Research On Driver Fatigue Detection Based On Face Micro-motion Features

Fatigue driving is one of the main factors causing traffic accidents,so fatigue detection and early warning of drivers is one of the research hotspots in the field of automobile safety assisted driving.this paper studies the driver fatigue detection method based on the video image,and the fatigue state of the driver is judged by combining the fretting information of the driver's eyes,mouth and head.The main contents of the study are as follows:(1)The improved SSH(Single Stage Headless Face Detector)face detection algorithm is studied.In order to improve the speed of SSH on driver's face detection,the influence of different basic networks on SSH performance is studied,and finally,VGG16 is used as the basic network;the method of two-layer detector is used to improve the speed of detection.The experimental results show that the improved SSH model can satisfy the detection of driver's facial image.(2)The driver facial tracking algorithm based on the combination of Cam Shift and Kalman is studied.the detected driver face image is tracked by Cam Shift face tracking algorithm,and then the driver face window is predicted by Kalman filtering.Finally,the advantages and disadvantages of face detection algorithm and tracking algorithm are compared by experiments.The experimental results show that the driver's facial tracking algorithm has good accuracy and real-time performance.(3)The extraction algorithm of fatigue features based on face key points,i.e.,SDM(Supervised Descent Method)face feature point localization method,is studied.Firstly,the ASM(Active Shape Model)face feature point localization algorithm and the AAM(Active Appearance Models)face feature point localization algorithm are studied,and the specific implementation process of these algorithms is expounded respectively.Then,the shortcomings of ASM and AAM are analyzed by experiments.Finally,a more accurate SDM algorithm is used to locate the key points of the driver's face.(4)Based on the key point position information of face,this paper presents a multi-information fusion method for driver fatigue and measurement.Firstly,the driver's eye fatigue detection method is studied by using the principle of PERCLOS(Percentage of Eyelid Closure Over the Pupil Over time),then the driver's mouth information is used to study the driver's mouth fatigue detection method by using the principle of PERYAW(Percentage of Yaw).Then using other key points of the head information,using the POSIT(Pose from Orthography and Scaling with Iteration)principle to estimate the driver's head posture information,by analyzing the driver nod frequency to study the driver fatigue.Finally,the driver's condition is judged by combining three criteria.The experiments show that the detection methods with multiple information indexes can have better detection accuracy.