Artificial neural network as a shape descriptor and recognizer
Yunnan An, Guan Pang · The HKU Scholars Hub (University of Hong Kong) · 1992
The shape recognition problem is defined as follows: Let M be a set of known shapes. Given an input pattern, one wants to know whether this input is a member of M, and find the shape of M that matches the input pattern. It is clear that the shape recognition problem can be viewed as a mapping, by which the membership and correspondence of the given pattern with the known shape set are determined. An artificial neural network (ANN), when operating in a serial-deterministic mode, performs a deterministic mapping from the possible 2n inputs to a set of stable states. From the description above, we can see that ANN could be used as a shape descriptor and shape recognizer. When an ANN acts as a pattern recognizer, an input is recognized (i.e. accepted) if and only if it is a stable state of the artificial neural network. This paper shows some general features when ANN is used as a shape descriptor and recognizer.