RECOGNIZING 3-D OBJECTS BY USING A HOPFIELD-STYLE OPTIMIZATION ALGORITHM FOR MATCHING PATCH-BASED DESCRIPTIONS

A new method is proposed for recognizing 3-D objects by using a Hopfie ld-style optimization algorithm based on matching patch-based image an d model descriptions. To obtain the image descriptions, range images a re employed to extract reliable high-level patch Features. In the opti mization process, the objective function is a Liapunov function which encodes a set of geometric constraints on the descriptions. The optimi zation is implemented in a Hopfield network with its interconnections encoding the imposed unary, binary and bounding edge constraints. At f irst, the paper makes an explanation on a new pre-processing method fo r deriving the required image description. It then presents the struct ure of the used Hopfield network that is able to recognize multiple mo del objects all at the same time. Experimental results based on synthe tic or real range images are also reported. (C) 1998 Pattern Recogniti on Society. Published by Elsevier Science Ltd. All rights reserved.