Light-adaptive face registration based on drivers' video

Abstract

Under real driving conditions, the fatigue monitoring system based on drivers' video is highly affected by light environment, which deteriorates the registration of facial information and thus the accuracy of surveillance. This paper, on the basis of AAM (Active Appearance Model)-based face registration method, analyzes the reasons of its failure when lighting conditions change and proposes an improved AAM algorithm. The experiments show that the improved AAM method can well register driver's face under changing lighting environment, and thus improve the accuracy in fatigue state recognition based on drivers' videos.