Terrestrial laser scanners provide a three-dimensional sampled (i.e. point cloud) representation of the surfaces of objects resulting in a very large number of points. They have great potential to improve the measurement and representation of remote and widespread objects for applications such as engineering metrology, cultural heritage recording and so on. However, most laser scanners have a limited field of view so it is necessary to collect data from several locations in order to obtain a complete representation of an object. This procedure is called the registration of multiple point clouds.

Existing registration methods, such as the Iterative Closest Point (ICP) by Besl and McKay (1992) or Chen and Medioni’s (1992) method, work well only if good a priori alignment is provided. This paper presents an automated registration method for the Geometric Primitive ICP with the RANSAC (GP-ICPR) that utilises geometric primitives, neighbourhood search and the positional uncertainty of laser scanners. In addition, the GP-ICPR was tested with both simulated and real-point clouds in terms of registration error and the convergence region. These tests on the convergence region of the GP-ICPR effectively demonstrate that the proposed method has about 1m translational and on the order of 10° rotational convergence with terrestrial laser scanner datasets. Although there is room for improvement to achieve a fully automated registration of three-dimensional point clouds, the current level of rotational and translational convergence region of the GP-ICPR was demonstrated to be effective for on-site registration of point clouds from a terrestrial laser scanner.