Automatic cleanness rating of steels using image analyser and inclusion charts

Abstract

A Quantimet 720 automatic image analyser has been used in a programme aiming to develop a new method for rating the cleanness of steels automatically. The method is based on mathematical pattern-recognition techniques, and combines the advantages of classical cleanness rating using picture charts with those of quantitative methods using automatic image analysers. The morphologies of the pictures containing inclusions were characterized by geometric quantities measured using the Quantimet. Classification of inclusion pictures was carried out on the basis of generalized ‘distance’ functions. The mathematical operations necessary for classification were performed by a PDP-1 1/23 computer linked to the Quantimet. Comparative tests on steel specimens and schematic inclusion pictures showed good agreement between the results obtained by automatic classification and those obtained by conventional visual assessment. The method seems to be suitable for the automatic classification of other microscopic images.