摘要
芯片缺陷检测是其品控的关键环节,关乎产品质量与良率。对近年来基于传统的机器视觉和基于深度学习的芯片缺陷检测方法进行梳理与分析,介绍了对封装前的芯片表面缺陷和封装体存在的印刷缺陷与引脚缺陷检测的相关研究。其中,对于深度学习缺陷检测方法,根据数据标签的不同,将其分为全监督学习、无监督学习和其他方法3大类进行归类介绍。此外,详细分析了芯片表面缺陷特性,以期为相关研究人员提供有效参考。
Chip defect detection is a key link in its quality control,which is related to product quality and yield.This paper reviews and analyzes the chip defect detection methods based on traditional machine vision and deep learning in recent years,and introduces the related research on the detection of chip surface defects before packaging,printing defects and pin defects after packaging.According to the different data labels,the deep learning defect detection methods are divided into three categories:supervised learning,unsupervised learning and other methods.Finally,the characteristics of chip surface defects are analyzed in detail,in order to provide an effective reference for relevant researchers.
作者
王新宇
蒋三新
WANG Xinyu;JIANG Sanxin(School of Electronic and Information Engineering,Shanghai Electric Power University,Shanghai 201306)
出处
《现代制造技术与装备》
2022年第5期94-98,共5页
Modern Manufacturing Technology and Equipment
关键词
芯片
缺陷检测
机器视觉
深度学习
卷积神经网络
chip
defect detection
machine vision
deep learning
convolutional neural network
