Correlation method for pattern recognition

Abstract

A new approach towards solving the pattern recognition problems in hybrid optical/digital correlators is suggested. The method is based on the authors' know-how-principle of transforming the field intensity or amplitude data on the identified object assuming that these data may be entered into a computer by ordinary methods. This principle allows to solve the unification of the correlation functions and recognition attributes selection problem using a unified, identified object's class-independent method. Presented experimental results confirm applicability of this approach.