Pattern Recognition of Microstructures using Neural Networks

Abstract

A pattern recognition system for microstructures using neural networks has been developed. The system consists of two neural networks, a feature extraction network and a classification network. The feature extraction network classifies local area patterns of an input image into reference vectors and generates the histgram of the classified vectors. The classification network learns the histgram data of sample images using the back-propagation algorithm and classifies the microstructures into pre-determined categories. The system shows high performance in the classification of graphite shapes in cast irons and of grain sizes of stainless steels.