双阈值热点区域提取的线材表面缺陷检测方法  

Wire Rod Surface Defect Detection Method with Double Threshold Region of Interest Extraction

作　　者：甘胜丰[1] 王典洪[2] 孙林 

机构地区：[1]湖北第二师范学院计算机学院,湖北武汉430083 [2]中国地质大学(武汉)机械与电子信息工程学院,湖北武汉430074 [3]沙钢钢铁研究院,江苏苏州215625

出　　处：《机械设计与制造》2016年第7期127-129,134,共4页Machinery Design & Manufacture

基　　金：国家自然科学基金(61271274);湖北省自然科学基金(2014CFB380)

摘　　要：机器视觉的表面质量检测方法已广泛应用于工业生产,但高速线材由于其外形和工况的特殊性,鲜有应用。研究设计了4维面阵CCD与双色环形光源为图像采集单元,组合对比度边界检测算法、梯度和彩色图像均值阈值热点区域提取和二维最大熵图像分割算法成基于双阈值热点区域提取的线材表面缺陷检测方法。实验结果表明,该方法弥补了单色光源条件下采用单一阈值提取缺陷热点区域的劣势,耗时1.25s可实现多种类二维和三维线材缺陷的准确分割。The technology of machine vision has been widely used in surface qual@ detection of industrial products, except widely used in high speed wire rod, which is high temperature and red column. The method of double -threshold Region of Interest extraction have been suggested, which is consisted of the edge detection algorithm, the threshold algorithm by gradient and the color image mean, and two dimensional maximum entropy image segmentation algorithm. All of these algorithms are used to process images extracted form the image acquisition equipment with 4 dimensional CCD and double color ring light source. The results show that this method can segment multi-classes two dimensional and three-dimensional defects accurately, and make up for the shortage of the single threshold defect image extraction.

关 键 词：线材缺陷 热点区域提取 图像梯度 彩色图像均值 

分 类 号：TH16[机械工程—机械制造及自动化] TP181[自动化与计算机技术—控制理论与控制工程]