Towards Automatic Registration of 3-D Views

Recent advances in the field of laser scanning technology along with the availability of more powerful computing resources have favoured the increasing interest of surveyors, architects, archaeologists towards laser scanners technology as a very promising alternative for cultural heritage surveying. Thousands of points can be acquired in short time with an accuracy that is adequate to build 3D models for single objects so as for whole environments. At the present, resulting 3D digital models offer an invaluable mean for documentation, archiving, structural analysis and restoration of the large amount of objects spanning from Civil Engineering to Cultural Heritage assets. Usually, the end products of the whole workflow (survey and modeling) are VR representations (Vrml, Flash), movies (AVI, DVx, Mpeg), Dense Digital Surface Models (DDSM) and orthophotos as well. The creation of a 3D model requires a lot of data about the object surface or volume, which have then to be somehow aggregated, regardless the data structure and the acquisition device used. In most cases, the data registration step is based on ICP, that iterativelly finds the mutual orientation between two range maps, starting from an initial estimate provided by an operator. This approach is often time-consuming, increases the final cost of the 3D model and represents one of the major limits to the wide spreading of real object models. In this paper an overview of an automatic range data registration system is presented, whose theoretical basis rely on the previous works of A. E. Johnson and L. Lucchese. The former proposed an innovative solution for the recognition of similarities between 3D surfaces, introducing the spin-image concept, while the latter developed an alternative procedure (Frequency Domain technique) for 3-D motion estimation of free-form surfaces, based on the Fourier transform of the 3-D intensity function, implicitly described by the registered time-sequences of range data.