TACK – Tunnlar Automatisk spriCK Övervakning genom Maskininlärning
Rock tunnels in Sweden are normally supported with a thin layer of fibre reinforced shotcrete in combination with rock bolts. Cracks in the shotcrete could lead to a failure of the support system and their existence and width should, therefore, be noted during the inspections of tunnels that are routinely performed. In the latest years, this work has been performed by in-situ surveys that are expensive and time-consuming. Recently, several studies highlight the potential of semi-automatic methods where a mobile mapping equipment (usually mounted on a vehicle) is used to capture the scene and to reconstruct the 3D model of a tunnel using a set of geomatics sensors (i.e., visible and infrared cameras, laser scanning, IMU). This digital representation of the tunnel is subsequently analyzed manually by visual inspection with the aim of seeking the crack and mark its extent. It is clear that, due to a large amount of collected data, these methods are inefficient and affected by human errors.
The project presented will entail a collaboration between KTH Royal Institute of Technology, University of La Sapienza and WSP Sweden. The aim of this research and development project is to investigate and develop a new technique to detect cracks on a tunnel lining using a hybrid approach of deep learning and photogrammetry. This will give detailed information regarding the condition of the whole tunnel which then will be used to assess its level of safety. With this technique, cracks will be automatically detected and measured from the imagery acquired using a customized mobile mapping system which leads to a highly efficient tunnel monitoring that can increase the overall safety of these infrastructures.
Kungliga Tekniska Högskolan
Andreas Sjölander, email@example.com
KTH Royal Institute of Technology, University of La Sapienza and WSP Sweden