Model-based matching using skewed symmetry information

A model-based approach is proposed for recovering the 3-D position of a symmetrical polygon in the models from its single image. First, a Hough-type voting method detects skewed symmetry in the image. Matching of the skewed symmetry axis to symmetry axes of models then makes both selection of models and recovering their orientation easy. The method is shown to possess robustness of matching from occlusion