基于人脸特征点分析的疲劳驾驶实时检测方法  

Real-time Fatigue Driving Detection based on Analysis of Facial Landmarks

作　　者：柳龙飞 伍世虔[1] 徐望明[2] LIU Longfei;WU Shiqian;XU Wangming(School of Machinery and Automation,Wuhan University of Science and Technology,Wuhan 430081,China;Sciool of Infortndtion,icience and Engineering,Wuhan University of Sciencc and Technology,430081)

机构地区：[1]武汉科技大学机械自动化学院,湖北武汉430081 [2]武汉科技大学信息科学与工程学院,湖北武汉430081

出　　处：《电视技术》2018年第12期27-30,55,共5页Video Engineering

基　　金：国家自然科学基金面上项目(61775172);武汉科技大学2017年研究生创新创业基金项目(JCX2017016)

摘　　要：为了有效监测驾驶员是否疲劳驾驶、避免交通事故的发生,提出了一种利用人脸特征点进行实时疲劳驾驶检测的新方法。对驾驶员驾驶时的面部图像进行实时监控,首先使用AdaBoost算法检测人脸,并利用ERT算法定位人脸特征点;然后根据人脸眼睛区域的特征点坐标信息计算眼睛纵横比EAR来描述眼睛张开程度,根据合适的EAR阈值可判断睁眼或闭眼状态;最后基于EAR实测值和EAR阈值对监控视频计算闭眼时间比例(PERCLOS)值度量驾驶员主观疲劳程度,将其与设定的疲劳度阈值进行比较即可判定是否疲劳驾驶。实验结果验证了所提出方法的有效性。In order toeffectivelydetect fatigue driving and avoid trafficaccidents,a new real-time fatigue driving detection approach using facial landmarlcs is proposed.Firstly,when the driver's face is monitored by acamera in real time,the faceregion is detected using AdaBoost algorithm and the facial landmarks are located using ERT algorithm.Then the value of Eye Aspect Ratio(EAR)is calculated from the locations of such facial landmarks in the eye regions to characterize the eye-open degree,and a suitable EAR thresliold is obtainedto judge the eyes are open orclose.Finally,the value of PERCLOS is calculated to measure the fatigue degree accordingto the values of EAR and EARthreshold from the monitoringvideo,andis compared witha given fatigue-degree threshold to determine whether thedriveis fatigue or not.Experimentresults validatesthe effectiveness of the proposed approach.

关 键 词：人脸特征点 疲劳驾驶 EAR PERCLOS 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]