Evaluation of the Convergence Region of an Automated Registration Method for 3D Laser Scanner Point Clouds.docVIP

Sensors 2009, 9, 355-375; doi:10.3390/OPEN ACCESS sensors ISSN 1424-8220 /journal/sensors Article Evaluation of the Convergence Region of an Automated Registration Method for 3D Laser Scanner Point Clouds Kwang-Ho Bae Department of Spatial Sciences, Curtin University of Technology, GPO Box U1987, Perth, WA 6845, Australia; E-Mail: K.H.Bae@.au; Tel.: +61-8-9266-7604; Fax: +61-8-9266-2703 Received: 8 December 2008; in revised form: 30 December 2008/ Accepted: 4 January 2009 / Published: 8 January 2009 Abstract: Using three dimensional point clouds from both simulated and real datasets from close and terrestrial laser scanners, the rotational and translational convergence regions of Geometric Primitive Iterative Closest Points (GP-ICP) are empirically evaluated. The results demonstrate the GP-ICP has a larger rotational convergence region than the existing methods, e.g., the Iterative Closest Point (ICP). Keywords: Convergence region; registration; terrestrial laser scanners, point clouds. 1. Introduction Laser scanners provide a three-dimensional (3D) sampled representation of the surfaces of objects with high spatial resolution and have been gained popularity in terrestrial and airborne applications such as 3D-reconstruction of terrain [17], tree height estimation [36] and building segmentation [27]. Those datasets are generally called 3D point clouds. Since laser scanners have a limited field of view, it is necessary to collect data from several locations in order to obtain a completed representation of an object. These data must be transformed into a common coordinate system for further analysis. This procedure is called the registration of point clouds. A method for the automated registration of two point clouds is summarised as follows: - - Finding correct correspondence from either selected point-to-point or point-to-surface pairs. Adjustment algorithms for estimating the