基于眼睛状态多特征融合的疲劳驾驶检测  

Fatigue driving detection based on multi-feature fusion of eye state

作　　者：任俊 魏霞[1] 黄德启[1] 刘栋 REN Jun;WEI Xia;HUANG De-qi;LIU Dong(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047

出　　处：《计算机工程与设计》2022年第11期3187-3194,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(51468062)。

摘　　要：为解决驾驶状态中光照及头部姿势变化等因素对眼睛状态检测影响的问题,提出一种基于多特征融合的眼睛状态识别方法。采用级联回归树算法定位人脸特征点,利用欧拉角对人脸特征点校正后得到人眼的纵横比特征、根据人眼二值图像得到累积黑色素差值特征以及人眼水平投影高宽比特征,在此基础上提出融合这3个特征并使用支持向量机分类器进行眼睛状态识别,根据筛选机制以及PERCLOSE准则判别疲劳状态。实验结果表明,该算法疲劳检测准确率为97.05%,可以检测多种姿态下的眼睛状态,满足实时性的要求。To solve the problem of the influence of illumination and the change of head posture on eye state detection in driving state,an eye state recognition method based on multi-feature fusion was proposed.Cascade regression tree algorithm was used to locate facial feature points,Euler angle of face feature points was used for correction to get aspect ratio characteristics of the human eye,accumulated melanin difference characteristics according to the human eye binary image and horizontal projection ratio characteristics.These three characteristics were fused and the support vector machine classifier was used for eye state reco-gnition.The fatigue state was determined according to the screening mechanism and PERCLOSE criterion.Experimental results show that the fatigue detection accuracy of the algorithm is 97.05%,which can detect the state of eyes in a variety of postures and meet the real-time requirements.

关 键 词：特征点检测 眼部状态识别 支持向量机分类器 多特征疲劳检测 眼睛筛选机制 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]