Automatic CAD-based construction of hybrid models for 3-D object recognition and localization

An automatic generation of object models from CAD data for object recognition and localization is presented in this paper. The model described here comprises the object's geometry and topology, its photometric properties and viewer-centered information. This information is automatically extracted from a CAD model and transformed into a form suitable for object recognition. Experiments demonstrating the object recognition and localization from stereo images using the developed model are presented.