–2628
Zhou Q (2017) Research on surface state detection technology of vehicle-mounted tunnel repair[D]. China Academy of Railway Sciences, Beijing
Xie X, Lu X, Tian H et al (2013) Measurement method of tunnel full section deformation based on ground 3D laser scanning technology. Chin J Rock Mech Eng 32(011):2214–2224
Huang S, Bao T, Li Y et al (2023) Semantic segmentation method for hydraulic concrete crack based on improved deeplab V3+ network. Adv Sci Technol Water Resourc 43(1):81–86
Huang Y, Liu X, Yuan Y et al (2012) Automatic detection technology for shield tunnel leakage. J Tongji Univ (Natural Science) 46(01):73–78
Zhu Y, Tang H (2023) Automatic damage detection and diagnosis for hydraulic structures using drones and artificial intelligence techniques. Remote Sens 15:615
Acknowledgements
This study was supported by the China Yangtze Power Co., Ltd. Research Project (Z212102007).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Copyright information
© 2024 The Author(s)
About this chapter
Cite this chapter
Deng, X., Zhou, Z., Chen, Z., Xu, Y., Cai, W. (2024). A Method and Device for Tunnel Defect Detection Based on Image Automatic Collection and 3D Reconstruction. In: Mei, G., Xu, Z., Zhang, F. (eds) Advanced Construction Technology and Research of Deep-Sea Tunnels. Lecture Notes in Civil Engineering, vol 490. Springer, Singapore. https://doi.org/10.1007/978-981-97-2417-8_13
Download citation
DOI: https://doi.org/10.1007/978-981-97-2417-8_13
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-97-2416-1
Online ISBN: 978-981-97-2417-8
eBook Packages: EngineeringEngineering (R0)