Vision-based distraction analysis tested on a realistic driving simulator

This paper presents a non intrusive approach to obtain driver´s face pose estimation based on stereo gray-level image processing. Face pose estimation is based on an automatic and incremental 3D model creation and its correct tracking. From this information, gaze focalization in the scene is calculated in order to detect driver distraction. Different distraction activities are inferred in a realistic simulator and a study of the incidence of these distracting activities in the driver´s behaviour is carried out. Some experimental results and conclusions are presented.