Invariant object recognition using Fahlman and Lebiere’s learning algorithm

Abstract

A new neural network system for object recognition is proposed which is invariant to translation, scaling and rotation. The system consists of two parts. The first is a preprocessor which obtains projection from the input image such that, for any rotation and scaling of standard image, the projection results are reduced to cyclically shifted ones, and then adopts the Rapid Transform [9] which makes the projected images cyclic shift invariant. The second part is a neural net classifier which receives the outputs of preprocessing part as the input signals. The most attractive feature of this system is that, by using only a simple shift invariant transformation (Rapid Transform) in conjunction with the projection of the input image plane, invariancy is achieved and the system is reasonably small. Experiments with six geometrical objects with different degree of scaling and rotation show that the proposed system performs excellent when the neural net classifier is trained by Fahlman and Lebiere’s learning algorithm [8].