摘要
针对PCB中存在的缺陷对象较小、类型较多、难以识别等问题,拟提出改进YOLOv5算法,以实现对该问题的有效处理。首先引入SEnet,通过学习,实现对各特征信道的重要性的自动提取,提高物体检测的精度。然后将解耦头思想引入到YOLOv5网络中,以提高故障检测的准确率并加速网络的收敛。实验结果显示,改进后YOLOv5算法的mAP@5值达到了92.5%,mAP@0.5:0.95值达到了47.5%,比原始分别提高了4.3%和2%。此外,每一种类型缺陷的精度都有了显著的提高,证明了算法的有效性。
Aiming at the problems that the defect objects in PCB are small,have many types,and are difficult to identify,an improved YOLOv5 algorithm is proposed to realize the effective treatment of this problem.Firstly,SEnet was introduced to automatically extract the importance of each feature channel through learning,and the accuracy of object detection was improved.Then,the decoupling head idea is introduced into the YOLOv5 network to improve the accuracy of fault detection and accelerate the convergence of the network.The experimental results show that the improved YOLOv5 algorithm's mAP@5 value reaches 92.5%,and the mAP@0.5:0.95 value reaches 47.5%,which are 4.3%and 2%higher than the original ones.In addition,the accuracy of each type of defect is significantly improved,which proves the effectiveness of the algorithm.
作者
迟盛元
白岩
孟祥民
Chi Shengyuan;Bai Yan*;Meng Xiangmin(College of Mechanical Engineering,Beihua University,Jilin,China)
出处
《科学技术创新》
2024年第1期106-109,共4页
Scientific and Technological Innovation
基金
吉林省科技发展计划项目(20210203109SF)
吉林省自然科学基金项目(202201ZYTS513)。
