基于深度卷积神经网络的电子玻璃缺陷分类方法  

Electronic glass defect classification method based on deep convolutional neural network

作　　者：李苑[1] 于浩 金良茂 曹志强 陈家睿 郑际杰 韩高荣[1] 刘涌[1] LI Yuan;YU Hao;JIN Liangmao;CAO Zhiqiang;CHEN Jiarui;ZHENG Jijie;HAN Gaorong;LIU Yong(Zhejiang University,Hangzhou 310058;Bengbu COE Technology Co.,Ltd.,Bengbu 233030;Innovation Center for Advanced Glass Materials(Anhui)Co.,Ltd.,Bengbu 233060)

机构地区：[1]浙江大学,浙江杭州310058 [2]蚌埠中光电科技有限公司,安徽蚌埠233030 [3]玻璃新材料创新中心(安徽)有限公司,安徽蚌埠2330060

出　　处：《中国建材科技》2024年第S01期17-23,共7页China Building Materials Science & Technology

基　　金：国家重点研发计划(2022YFB3603300)。

摘　　要：电子玻璃是信息显示产业的关键基础材料之一。近年来,显示产业向大尺寸化、超高清和轻薄化发展,对于电子玻璃基板的质量提出了更高的要求。机器视觉检测具有速度快、精度高、成本低、稳定性好等优点,被广泛应用于各种工业场景中。图像处理算法、识别分类算法是机器视觉检测的关键技术。本文针对基于深度卷积神经网络的整图分类方法在电子玻璃表面缺陷检测领域的应用,从图像数据处理、卷积神经网络构建、训练调参、评价标准等方面介绍其研究进展,并总结部分应用实例,对电子玻璃缺陷分类未来的研究方向进行展望。Electronic glass is one of the critical foundational materials in the information display industry.In recent years,the display industry has developed towards large-sized,ultra-high definition,and lightweight,putting higher requirements for the quality of electronic glass substrates.Machine vision inspection has the advantages of fast speed,high accuracy,low cost,and good stability,and is widely used in various industrial scenarios.Image processing and recognition classification algorithms are vital technologies in machine vision detection.The present work focuses on applying deep convolutional neural network-based whole-image classification methods in the field of electronic glass surface defect detection.It introduces the research progress from image data processing,convolutional neural network construction,training parameter tuning,and evaluation standards,and it summarizes some application examples.It also looks forward to future research directions for electronic glass defect classification.

关 键 词：电子玻璃 机器视觉 深度卷积神经网络 缺陷分类 

分 类 号：TQ171[化学工程—玻璃工业]