基于机器视觉的PCB表面缺陷检测研究综述  

A review of research on PCB surface defect detection based on machine vision

作　　者：徐一奇 肖金球[1,2,3] 汪俞成 顾逸韬 赵红华 XU Yiqi;XIAO Jinqiu;WANG Yucheng;GU Yitao;ZHAO Honghua(College of Electronics and Information Engineering,Suzhou University of Science and Technology,Suzhou 215009,China;Intelligent Measurement and Control Engineering Technology Research Center,Suzhou 215009,China;College of Physical Science and Technology,Suzhou University of Science and Technology,Suzhou 215009,China)

机构地区：[1]苏州科技大学电子与信息工程学院,江苏苏州215009 [2]苏州市智能测控工程技术研究中心,江苏苏州215009 [3]苏州科技大学物理科学与技术学院,江苏苏州215009

出　　处：《微电子学与计算机》2025年第4期1-15,共15页Microelectronics & Computer

基　　金：国家住房与城乡建设部项目(341111601);江苏省住房和城乡建设厅项目(34173164);江苏省研究生创新工程基金(KYCX21_3010)。

摘　　要：印刷电路板(Printed Circuit Board,PCB)是几乎每种电子产品中必备的组件,其优劣直接影响了电子产品的质量。随着集成电路和半导体技术的快速发展,PCB也趋于精小化。因此,对PCB中的缺陷进行高精度和快速检测成为了一大挑战。对PCB的各种缺陷检测方法进行了分析研究,详细讨论了传统的基于图像处理、基于机器学习和基于深度学习的缺陷检测方法,对它们的算法性能,优点和局限性进行比较,总结了PCB缺陷检测领域当前面临的挑战并展望未来缺陷检测的研究趋势。Printed Circuit Boards(PCBs)are essential components in nearly all electronic products,and their characteristics significantly influence the quality of these products.With the rapid advancement of integrated circuit and semiconductor technologies,PCBs have increasingly become more compact.Consequently,achieving high-precision and efficient defect detection in PCBs has emerged as a critical challenge.This study analyzes and evaluates various defect detection methods for PCBs,with a detailed examination of traditional approaches based on image processing,machine learning,and deep learning.The algorithmic performance,strengths,and limitations of these methods are compared,and the current challenges in PCB defect detection are summarized.Additionally,this work forecasts future research trends in the field of PCB defect detection.

关 键 词：PCB 缺陷检测 机器视觉 深度学习 

分 类 号：TN407[电子电信—微电子学与固体电子学] TN41