Registration of multi-resolution point clouds from terrestrial laser scanners

Abstract

Registration of point clouds from terrestrial laser scanners plays the key role in feature extraction, 3D object modeling and 3D scene classification. An automated registration algorithm of pair-wise point clouds is presented, which is based on multi-resolution data created from raw point clouds, using geometric properties and point to point improved adjustment method, iteratively calculates six transformation parameters. RANSAC algorithm is also used to get reliable corresponding points. Convergence region and rate of proposed point clouds registration algorithm have been tested on a variety of data sets. Quality analysis of registration method is tested by using check points.