evaluation of intensity augmented ICP for terrestrial LiDAR data registration
Final pub: Lohani, B., & Sasidharan, S. (2017). An evaluation of intensity augmented ICP for terrestrial LiDAR data registration. Journal of Geomatics, 11(2). While using laser scanner for map-making or developing 3D models of objects, it is important to scan asite or an object from multiple viewpoints. These different scans are integrated to generate a complete point cloudwhich is then used for developing the map or 3D model of the site. ICP (Iterative Closest Point) is a standard algorithmfor registration of point clouds. However, in the absence of marked features which are geometrically distinct in thepoint clouds, which are being combined, this method sometime fails. This paper exploits the radiometric data that arealways obtained along with the coordinates and devises a novel approach for scan registration. Before usingradiometric data in registration process, the data are normalized. The algorithm presented in this paper works in twostages- Intensity Augmented ICP (IAICP) for coarse registration stage and conventional geometric ICP at the fineregistration stage. The proposed approach is successfully applied to a few test data captured by Optech ILRIS 36-Dresulting in an accurate estimation of the transformation parameters. A comparison of the conventional and intensityaugmented registration approaches is also presented. The results indicate the supremacy of IAICP over the ICP, as thelatter is found to fail in geometrically confusing cases while the intensity augmented ICP gives satisfactory result insuch cases.