基于TAC-YOLOv11s的PCB缺陷检测与实例分割算法  

PCB Defect Detection and Instance Segmentation Algorithm Based on TAC-YOLOv11s

作　　者：王欣璐 郑晓亮[2] 来文豪 WANG Xinlu;ZHENG Xiaoliang;LAI Wenhao(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan 232001,China;School of Public Safety and Emergency Management,Anhui University of Science and Technology,Hefei 231131,China)

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001 [2]安徽理工大学公共安全与应急管理学院,合肥231131

出　　处：《湖北民族大学学报(自然科学版)》2025年第1期80-85,共6页Journal of Hubei Minzu University:Natural Science Edition

基　　金：安徽理工大学高层次引进人才科研启动基金资助项目(2021yjrc02)。

摘　　要：针对印刷电路板(printed circuit board,PCB)缺陷目标小导致识别精度低的问题,提出基于三重注意力跨阶段连接-你只看一次版本11小型(triplet attention and cross stage connections-you only look once version 11 small,TAC-YOLOv11s)的PCB缺陷检测与实例分割算法。首先,设计了跨阶段部分连接(cross stage partial connections,CSPC)特征提取网络,增强网络的特征表达能力;其次,增加了小目标分割头(small object segmentation head,SO)模块,提高对小目标的检测和分割能力;然后,加入了三重注意力(triplet attention,TA)机制,增加对小目标的定位和捕获;最后,采用广义交并比(generalized intersection over union,GIoU)损失函数,优化算法性能。结果表明,与原始YOLOv11s算法相比,TAC-YOLOv11s算法的边界框和掩膜精确率分别提升了11.1%和8.2%,50%交并比阈值下的平均精确率均值分别提升了30.4%和34.3%,证明了算法的优越性。TAC-YOLOv11s算法对实现PCB缺陷的高精度检测与分割具有重要意义。To address the issue of low recognition accuracy caused by the small size of targets in printed circuit board(PCB)defects,a detection and instance segmentation algorithm based on triplet attention and cross stage connections-you only look once version 11 small(TAC-YOLOv11s)was proposed.Firstly,a cross stage partial connections(CSPC)feature extraction network was designed to enhance the network′s feature representation capability.Secondly,a small object segmentation head(SO)module was added to improve the detection and segmentation ability for small objects.Thirdly,a triplet attention(TA)mechanism was incorporated to increase the localization and capture of small targets.Lastly,generalized intersection over union(GIoU)loss function was adopted to optimize the performance of the algorithm.The results demonstrated that the TAC-YOLOv11s algorithm improved by 11.1%and 8.2%in bounding box and mask precision,respectively,and the mean average precision with an intersection over union threshold of 50%for bounding boxes and masks increased by 30.4%and 34.3%,respectively,compared to the original YOLOv11s algorithm,thoroughly validating the superiority of this algorithm.TAC-YOLOv11s algorithm signified its importance in achieving high-precision detection and segmentation of PCB defects.

关 键 词：印刷电路板 缺陷检测 实例分割 YOLOv11s 小目标 

分 类 号：TM755[电气工程—电力系统及自动化]