A fast detection method for region of defect on strip steel surface

Abstract

In order to meet the increasing demands of high efficiency and accuracy for strip steel production line, a fast detection method for region of defect (ROD) on strip steel surface is proposed in this paper. Firstly, the efficiency requirement of ROD detection algorithm is described. Secondly, mean filter improved in speed is used to filter noise. Then, five statistical projection features are extracted from detection region on surface image. Finally, based on distinct feature vector dataset, extreme learning machine (ELM) classifier, region of background (ROB) pre-detection and classifiers selection are combined together to realize two-class classification of ROD and ROB. Experimental results show that the novel method proposed in this paper not only is of high detection accuracy and efficiency but also can satisfy on-line ROD detection.