基于改进YOLOv5的驾驶员疲劳状态检测  

Driver fatigue state detection based on improved YOLOv5

作　　者：李浩 马晓 周万珍[1] LI Hao;MA Xiao;ZHOU Wanzhen(School of Information Science and Engineering,Hebei University of Science and Technology,Shijiazhuang,Hebei 050018,China;Supervision Office,Cangzhou Open University,Cangzhou,Hebei 061001,China)

机构地区：[1]河北科技大学信息科学与工程学院,河北石家庄050018 [2]沧州开放大学监察室,河北沧州061001

出　　处：《河北工业科技》2024年第1期17-26,共10页Hebei Journal of Industrial Science and Technology

基　　金：河北省自然科学基金(F2018208116)。

摘　　要：为了减少因疲劳驾驶而造成的意外交通事故,提出了一种基于改进的YOLOv5网络模型,对驾驶员的疲劳状态进行检测。首先,使用轻量型网络MobileNetV3替换原YOLOv5主干网络;其次,在颈部网络各个C3模块中融入ECA注意力机制;最后,通过检测网络对眼睛的开合度和嘴巴有无打哈的状态进行定位和识别,使用多指标对驾驶员进行疲劳判定,并自建疲劳检测数据集进行实验。结果表明:改进的YOLOv5模型参数量、计算量、体积分别减小至原模型的48%、38%、50%,解决了原模型参数量、计算量、体积过大的问题;mAP值由98.6%提升至99.1%,精确率由95.9%提升至96.8%,检测速率由115 f/s提升至119 f/s,进一步提高了模型的检测精度和检测速度。改进的YOLOv5模型具备轻量化、高精度、高速率的特点,可为疲劳驾驶预警提供参考。In order to reduce the sudden traffic accidents caused by fatigue driving,a network model based on improved YOLOv5 was proposed to detect the fatigue state of drivers.Firstly,the original YOLOv5 backbone network was replaced by the lightweight network MobileNetV3.Secondly,the ECA attention mechanism was incorporated into each C3 module of the neck network.Finally,the degree of eyes opening and closing and mouth with or without snorting were located and identified by the detection network,and then the multiple indexes were used to judge the fatigue state of the driver,and the self-built fatigue detection dataset was used for experiments.The results show that the improved YOLOv5 model′s number of parameters,computations,and volume are reduced to 48%,38%,50%of the original model,respectively,which solves the problem of excessive number of parameters,computations and volume of the original model.The mAP value is increased from 98.6%to 99.1%,the accuracy is increased from 95.9%to 96.8%,and the detection rate is increased from 115 f/s to 119 f/s,all of which further improves the detection accuracy and speed of the mode.The improved YOLOv5 model has the characteristics of lightweight,high precision and high speed,which can provide reference for fatigue driving early warning.

关 键 词：计算机图像处理 YOLOv5 MobileNetV3 ECA 轻量化 疲劳驾驶 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]