浮法玻璃缺陷检测中的模式识别处理  

Pattern Recognition Processing in Defect Detection of Float Glass

作　　者：李淑芳[1] 张利辉[1] 陆琳琳[1] 赵尔男[1] 梁红[1] 

机构地区：[1]中国环境管理干部学院信息工程系,河北秦皇岛066004

出　　处：《科技广场》2012年第9期32-34,共3页Science Mosaic

基　　金：秦皇岛市科技局资助项目(编号:201101A060)

摘　　要：目前,在基于机器视觉的浮法玻璃质量在线监测系统中,首先需要对玻璃图像进行采集,对其进行处理和分析,以进行玻璃缺陷识别,然后用于后续操作中。本文根据对浮法玻璃各类缺陷的特征研究,通过分析反向传播学习神经网络识别算法的优缺点,采用一种改进方式用于浮法玻璃的缺陷分类和字符识别。经过实验数据测试表明,改进后的算法具有收敛速度快、识别误判率低的特点,能够为浮法玻璃质量在线检测系统中的玻璃打标和切割等操作提供准确、有效的信息,对于减少人工误差、保证产品质量等方面具有重要的意义。At present, machine vision is based on online quality detection system for glass production, glass and image acquisition, processing and analysis in order to provide effective information for subsequent operations. According to the characteristics of glass defects, this paper analyzes the advantages and disadvantages of the traditional BP algorithm, where an improved BP neural network recognition algorithm is applied to the glass defect classification and character recognition. The experimental results show that convergence speed of the algorithm is fast and the identification of false positives is low, compared with traditional BP recognition algorithm, and it can provide accurate and efficient information for glass marking and cutting of glass quality on-line detection system.

关 键 词：玻璃缺陷检测 BP算法 特征提取 模式识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]