Non-rigid object recognition using principal component analysis and geometric hashing

Abstract

A novel approach is proposed to recognize non-rigid 3D objects from 2D images using principal component analysis and geometric hashing. For all of the models that we want to be able to recognize, we calculate the statistic of point features using principal component analysis and then, calculate the invariants of them. In recognition stage, we calculate the needed invariants from an unknown image and used as indexing keys to retrieve from the model base the possible matches with the model features. We hypothesize the existence of an instance of the model if a model's features scores enough hits on the vote count.