Matching correspondence between images and 3D model in a reconstruction process

3D reconstruction from photographies is an active research trend. The resolution of the sensors is increasing and the data processing is more accurate (not only restricted to calibrated stereo vision). In archaeological research it becomes a common way to safeguard some views of an ancient site, coupled with a manner to describe in 3D the artifacts with the same set of photographies. Archaeological scientist are now facing a complex problem to handle these digital data. An important usage is to describe semantically the artifacts. It is generally made ”by hand”, supplied by the knowledge of the scientists. We propose a solution that can perform a part of this work automatically, to generate descriptions of the obtained geometry. It combines image processing, geometry processing, 3D reconstruction. This paper aims at presenting an algorithm for 2D/3D point matching. The 3D reconstruction process of model from multiple views based on SIFT algorithm. The matching process uses a 2D mask pattern to lookup the 3D corresponding point. Experimental results show that our matching algorithm is precise, highly flexible, and can be successfully applied to a variety of 3D shapes.