We developed two systems to improve the efficiency of tunnel inspections. In railway tunnel maintenance, periodic inspections are mandatory. Structural conditions are currently evaluated through visual and hammering inspections. However, these inspections are time-consuming because inspectors must compare paper-based reference documents with the tunnel surfaces under poor lighting conditions. To address these challenges, we first developed an AI-based damage detection and soundness assessment application that automatically identifies surface defects and repair marks in tunnel images and evaluates overall tunnel soundness. Secondly, we developed a mobile projection system that uses red mesh overlays to visualize AI-detected defects on the tunnel lining and indicate critical areas. This paper presents these systems and reports on field validation results demonstrating their effectiveness.
YAMASHITA et al. (Fri,) studied this question.