Mobile LiDAR-Based Convergence Detection in Underground Tunnel Environments

Abstract

This paper presents a mobile LiDAR-based method for remotely identifying convergence (i.e., naturally occurring deformation) in excavated underground tunnel environments. A mobile LiDAR system is used to collect and generate two independent 3D point clouds of the excavated environment. In the absence of actual convergence, simulated deformation is applied to one of the two point clouds based on a simple convergence model. Registration of the 3D data is performed by using a rough alignment based on principal components, followed by a piecewise iterative closest point (ICP) algorithm. The residual point-to-surface distances are then used as a deformation signal, which is filtered using a modal analysis based on expected deformation shapes as well as a median filter. It was found that convergence deformations of 0.05 m could be confidently identified and deformations as low as 0.0125 m could be detected within residual deformation data with a mean absolute error of approximately 0.0235 m. The proposed technique therefore allows deformations on the same order as background noise to be characterized and flagged for further inspection by mine operators.