一种基于机器视觉的玻璃质量在线检测系统  

Machine-vision Based Online Quality Detection System for Glass Production

作　　者：易乔木[1] 程金树[2] 周洋[2] 陈幼平[2] 

机构地区：[1]华中科技大学机械科学与工程学院,武汉430074 [2]武汉理工大学材料科学与工程学院,武汉430070

出　　处：《武汉理工大学学报》2007年第5期23-26,共4页Journal of Wuhan University of Technology

基　　金：湖北省自然科学基金(2005ABA269);武汉市科技攻关计划

摘　　要：对玻璃缺陷识别原理进行了介绍,并对基于机器视觉的运动图像实时处理算法进行了研究。由此,提出了玻璃质量在线检测系统的总体设计方案,并提出了一种基于自适应阈值选取的图像二值化分割、缺陷特征提取并通过神经网络分类器识别不同种类缺陷的新算法。在此基础上,研制开发了一个玻璃质量在线检测系统,并将其运行于工业现场。实际使用情况表明该系统是可行且有效的。Based on the analysis of the domestic trends of the online quality detection system for glass production, in this paper, the general principle of glass defect recognition was discussed and the related machine vision based real-time processing algorithm for motion image was studied. A general implementation scheme of the online quality detection system for glass production was proposed. An algorithm of self-adapted threshold selection was put forward for image binary processing and defect feature abstraction. A neural network based sorter was designed to distinguish different defect types. Furthermore, an online quality detection system of glass production was implemented and such a system was running on the field. It was shown that the system was feasible and effective in practice.

关 键 词：机器视觉 在线检测 图像处理 特征提取 神经网络 

分 类 号：TH12[机械工程—机械设计及理论]