Modeling tunnel profile using Gaussian process

Tunnel profile is an important quality measure for deformation monitoring throughout lifetime of a tunnel. Recently, statistical models such as cylindrical regression have been introduced to model the tunnel profile with the help of Terrestrial Laser Scanning (TLS) surveyed point cloud data. However, these models cannot capture the local variability of tunnel surface, which is important for deformation monitoring. This paper presents a Gaussian process (GP) based approach to model the tunnel profile. We introduce a reading-surface profile which uniquely determines a three-dimensional tunnel model in Cartesian coordinate system. Based on the reading-surface, we can fit the GP model and obtain continuous prediction of the tunnel surface. We present a case study to demonstrate that our model is flexible to capture the local variability of the tunnel surface, and therefore has great potential for deformation monitoring.