摘要
The surface quality of the steel slab directly affects the performance and quality of the final steel products.Due to the raw material,rolling equipment and processing technology,however,different types of surface defects might exist on the steel slab surface,such as scarring,crack,roll marks,scratches,pinholes,phosphorus leather,holes,pitting,etc.These defects not only affect the appearance of the steel products,but also reduce the performances of steel products,such as the corrosion resistance,wear resistance and fatigue strength of the product.Thus,it is very important to detect and identify automatically surface defects of steel slab products on-line and to control and improve the surface quality of the steel products.This is a very complex problem.In this paper,we introduce some machine vision techniques to detect and identify defects from the steel slab surface.An adaboost based machine leaming method is proposed to detect defects from steel slab surface.It is usually able to remove pseudo-defects.A directional detector based on wavelet transformation is used to detect directional structural crack feature from the complex steel slab surface.A saliency based method combining elastic threshold method and Gabor wavelet filter is used to detect crack from the steel slab surface.Experimental results show the effectiveness of these methods.