改进YOLOv5s的摩托车头盔佩戴检测算法  

Improved Motorcycle HelmetWearing Detection Algorithm of YOLOv5s

作　　者：张鑫 周顺勇 ZHANG Xin;ZHOU Shunyong(School of Automation and Information Engineering,Sichuan University of Science&Engineering,Yibin 644000,China;Artificial Intelligence Key Laboratory of Sichuan Province,Yibin 644000,China)

机构地区：[1]四川轻化工大学自动化与信息工程学院,四川宜宾644000 [2]人工智能四川省重点实验室,四川宜宾644000

出　　处：《四川轻化工大学学报（自然科学版）》2023年第3期50-58,共9页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

基　　金：四川省科技厅项目(2020YFG0178);四川省科技厅省院校合作项目(2020YFSY0027)。

摘　　要：针对摩托车头盔佩戴检测准确率低和检测速率慢的问题,提出一种基于YOLOv5s的改进摩托车头盔检测算法。首先,在YOLOv5s的多尺度特征检测中增加浅层检测尺度和4倍上采样特征融合结构,以提升检测准确率。其次,引入卷积注意力模块(Convolutional Block Attention Module,CBAM),以提升对聚集目标的关注,有效解决因遮挡、重叠导致的漏检和误检问题。最后,使用MobilenetV3的Block结构替换主干网络及颈部中的瓶颈结构,实现了降低网络参数量的目的。实验结果表明,相较于YOLOv5s算法,改进算法的mAP提高了2.91%,检测速率达到了36 frame/s,在保证较高检测速率的同时检测精度更高,具有一定的应用价值。Aiming at the low accuracy and slow detection rate of motorcycle helmet wearing detection,an improved motorcycle helmet detection algorithm based on YOLOv5s has been proposed.Firstly,a shallow detection scale and a 4-fold upsampling feature fusion structure are added in the multi-scale feature detection of YOLOv5s,to improve the detection accuracy.Secondly,the Convolutional Block Attention Module(CBAM)is introduced to improve the focus on the aggregation target,which effectively solves the problem of missed detection and false detection caused by occlusion and overlap.Finally,the block structure of MobilenetV3 is used to replace the bottleneck structure in the backbone network and neck,achieving the purpose of reducing the amount of network parameters.The experimental results show that,compared with the YOLOv5s algorithm,the mAP of the improved algorithm is increased by 2.91%,and the detection rate reaches 36 frame/s.The detection accuracy is higher with ensuring a higher detection rate,which exhibits certain application value.

关 键 词：头盔检测 YOLOv5s模型 CBAM注意力机制 MobilenetV3网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]