Study of automatic surface inspection for cold rolled steel strip

Abstract

The paper describes a feasibility study of on-line inspection of visible defects in the surface of cold rolled steel strip by a experimental inspection system. The system incorporates CCD linear image sensor and illumination placed in specular reflection direction . To cope with the wide variety of different defect mixed with lot of noise and continuous changing of environment, the paper propose a segmentation algorithm using a special two dimensional filter and adaptive threshold varying with environment to get a quite accurate defect binary image. Defect deleanation and classification algorithm analyses defects, removes noise and extracts features of defects, then classify them by statistics decision tree.