Finding anomalies in high-density LiDAR point clouds
Modern three-dimensional (3D) terrestrial scanning systems such as the TITAN system make it possible to acquire precise, geo-referenced datasets over areas spanning many square kilometers. These systems consist of several vehicle-mounted lasers for acquiring point data, a high-precision GPS unit for providing geo-reference data, and an inertial measurement unit (IMU) for tracking vehicle motion. Two liabilities of this type of approach are that the navigational accuracy degrades when the GPS signal is lost, and moving objects can cause data artifacts. These induce particular anomalies in the acquired data that must be eventually corrected, often by hand, during the post-processing stage. The goal of this paper is to show that by exploiting the configuration of the LiDAR sensors, such anomalies can often be detected automatically from the dataset. In particular, we demonstrate that under an appropriate tessellation, iterative closest point (ICP) algorithms can be used to reliably localize anomalies and provide an estimate of their magnitude.
Harrison, J.W., Feme, F.P., Hefford, S.W., Samson, C, Kusevic, K., Mrstik, P., & Lies, P.J.W. (2009). Finding anomalies in high-density LiDAR point clouds. Geomatica, 63(4), 397–405.