Abstract
This paper presents a system to discriminate driver's posture using face tracking for driver support. A camera is arranged near a room mirror of a car, since the setting does not conflict with the law and is easy to implement. A driver's face region is firstly detected by Template Matching. The region is tracked using color information. Cam-Shift Tracking or Mean-Shift Tracking is used for the face tracking. Finally, driver's pasture is discriminated using the location of the face in images. The area for the discrimination is set up empirically. The effectiveness of the system is evaluated by experiments for a driver wearing glasses, a cap, or a mask.
