Signal processing for automatic optical su rface inspection of steel strip

Abstract

This paper reports on a continuing collaborative research project involving the Instrument Systems Centre (now the Measurement and Instrumentation Centre) of the City University, the British Steel Corporation, and the Sira Institute, and concerned with automatic optical surface inspection for cold rolled steel strip. The steel surface is interrogated by a high-speed laser scanning system and the resulting multi-channel response signal is processed to achieve detection, delineation (ie, boundary extraction), and identification (by type) of significant surface defects. Processing is organised in a hierarchy, using multiple matched filters and modified image processing operations for detection and delineation, and the techniques of automatic pattern recognition for identification. Extensive studies in computer simulation are described for the development and evaluation of this processing system. The resulting structure is relevant to many similar problems of plain surface inspection.