手机表面缺陷的机器视觉检测方法研究进展  

Research progress of mobile phone surface defect detection based on machine vision

作　　者：吴一全[1] 庞雅轩 WU Yiquan;PANG Yaxuan(School of Electronic and Information Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)

机构地区：[1]南京航空航天大学电子信息工程学院,江苏南京211106

出　　处：《智能系统学报》2025年第1期33-51,共19页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(61573183).

摘　　要：智能手机在现代人们的学习、工作与生活中扮演着十分重要的角色,手机的大批量生产给手机表面(手机屏幕玻璃盖板、手机外壳)缺陷检测工作提出了更高的要求,而基于机器视觉的检测方式能够更加快速准确地实现对手机表面缺陷的检测。以该领域面临的挑战为思路,总结了近10年来基于机器视觉的手机表面缺陷检测的研究进展。首先列举了手机表面存在的典型缺陷,并分析了机器视觉应用于手机表面缺陷检测工作中面临的部分难题,其中包括算法的精度、实时性、鲁棒性3个方面;然后分别针对上述问题的改进方法进行了分析与对比;进一步总结了目前可供使用的手机表面缺陷数据集及算法的性能评价指标;最后根据手机表面缺陷检测领域面临的问题进行了总结与展望。Nowadays,smartphones play an important role in our learning,working,and daily lives.Mass production of smartphones has raised higher requirements for defect detection on the phone surface,including the glass cover and phone shell.Machine vision-based detection methods can achieve faster and more accurate detection of surface defects on smartphones.Taking the challenge in this field as a guide,this paper concludes the research progress of machine vision-based smartphone surface defect detection over the past decade.First,typical defects on the phone surface are listed,and some challenges faced by machine vision in smartphone surface defect detection are analyzed,including algorithm accuracy,real-time performance,and robustness.Then,improvement methods for the above problems are analyzed and compared.In addition,available datasets for smartphone surface defect detection and performance evaluation metrics for algorithms are summarized.Finally,a summary and outlook are provided based on the challenges faced in the field of smartphone surface defect detection.

关 键 词：机器视觉 缺陷检测 手机屏幕玻璃盖板 手机外壳 深度学习 数据集 性能评价指标 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]