Locating Eye and Nose Features Precisely in IR Video Sequences for Predicting Driver Fatigue

It is important to detect the driver fatigue at early stage in order to prevent accidents caused by tired operation. This paper presents a non-intrusive computer vision system for real-time monitoring driver's fatigue, which is based on active remote infrared (IR) illumination for real time eye tracking. In the system, using the image captured by IR camera, a more efficient method-luminance weight contrasting circle template with distance divisive is developed to detect the bright-pupils accurately in a short time. And a 3D face model with features of eyes and nostrils is constructed helping to predict and track the positions of eyes. The system can be equipped in the panel of a car and work well in most lighting conditions. Some valuable experimental results and some statistic data about the system are shown.