Automated visual classification of frequent defects in flat steel coils

Surface inspection is one of the most important facets of quality-control systems in the steel manufacturing and processing industry. A number of surface defects can be detected by a visual inspection. However, human visual inspection becomes a hard task due to the high-processing speed. In the present paper, the automated visual inspection of flat steel is approached. A detailed description is given of the main aspects involved, concerning image acquisition, image processing algorithms, architecture design, the custom software developed, and data transmission and synchronization. Particular attention is paid to feature extraction and classification. Six kinds of defects are finely classified: weld, white rust, transporter marks, pitting corrosion, protuberance of zinc, and rolled marks. The proposed solution has been implemented and tested in a real industrial environment, in a flat steel cutting factory, showing suitable results. A successful classification ratio of about 87% of the known defects has been obtained, which is considered a reliable result.