期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2007
卷号：XXXVI-3/W52
页码：178-183
出版社：Copernicus Publications
摘要：Several approaches for automatic registration of terrestrial LIDAR data exist. However, they normally can not be compared to each other because of a lack of reference data. This is especially true for applications in urban areas. One dataset available for this purpose is a set of eight LIDAR scans from Agia Sanmarina, a Byzantine church in Greece, which has been made available by the ISPRS working group V/3 on terrestrial laser scanning. We have tested our plane based approach for automatic registration on this particular dataset: The point clouds are first split into a regular raster, then for each raster cell, the dominant plane is robustly estimated and denoted as surface element. Coarse registration is carried out via grouping the surface elements to large planes and a generate-and-test strategy to find transformation parameters that maximize the inlier count. Pairwise fine registration is accomplished using a variant of the ICP (iterative closest point) algorithm that is based on matching surface elements instead of 3D points. In addition to this, the theoretical framework for a simultaneous fine registration of multiple datasets is presented in this paper
关键词：Terrestrial; LIDAR; Building; Point Cloud; Registration; Algorithms; Automation; Performance
