Unmanned Aerial Vehicle (UAV) navigation and image processing techniques for underground tunnel infrastructure assessment

In many countries across the world, hundreds of kilometres of tunnel infrastructure networks require regular inspection and maintenance to ensure the safety and serviceability of lifeline transportation systems. Large-scale underground infrastructure comprises twin tunnel cross passages, shafts, caverns, and other complex geometries, which poses great challenges for efficient onsite inspection and even health and safety issues for workers in underground environments. Compared to conventional inspections by workers and ground vehicles or robotics, Unmanned Aerial Vehicles (UAVs) emerge as a cost-effective and adaptive monitoring tool, particularly suited for the access to shafts and caverns. However, there are challenges in implementing UAV-based inspection for tunnels, including GPS-denied environments, complex geometries, and limited flight time. To address these issues, this study proposes a UAV inspection method to provide an efficient and reliable UAV-based solution for tunnel infrastructure assessment. The Proximity Move-Pause-Photo for Surface Defect Inspection (PMPP-SDI) methodology is a technique for UAV tunnel inspection. The method uses flexible flight routines to follow the tunnel lining, enabling comprehensive coverage of the tunnel surface. This real-time reactive control generates coloured imagery data, as well as a densified point cloud for precise local spatial information. The method demonstrates a balance between cost and performance of automatic tunnel inspection. And it has the potential to reveal defects and irregularities at a smaller scale compared to previous research using tunnel segments as the unit of analysis. It also allows for image processing-based surface defect identification. An innovative image processing workflow is presented for UAV tunnel image processing. This workflow allows the generation of 3D tunnel models, an area that hasn't been accomplished in previous research. This workflow facilitates tunnel geometry and defect analysis for comprehensive tunnel health inspection in both spatial information and surface defects. Spatial information extraction is enabled using a Digital Surface Model (DSM), which supports UAV control examination and geometry examination. Additionally, surface defect identification and quantification are carried out through segmentation of the orthomosaics of the model. Furthermore, the workflow integrates machine learning (ML)-driven point cloud trimming and defect identification to ensure precise model reconstruction and defect analysis, thereby improving overall accuracy in tunnel inspections. A field test of the proposed UAV navigation and image processing method is carried out in a railway tunnel, targeting defect detection and tunnel health assessment. The uneven illumination and lack of GNSS signals in the tunnel make manual inspection challenging. The findings reveal the tunnel lining's deformation pattern, with distinct deformation zones identified based on the colour of the stone and brick tunnel lining. The study also explains the structural health of the railway tunnel under moisture impact. These results demonstrate the potential of UAV-based tunnel inspection for infrastructure assessment, offering advantages in terms of safety, efficiency, cost savings, and data accuracy. In the second field study, a road tunnel's surface defect distribution was examined using the PMPP-SDI method. The tunnel inspection faced challenges due to its intricate geometry, including twin tunnels, vehicle cross passages (VCPs), and laybys. Specific attention was given to compressive defect distribution during the analysis. The results identified three separate zones depending on their distance from VCPs, each displaying varying surface defect densities and distributions. High-density cracks and spalling in vulnerable zones suggest further specific monitoring and maintenance. The test showcases the PMPP-SDI method's value in evaluating and prioritizing maintenance for complex transportation infrastructure. In conclusion, this thesis presents a systematic UAV-based tunnel inspection method, consisting of PMPP-SDI for real-time underground navigation and innovative image processing workflow. The effectiveness of the proposed inspection method was evaluated in two field tests on a road tunnel and a railway tunnel, respectively. Field test results identified distribution patterns of surface defects and deformation, highlighting the potential of UAV-based tunnel inspection for infrastructure assessment. Moreover, the study demonstrated the potential of ML-driven technologies to enhance model quality, suggesting further research directions for the advancement of UAV-based tunnel inspection systems.

Grantor dc:publisher

University College Cork

Year dc:date.issued

2024

Author dc:creator

• Zhang, Ran

Advisor dc:contributor.advisor

• Li, Zili

• Underground tunnel monitoring
• Unmanned Aerial Vehicles (UAVs)
• UAV photogrammetry
• Surface defect inspection
• Automatic inspection

Statement dc:rights

• © 2024, Ran Zhang.

Licence dc:rights.uri

• https://creativecommons.org/licenses/by-nc/4.0/

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10468/16993

OAI identifier oai:identifier

oai:cora.ucc.ie:10468/16993