基于机器视觉玻璃器皿图像中缺陷检测系统设计探究  

Research on the Design of Defect Inspection Image System by Machine Vision in Glassware

作　　者：郑哲来 ZHENG Zhelai(Anhui Zhengcheng Glass Technology Co.,Ltd.,Maanshan 243000,China)

机构地区：[1]安徽正成玻璃科技有限公司,马鞍山243000

出　　处：《玻璃》2025年第2期12-17,共6页Glass

摘　　要：某玻璃器皿企业的生产线需要设计一套自动化的缺陷检测系统,旨在提高产品质量检测效率。为了达到以上目标,研究过程以机器视觉技术为基础,设计了相应的检测系统,其硬件部分包括数码相机、机械手、工业计算机、光源以及称重传感器,软件部分的设计重点为视觉检测算法。该算法运用BP神经网络开展模型训练与缺陷识别,输入参数涵盖缺陷周长、面积、长宽比等。对系统的缺陷检测效果进行验证,结果显示,其对气泡、结石、夹杂物、裂纹四类缺陷的识别准确率分别达到95%、96%、97%、93%。A production line of a glassware enterprise needs to design a set of automatic defect detection system to improve the efficiency of product quality detection.Based on machine vision technology,a corresponding detection system is designed for achieving the above goals.The hardware includes digital cameras,robotic arms,industrial computers,light sources and weighing sensors.The software focuses on visual inspection algorithms.The emphasis of the algorithm is on the model training and defect identification by means of BP neural network,and the input parameters include dimension of the defect,such as perimeter,area,length-to-width ratio,etc.The detection effect of the system is verified,and the results show that the accuracy rate of identification of four kinds of defects,such as bubbles,stones,inclusions and cracks,reached 95%,96%,97%and 93%respectively.

关 键 词：机器视觉 玻璃器皿 缺陷检测系统 

分 类 号：TQ171[化学工程—玻璃工业]