A Novel Approach for Real Time Eye State Detection in Fatigue Awareness System

摘要

This paper proposes a novel eye state detection approach to construct an efficient real time driver fatigue awareness system with an ordinary webcam. Eye state detection has given big challenges to researchers as eye block takes only a small part of input image and can show at various appearances for its flexibility. Moreover, light illumination and viewpoint changes cause more confusions and difficulties for PC to robustly extract eye structure such as contours and iris circles. We transfer this tough problem to a classification problem by combining a discriminative feature, namely Color Correlogram, with machine learning method (Standard Adaboost in this paper). The novelty of this work is that we can efficiently and robustly detect eye states in real time with a single ordinary webcam, even in somewhat harsh conditions such as certain lighting changes, head rotation and different objects. Experimental evidence supports this method well and human fatigue conditions are simultaneously measured based on eye states.