基于改進YOLOv8n的PCBA外觀缺陷檢測

打開文本圖片集

中圖分類號：TN911.73-34 文獻標識碼：A 文章編號：1004-373X（2025）17-0176-05

DOI：10.16652/j.issn.1004-373x.2025.17.026 引用格式：，.基于改進YOLOv8n的PCBA外觀缺陷檢測[J].現(xiàn)代電子技術(shù)，2025，48（17）：：176-180.

PCBA surfacedefectdetection based on improved YOLOv8n

ZHOU Zhiwei，HAN Bin

SchoolofInformationandControlEngineering，SouthwestUniversityofScienceandTechnologyMianyang6Chir

Abstract：Thedesignsof modernPCBA（printed circuit boardasembly）arecomplex，andtheirdensitiesare getting higher andhigher.Thedistancesbetweencircuitsandcomponentsaregetingcloser.Thishigh-densitylayoutresultsinmoretypesand smallrrangesofdefectsonPCBA，increasing thedificultyof defectdetection.ImprovingtheLSKmoduleinYOLOv8nby combininglargekernelandsmallkernelconvolutionscancapturefeaturesofdiferentscales，therebyimprovingthereliabilityof appearance defect detection.Inviewof this，aPCBAappearance defectdetection method basedonimproved YOLOv8nis studied.FirstlyeferigtotenetworkstructureofSlim-neck，theetworkstructureofNeckisimproedtoachievelighweight. Secondly，theintroductionofLSK modulesenhances themeanaverage precision （mAP）reduced due tothe lightweightnetwork structure.Then，byintroducingtheSEmodule，thenetworkstructureoftheHeadisimprovedtofurtherenhancethedetection performanceof themodel.Finally，theMPDIUlossfunctionisintroduced toenhance theabilityofsmallojectdetection.The experimentalresultsshowthatthemAPoftheimproved modelproposed inthispaperreaches95.2%onthePCBAappearance defect dataset， increasing by 3.1% in comparison with that of YOLOv8n. The validity of the proposed model is verified.

Keywords：PCBA;appearance inspection;defect detection; improved YOLOv8n; lossfunction;lightweightnetwork

0 引言

印制電路板（PCB）作為電子元器件安裝和電路電氣連接的載體，是各類型電子產(chǎn)品的重要組成部分，提高其生產(chǎn)制造水平，對于計算機、消費電子、通信設(shè)備等行業(yè)發(fā)展有著至關(guān)重要的作用。（剩余5216字）