Automatic Pattern Classification of Real Metallographic Images

Abstract

This paper addresses the problem of automatic pattern classification in real metallographic images from the steel plant ArcelorMittal Ostrava plc (Ostrava, Czech Republic). Images of manufactured metal plates contain dark dots, i.e. imperfections. We monitor the process quality in the steel plant by determining automatically the number and sizes of these dots which represent plates' imperfections. The proposed algorithm segments the area of plates that contains dots, identifies rows of pixels that contain them, marks and counts them. The obtained results are promising and confirm that the proposed algorithm should serve as the foundation for future research in this area.