Vision-based change detection for inspection of tunnel liners

Abstract

Tunnel inspections may demand personnel to access hazardous environments soliciting the need for robotic operations to minimize human intervention. CERN, the European Organisation for Nuclear Research, has a number of tunnel infrastructures, including the tunnel hosting the Large Hadron Collider (LHC). A Train Inspection Monorail (TIM) was installed in the LHC tunnel to reduce personnel intervention. It gathers data from various sensors and captures images which, up till now, were only used for data record purposes. In this paper we present a computer vision system, TInspect, that uses a robust hybrid change detection algorithm to monitor changes on the LHC tunnel linings. The system achieves a high sensitivity of 83.5% and 82.8% precision, and an average accuracy of 81.4%. The proposed system is also configurable through different parameters to adapt to different scenarios, making it useable in other tunnel environments and therefore not exclusive to the LHC tunnel.