3D real object recognition on the basis of moment invariants and neural networks

Abstract

In this study, recognition system of the completely visible 3D solid objects of the real life is presented. The synthesis of analyzing two-dimensional images that are taken from different angle of views of the objects is the main process that leads us to achieve our objective. The selection of "Good" features those satisfying two requirements (small intraclass invariance, large interclass separation) is a crucial step. A flexible recognition system that can compute the good features for a high classification is investigated. For object recognition regardless of its orientation, size and position feature vectors are computed with the assistance of nonlinear moment invariant functions. After an efficient feature extraction, the main focus of this study, recognition performance of artificial classifiers in conjunction with moment-based feature sets, is introduced. © Springer-Verlag 2004.