Identification, Based On The Integration Of Driver Fatigue

Driver fatigue is one of the chief causes of traffic accident. Hence,it is significant to detect driver's fatigue status and decrease accidend rate. In this paper, we present a method of driver fatigue detection which combine multiple visual cues as follows: ERCLOS, AECS, GAZEDIS, PERSAC,NodFreq and YawnFreq.The main work of this paper is:1) The basic frame of driver fatigue detection system which combine multiple visual cues is presented at first.2) In order to implement precised location of eyes, the thesis proposes an algorithm combined with grayness projection and block complexity, and use a kalman particle filter based on color and texture for eyes tracking, improve the precision of tracking.3) Proposing extraction methods of fatigue parameters: Gabor filter is used for extracting texture features from image of eyes; Then the RBF neural network is introduced to estimate opening degrees of the eyes; And In order to extract gaze features, the Generalized Regression Neural Networks is used for gaze position prediction; Finally a method based on image segmentation is used for getting the size of mouth.4) Fuzzy neural network is used for dealing with uncertainty of human fatugue generation. And Bagging algorithm is introduced to improve the precision of the classifer.All of features are based on ocular measures that makes the image gather simple, all the things needed is a CCD camera. We have performed some experiments to make sure that our method works effective. Experiment results show that our method yields a much more robust and accurate results than using a single feature.