浮法玻璃缺陷在线识别技术的研究  

Research on Online Recognition Technology for Float Glass Defects

作　　者：刘怀广[1] 陈幼平[1] 谢经明[1] 彭向前[1] 

机构地区：[1]华中科技大学机械科学与工程学院,湖北武汉430074

出　　处：《小型微型计算机系统》2011年第4期738-742,共5页Journal of Chinese Computer Systems

基　　金：湖北省科技攻关重点项目(20081032112)资助

摘　　要：表面缺陷是影响浮法玻璃质量的主要因素,针对目前国内在浮法玻璃缺陷识别正确率不高的现状,本文结合玻璃缺陷低灰度的特点,利用两次一维OTSU缺陷分割方法,实现了缺陷核心的有效分割.然后根据不同缺陷核心灰度分部的特点,提出12种具有统计性的特征,利用改进的神经网络的非线性映射能力实现了缺陷的正确识别.最后的实验结果证明,算法具有较好的效果.Defects existing is the main factor effecting quality, of float glass. As the low correctness of float glass defects in our country, a segmentation method of twice OTSU was proposed according to the low grey level of glass defects. The defect cores were separated correctly based on the method. Twelve statistical characters were obtained for defect recognition. Finally, defects were classified correctly with the nonlinear mapping ability of neural network. The final experiment results showed efficiency of the method.

关 键 词：图像处理 特征提取 缺陷识别 神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]