Abstract: Driver behavior surveillance and warning system plays an important role in reducing the traffic accidents.Monitoring and analyzing the fatigue and attention state by using the machine vision is one of hotspots in the assistant safe driving system.Machine vision affected by light condition is very serious，even impact on the reliability of the image processing algorithms.In special lighting conditions of low-light and bright light，the monitoring idea for fatigue state of the driver is introduced firstly.And then the algorithm is proposed that based on purkinje spots and gray variance locates the face of a driver，using the location of the extremums in the projective curve divides the facial organs into separate areas，obtains the location of the eye’s contours，and using PERCLOS to determine the mental state of the driver.Based on the result of the location of the face and organs of the driver，analyzing the rotary motion of the head，an algorithm for calculating the facial rotary angle of drivers is proposed and then a strategy is obtained used for judging whether the driver’s attention is disturbed or not.Practice proves that the algorithm proposed acts in real time and obtains excellent effect in special lighting conditions of low-light and bright light，which provide more accurate basis for analyzing the driver’s attention state.

Key words: safety driving assistant, machine vision, fatigue state, attention state

摘要： 驾驶人驾驶行为监测及预警系统对于提高行车安全性及降低交通事故等问题具有重要作用，而利用机器视觉对驾驶人疲劳状态及注意力状态进行监测和分析是安全辅助驾驶领域内的研究热点之一。机器视觉受光照条件影响非常严重，甚至直接影响到图像处理算法的可靠性。在弱光和强光两种特殊照明条件下，介绍了驾驶人疲劳状态的检测思路，提出利用普尔钦斑点及投影曲线极点位置分割面部器官独立区域，最终获得眼睛的轮廓状态，利用PERCLOS判断驾驶人的疲劳状态。在驾驶人面部及面部器官定位的基础上，对驾驶人的头部旋转运动进行分析，提出了计算驾驶人头部旋转角度的计算方法，以驾驶人头部旋转角度为依据判断驾驶人的注意力是否分散。实践证明，在弱光和强光两种特殊照明条件之下，算法实时性好，准确率较高，效果非常理想，能够为驾驶人疲劳状态及注意力状态分析提供较为准确的依据。

关键词: 安全辅助驾驶, 机器视觉, 疲劳状态, 注意力状态