CAJ | 학술논문

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。
목전,재기우궤기시각적부법파리질량재선감측계통중,수선수요대파리도상진행채집,대기진행처리화분석,이진행파리결함식별,연후용우후속조작중。본문근거대부법파리각류결함적특정연구,통과분석반향전파학습신경망락식별산법적우결점,채용일충개진방식용우부법파리적결함분류화자부식별。경과실험수거측시표명,개진후적산법구유수렴속도쾌、식별오판솔저적특점,능구위부법파리질량재선검측계통중적파리타표화절할등조작제공준학、유효적신식,대우감소인공오차、보증산품질량등방면구유중요적의의。
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.