Fast digitizing of 3D shapes by automatic alignment of multiple range maps

Range maps registration still represents one of the most time consuming phases in the digitization of 3D shapes due to the high user intervention required. The traditional approaches are, in fact, based on manual rough alignments, followed by ICP refining techniques. On the other hand, existing unattended methodologies, based on the automatic searching for correspondent morphological singularities on adjacent point clouds, do not seem to guarantee sufficient robustness and flexibility in the fast reconstruction of target objects. This paper presents a methodology to acquire free-form shapes by combining a 3D stereo vision system and a fully automatic range maps registration process. The automatic alignment is carried out without any assumptions about the initial positions of the point clouds and is based on the automatic detection of fiducial markers, located on the surface object, by processing grey intensity images.