Neural networks for the recognition and pose estimation of 3D objects froma single 2D perspective view

In this paper we present a neural network (N-N) based system for recognitio n and pose estimation of 3D objects from a single 2D perspective view. We d evelop an appearance based neural approach for this task. First the object is represented in a feature vector derived by a principal component network . Then a NN classifier trained with Resilient backpropagation (Rprop) algor ithm is applied to identify it. Next pose parameters are obtained by four N N estimators trained on the same feature vector. Performance on recognition and pose estimation for real images under occlusions are shown. Comparativ e studies with two other approaches are carried out. (C) 2001 Elsevier Scie nce B.V. All rights reserved.