Automatic Detection of Shield Tunnel Leakages Based on Terrestrial Mobile LiDAR Intensity Images Using Deep Learning

Abstract

Water leakages are very important signals that characterize the serious potential structural damages or flaws in shield tunnels. Automatic, timely, and accurate detection of water leakages is of great significance to the safe operation and maintenance for shield tunnels. However, existing methods (e.g., passive optical images) are highly limited by the confined spaces and dim light conditions in shield tunnels. This paper proposes a new method that uses the intensity images of terrestrial mobile LiDAR (Light Detection and Ranging) for automatic leakage detection in shield tunnels based on deep learning. A self-developed terrestrial mobile LiDAR system (Faro Focus X330) are used to simultaneously obtain the point clouds and intensity information of the shield tunnels. The original intensity data are corrected for the distance effect to generate intensity images. A improved Fully Convolutional Network (FCN) based VGG-19 that is a network structure in deep learning is constructed to achieve accurate tunnel leakage detection based on intensity images. The results show that the proposed method can rapidly and accurately detect leakages in shield tunnels and the leakage detection results are not affected by the tunnel attachments (e.g., bolt holes, cables, and metal facilities). The error rate and segmentation speed are 4.8% and 0.6 s, respectively.