Abstract: In view of the lack of current intelligent fatigue driving monitoring products, a design scheme based on machine vision theory is proposed. When fitting the eye contour, due to the high complexity of the algorithm of ellipse based on random voting mechanism, a modified method is proposed to reduce the time complexity of the algorithm, by narrowing the range of fitting points. When one judges the fatigue state, the PERCLOS method can not take into account the accuracy and real-time problem. By modifying the parameters appropriately, the method of continuous frame analysis is presented, so that the driver’s state can be distinguished, and the real-time performance is guaranteed. Experimental results show that the algorithm improves the accuracy and real-time performance of the system compared with the previous algorithm.

Key words: rotating target method, voting mechanism, normalization, continuous frame analysis