Rapid Detection Method Of Fatigue Driving Based On Custom Face Detector

With the gradual increase of car ownership,traffic accidents are becoming more frequent,resulting in casualties and economic losses,among which fatigue driving is the most important cause of traffic accidents.Usually,before the occurrence of fatigue driving accidents,the driver will show more obvious fatigue characteristics,so it is of great practical significance to study the fast,accurate and effective detection methods of fatigue driving,provide fatigue warning for drivers,and reduce the occurrence of traffic accidents.Road traffic safety has always been the focus of research of experts and scholars around the world,and in many studies on fatigue driving detection,machine vision methods have become the mainstream technology for detecting driver behavior characteristics with the advantages of non-contact,high precision and low cost.In view of the problems of single fatigue features,large detection model and slow speed in the existing literature,this paper proposes a fatigue detection method based on custom face detector,and based on this,it quickly identifies fatigue driving state with multi-feature fusion.The specific description is as follows:Firstly,aiming at the problem that the face detection model is large and affects the detection speed of fatigue driving,a custom face detector that only includes the eyes and mouth is proposed,and the head pose estimation is performed by using the driver’s two-dimensional face key points combined with coordinate system transformation and camera calibration.Then,aiming at the problem of driver fatigue feature extraction,the EAR,MAR algorithm and head pose Euler angle are used to extract multiple fatigue features of the driver’s eyes,mouth and head attitude,and in view of the driver’s possible local occlusion,face occlusion recognition is carried out based on SVM.Aiming at the problem of fatigue characteristics,the fatigue characteristics of eyes,mouth and head posture,and the fatigue characteristics of PERCLOS,blink frequency,eye closing duration,yawning times,head lowering frequency and head abnormal angle were combined to determine the fatigue state of the driver by DS evidence theory and early warning.Finally,the proposed method is simulated on the Raspberry Pi experimental platform.Experiments show that the driver fatigue detection method based on custom face detector proposed in this paper has high accuracy and robustness on fatigue datasets,and can provide good fatigue warning.