A Real-time Tunnel Surface Inspection System using Edge-AI on Drone

Inspection of tunnel lining surface is an essential yet challenging task in tunnel construction. Timely identification of tunnel lining defects (e.g. water leakage, cracks and spalling, etc.) is critical for the quality control in the construction stage. Conventional methods rely on visual inspection and manual recording of the inspection results inside the congested tunnel environment, which is time-consuming and inefficient. In recent years, the advancements in AI and drone technologies show promising prospects in tunnel inspection. In this study, an AI-powered inspection system using edge-computing device on a specially configured drone is developed for efficient and accurate inspection of the completed tunnel lining surface. The proposed sensor-fusion method for defect detection and SAHI-based method to detect small cracks, spalling, and water leakages on the tunnel lining surface demonstrated satisfactory results in the pilot application in the construction of Trunk Road T2 and Cha Kwo Ling Tunnel in Hong Kong. This is the first reported onboard AI tunnel surface inspection system in the world that can be applied in other tunneling and infrastructure projects in the foreseeable future.