Research On Fatigue Detection Based On Eye Image Characteristics

Traffic accident is the most common man-made disasters in modern society,causing huge personal injury and property loss.47% of traffic accidents are caused by fatigue driving.Therefore,fatigue driving has become a social problem which need immediate attention.How to detect and relieve short-term fatigue efficiently and easily by non-contact method has become a hot topic in the current research.Based on the theoretical analysis of eye image,this thesis takes infrared light as the light source,designs the algorithm of real-time fatigue detection,which is based on the characteristics of eye image,and achieves the human fatigue detection system.The entire process is divided into three steps: face detection,human eye tracking and fatigue identification.The first step is the detection of human face.In this thesis,the detection image is acquired by camera and subject to homomorphic filtering,equalization and normalization.In the light of Ada Boost algorithm and the class Haar feature,the strong classifiers are combined into a cascade classifier to get the image target and then detects the face.The second step is the human eye location tracking.The traditional target location methods are of complex computation and do not perform well in real-time aspect and it requires face image with affluent edge information.Based on the detected face from the first step,this thesis improves the Harris corner detection algorithm,which greatly reduces the computational cost.Taking advantages of integrated Kalman filter and particle filter,the human eye pupil can be tracked,which achieves the purpose of accurate positioning and rapid tracking.Finally,the thesis uses the method of calculating the difference value of the eyelids to judge the mental state of the human body.And it considers two parameters,including PERCLOS principle and blink frequency,to judge whether the body is fatigued and the fatigue degree.At last,when the body is detected as fatigued in tired,t DCS(Transcranial Direct Current Stimulation),a safe and convenient method,is been used to improve the excitability of the brain by stimulating the transcranial brain,so as to relieve fatigue in a short term.This thesis has tested the images of video database and volunteers.The results show that the method of this thesis has improved in terms of real-time and accuracy compared with the current mainstream detection algorithm.