Fatigue Driving Detection Based On Blink Detection And Pupil Localization

In recent years,the rapid development of road traffic has brought about a threat to people's lives in China.Fatigue driving is one of the most important reasons in many traffic accidents.Therefore,detecting whether the driver is in a state of fatigue has important research value and social significance for preventing fatigue driving and reducing traffic accidents.In fact,the state of the human eyes changes inevitably when the driver is in a state of fatigue.For this reason,this paper focuses on this issue.Firstly,the face detection and the human eyes positioning are studied and analyzed.The cascade classifier of AdaBoost algorithm is used for face detection and the method of gray image projection is used to locate the human eyes.Then,on this basis,this paper focus on the PERCLOS(Percentage Of Eyelid Closure Over The Pupil Overtime),the frequency of eye blinking and the pupil size.In order to judge the state of eyes effectively,a method of blink detection that based on the combination of boundary extraction and dynamic threshold segmentation is proposed.At the same time,the contour of the pupil is extracted by the improved method of star-ray,and then the improved method of least-squares ellipse fitting is adopted.The process of selecting the edge sample points is simplified,the exact matching of the pupils is completed,and the accurate extraction of the pupil position and size is achieved.Finally,the various face databases were sorted out,analyzed,filtered,and part of the image data was collected by myself,form a certain scale of test library.The research result verified that this paper can be accurately and efficiently by using Microsoft Visual Studio 2010 and OpenCV function library.The completion of blink detection and pupil positioning that can be applied for detecting fatigue driving.