Study On The Eye Detection And Tracking In Driver Fatigue Monitoring System

Fatigue Driving is regarded to be an important reason of traffic accidents. In order to reduce these traffic accidents, driver's fatigue level has been estimated according to his eyes state in this thesis. Eye detection and tracking methods of the driver have been studied and improved importantly.This thesis mainly on the following aspects:Firstly, a new method has been proposed in this thesis that has been used to detect driver's eyes. In this method, image processing and SVM were conjoined. Before two eyes are detected precisely, an eye pair is located. The structure of the eye region is considered as a stable and robust feature which can help distinguish eye pair from other patterns. Initially, the facial image is binaried and edge-detect to obtain the eye candidates. Then, all the eye candidates are sent to an SVM classifier that verified the candidates and obtains the real eye pairs. Finally, the real eye pair is located according to the verification results. The method is superior in detecting speed and veracity.Secondly, novel eye tracking method has been proposed that has conjoined Kalman filter and Mean Shift, the whole tracking process has been divided into three stages: first, the eye model is constructed which bases on eye's location from the initial frame, Kalman filter is used to estimate eye position then, and Mean Shift is used to correct eye position at last. This eye tracking method is effective for realistic lighting condition with the less than 15 angle's head rotation.