The detection and documentation of water leakage defects are critical for asset management of ageing underground infrastructure. This study develops a computer vision-based method to demarcate tunnel leakage areas using a robot-mounted imaging system. The robotic system, the Tunnel Inspection Monorail (TIM) equipped with a 360-degree camera, is deployed for the remote video data acquisition. A customized Omnidirectional Image (ODI) processing pipeline is proposed to transform the gathered omnidirectional video into two-dimensional tunnel mosaics with units of 10-metre length. Furthermore, an ensemble deep learning strategy leveraging transfer-learning is employed to generate a scaled demarcation map of tunnel leakage areas. The proposed method was deployed to inspect a 3.45 km-long tunnel alignment at CERN, the European Organization for Nuclear Research. Results demonstrate that the method can visualize the tunnel lining condition over several kilometres augmented with the automated leakage demarcation, facilitating the identification of severely leakage-effected chainages. In addition, the method allows to reveal the spatial distribution characteristics of leakage defects in the large-scale underground infrastructure, both longitudinally and circumferentially. The method proves its practical significance in terms of remote, automatic, intuitive, and scalable water leakage monitoring with the enhanced efficiency and accuracy compared to the manual inspection. The subsequent statistical analysis reveals inherent leakage defect characteristics and enhances the understanding of the waterproofing deterioration of the ageing CERN tunnels.
Ouyang et al. (Mon,) studied this question.