Research On Driver Fatigue Detection Based On Multi-feature Integration

With the development of society and economy, increasing motor vehicles, the personal safety attract more attention than ever before in the whole world.Driver fatigue has become one of the main cause of the accidents.Therefore, many countries are actively doing research on the fatigue driving, especially in the western developed countries where the study is more comprehensive in this regard.In today's society, driving fatigue detection is critical.In this paper, the driver fatigue detection system is divided into four parts:face detection,face tracking, feature extraction of the human eye and mouth and driver fatigue identification.Face detection plays a huge role in the driver fatigue detection system, a new method of face detection based on color detection and PCA is presented in this paper. Firstly, it uses color detection to locate candidate regions and then uses the PCA to detect face. Experiments show that this method may greatly reduce the number of the resemble face region and improve the efficiency and accuracy of face detection, besides performs better in different light conditions.In the process of tracking, it uses Kalman filter to track human faces so that it improves the speed of detection and accuracy of locating.Histogram-based threshold segmentation is used to detect the mouth region and locate the mouth. The priori knowledge is userd to determine the scope of eye and obtain the eye feature point by using edge extraction. And then a new method of correcting the inaccuracies of the eye's feature points is presented based on the change of the slope of the eye's feature points.Finally, The BP neural network is used for fatigue detection according to multi-feature information of obtained eyes and mouth and campare with the method based PERCLOS principle.