Surface Defect Detection And Classification For Marine Steel-Plate On-line

As one of the main products in the form of iron and steel industry, the steel plate is essential raw materials in the field of shipbuilding, chemical manufacturing, aerospace and other industrial areas; the surface quality has a direct impact on the quality and performance of the steel produced products, a tiny surface defect will reduce the corrosion resistance, abrasion resistance, and ultimately bring catastrophic harm to a molded product. So it becomes very important doing real-time defect detection and defect classification on the steel surface effectively and efficiently.This paper developed a set of intelligent machine vision-based goal-line defect detection system according to the defect site characteristics and needs of marine steel plate, ship plate surface, which is on the basis of the analysis of defects in steel surface line detection methods home and abroad. The system’s hardware system consists of image acquisition system, image processing and display system, lighting system, alarm system, control system and dust control system. Software system consists of image acquisition software and image processing software and image display software and machine control software.This paper describes image processing algorithms process for the marine online defect extraction of steel-plate surface, especially for the image plate surface pretreatment process in the case where some information of the plate edge is etched in the background lighting, After the image edge detection, through the local edge links by inflating and the overall edge links based on the Hough transform, completing the edge extraction in the complex case. Then do the threshold extracting using the Fisher threshold algorithm, and highlight a TF fast clustering method of filtering used on the image surface defects clustering process.Due to the processing technology and equipment and other various reasons of the500samples collected at the scene defects, the surface are prone to various types of defects, such as scratches, eye of a needle, scarring, pitting, indentation and orange peel. In the analysis of the existing feature selection methods of various defects and defect classification, this paper select a defect classification method to meet the real-time nature of the field defect classification. And grading the steel-plate images with an AHP and fuzzy decision theory method, finally, excluding the pseudo defects which is caused during transport due to the fall on the area of artificial sand or paint marks based on their inherent gray and geometric characteristics.