Fatigue Detecting Based On Facial Visual Features

As the process of urbanization continues to accelerate,transportation has become an indispensable part of people's lives.However,fast-paced living and working conditions make many people often tired.When people are in a state of fatigue,driving a car can more easily lead to traffic accidents.In recent years,with the continuous increase of traffic accidents caused by fatigue driving,aroused widespread concern.In order to reduce traffic accidents caused by fatigue driving,academics and industry have conducted in-depth research.This paper presents a fatigue detection method based on facial visual features.In this paper,the method of face detection is studied firstly,and the face detection methods based on Adaboost and Faceness Net are compared and analyzed.On this basis,this paper presents a fatigue detection method based on face feature points and a fatigue detection method based on deep learning.The fatigue detection method based on the facial feature points is to establish the state detection model through the location of the facial feature points and the relationship between them.And the state detection model is used to detect the closed state of eyes and mouth in each frame of image.According to the proposed method of fatigue determination,the fatigue state of each frame of image is detected.Taking into account the fatigue is a continuous process,therefore,this article uses the time window to determine the fatigue state of this period of time.The fatigue state of each time window is determined based on the proportional relationship between the number of fatigue frames and the total number of time windows.The fatigue detection method based on deep learning is to detect the closed state of eyes and mouth of the face in each frame of image through an improved SSD object detection method.The process is to detect the different states of the eyes and mouth as different categories.Finally,the fatigue state in each time window is detected according to the fatigue criterion proposed in this paper.In the method of this paper,two kinds of fatigue evaluation criteria are proposed according to the concept of time window: the accuracy rate based on the number of fatigue times;and the time coverage based on time.Through experimental verification and analysis,the proposed method has higher accuracy and time coverage under these two evaluation criteria.