基于YOLOv8的铝电解电容外观缺陷检测方法  

Appearance Defect Detection Method of Aluminum Electrolytic Capacitor Based on YOLOv8

作　　者：李泽沁 赵子荣 盛磊 曾良涛 姜丽 谭德立 LI Zeqin;ZHAO Zirong;SHENG Lei;ZENG Liangtao;JIANG Li;TAN Deli(School of Computer Science,Neusoft Institute Guangdong,Foshan 528225,China)

机构地区：[1]广东东软学院计算机学院,广东佛山528225

出　　处：《现代信息科技》2024年第20期55-60,共6页Modern Information Technology

基　　金：广东省大学生创新创业训练项目(S202312574017);广东大学生科技创新培育专项资金(pdjh2024a469)。

摘　　要：传统铝电解电容质检依赖人工目视,以识别外壳划痕、破损等缺陷,这种检测方法存在检测准确性和效率的问题。为解决这一问题,提出并实现了一种基于YOLOv8的电容缺陷检测方案。通过构建电容缺陷数据库,并在YOLOv8模型中进行凹陷、划痕、破损等缺陷的训练,方案成功实现了电容外观缺陷检测功能。实验数据结果显示,该模型的mAP@50超过87%。与传统检测方法相比,基于YOLOv8的电容缺陷检测方案具有更高的准确性,效率更高。进一步构建电容外观缺陷数据库,可以提高检测准确性和效率,为电容工业生产的缺陷检测提供了可行的解决方案。Traditional aluminum electrolytic capacitor quality inspection relies on manual visual inspection to detect defects such as scratches and damages on the casing,and this detection method encounters problems with accuracy and efficiency.To address the problems,this paper proposes and implements a capacitor defect detection scheme based on YOLOv8.By constructing a capacitor defect database and training the YOLOv8 model on defects such as dents,scratches,and damages,the scheme successfully achieves capacitor appearance defect detection function.Experimental data results show that the model's mAP@50 exceeds 87%.Compared to traditional detection methods,the capacitor defect detection scheme based on YOLOv8 has higher accuracy and efficiency.Further construction of a capacitor appearance defect database can enhance detection accuracy and efficiency,providing a feasible solution for defect detection in capacitor industrial production.

关 键 词：缺陷检测 YOLOv8 深度学习 卷积神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]