Techniques for 3D Object Automatic Modeling

Generating 3D models of real objects is an ever more growing field in these last years. 3D modeling relies on registering range data acquired by 3D laser scanner, which allow to describe with high accuracy complex solid bodies, such as sulptures. These kinds of devices are based on sensors, which are able to retrieve the object shape via laser beam, without any phisical contact with the object itself. This paper concerns with the improvements and test results related to 3D motion estimate technique, currently applied at the University of Padua for automatic 3D modeling. The basics of this method were already explained in paper “Estimating 3D rototranslations from range data by frequency domain technique”, presented during previous 3D optical Conference, helded in Zurich 1997. As stated from the title, the generation of 3D models is approached working in the frequency domain, applying the Fourier transform to acquired object digital images. In this paper we will present the theoretical and practical modifications applied to abovementioned technique, along with the new algorithm implemented for improving the estimates of rototranslations between image pairs in 3D space. Furthermore assessment on robustness of proposed algorithm, on estimate accuracy and related test results will be also provided.