A neural network system for robot vision

Abstract

The authors describe an approach to the development of a system of multiple neural networks for the purpose of recognizing and learning objects with position, rotation, and scaling independence. The approach involves the linking of several neural-network schemes (the Kohonen self-organizing network, the neocognitron, and the adaptive resonance theory network) into a larger system dedicated towards robot vision. Linking these networks together in a specified order should allow each network to contribute its strength to the overall process of recognizing and learning objects.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>