摘要
针对印刷电路板(PCB)缺陷微小难测问题,提出了一种基于改进YOLOv8s的PCB缺陷检测算法,该算法添加Transformer编码单元,并引入DwConv网络标准卷积,实现了提升实时检测PCB缺陷的精度,实验结果表明,改进之后的YOLOv8s模型,P_(mA)从0.909提升到了0.951,增加了4.2%。通过与其他主流目标检测方法相比,改进YOLOv8算法展示了更好的检测精度。
Aiming at the problem that defects in printed circuit boards(PCB)are tiny and difficult to detect,a PCB defect detection algorithm based on improved YOLOv8s is proposed.This algorithm adds a Transformer encoding unit and introduces the standard convolution of the DwConv network,realizing an improvement in the accuracy of real-time detection of PCB defects.Experimental results show that for the improved YOLOv8s model,the P_(mA) is increased from 0.909 to 0.951,an increase of 4.2 percentage points.Compared with other mainstream object detection methods,the improved YOLOv8 algorithm shows better detection accuracy.
作者
罗青青
舒升
周正贵
方银银
LUO Qingqing;SHU Sheng;ZHOU Zhenggui;FANG Yinyin(Anhui Business College,Wuhu 241002,China;Anhui Province Key Laboratory of Optoelectric Materials Science and Technology,Wuhu 241002,China)
出处
《齐鲁工业大学学报》
2025年第5期38-44,共7页
Journal of Qilu University of Technology
基金
安徽省自然科学研究重点项目(KJ2021A1483)
安徽省高校自然科学重点项目(2023AH052295,2024AH050530)
芜湖市第二批科技计划项目(2024kj040)。
