Research And Implementation Of Fatigue Driving Detection Algorithm Based On Facial Features

With the continuous development of road traffic,cars have gradually become an indispensable tool of travel.Although cars bring convenience to people’s travel,they bring some negative effects to traffic safety.According to the survey,incidence of traffic accidents caused by cars has been increasing in recent years,and many of these accidents are caused by the behavior of the fatigue driving.Therefore,it is necessary to research and implement an efficient and accurate fatigue driving detection system to keep from traffic accidents caused by fatigue driving.As an important part of the body,face contains abundant feature information.When the drivers are fatigue,eye closing time and blinking frequency are different from normal driving condition.And drivers may yawn when they are fatigue.So this paper mainly researchs fatigue driving detection based on facial features.The main work is illustrated as follows:Based on the face tracking algorithm of KCF,a face tracking algorithm based MCKCF have researched and designed,which improves the accuracy of face tracking.KCF only uses a single HOG feature,which decreases the accuracy of face tracking in complex environments.Moreover,KCF algorithm needs to manually mark the tracked target in the initial frame.And when the target disappears and comes back again,KCF algorithm may not be able to retrieve the target immediately,or even missing the target.Therefore,this paper uses multi-scale CNN features and MTCNN to optimize KCF.Then a face tracking algorithm based on MC-KCF is designed and implemented.Based on the face key point detection model of DCNN,this paper researchs and designs a face key point detection model based on BL-DCNN.This model can obtain 20 face key points,and the accuracy and speed of detection are improved.The face key points detection model based on DCNN can only obtains 5 face key points.Although the location of key facial features such as eyes and mouth can be obtained,the size of eyes and mouth can not be determined.And the detection speed of DCNN is slow.Therefore,this paper optimizes the detection process and network structure of DCNN based on Bottleneck layer in MobileNet model.Then a face key points detection model based on BL-DCNN is designed and implemented.Finally,based on the detection result of face key points,the key features of face such as eyes and mouth are located.Based on the location result of key facial features such as eyes and mouth,a fatigue driving detection method combining the blinking frequency,the time of eyes closing and yawning is proposed.The accuracy of fatigue driving detection is improved.For recognition of the eye state,this paper proposes an eye state recognition model based on CNN,which combines the eye angle for auxiliary recognition.Moreover,for recognition of the yawning,this paper uses face key points of near the mouth to calculate the duration of mouth opening to determine.This paper designs and implements a fatigue driving detection system based on facial features.These proposed and improved algorithms are integrated into this system.Finally,the method and system proposed in this paper are verified by experiments.