A Rapid Facial-Direction Detection System Based on Symmetrical Measurement

This study presents a method of detecting driver’s facial direction, which is necessary to judge the timing of alert and control the vehicle speed for a Front Collision Avoidance Assistance System. The method is robust against various facial and lighting conditions using three features that are robust against variety. One is a histogram of the facial vertical edges that exist on the faces of almost all people. The second is a White-Black-White edge, which extracts a pair of upward and downward horizontal edges. The edges particularly emphasize facial features patterns like the eyes and mouth. The third is symmetry of facial features: a face is nearly symmetrical around a vertical center axis. These features are apparent in low resolution images. Using them, a rapid facial-direction detection system can be realized which is robust against facial variety and lighting conditions.