基于视频图像信息提取的疲劳驾驶检测技术研究

【作者】 朱婧；

【导师】 薄煜明；

【作者基本信息】 南京理工大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 机动车辆与日俱增,随之而来的车辆交通安全问题也越来越受到社会的广泛关注。调查表明,疲劳驾驶在造成交通事故的危险因素中高居第三位,在死亡交通事故原因中居首位。因此,研制疲劳驾驶检测预警系统,对于避免交通事故,提高交通安全性有着重要意义。本文在研习了国内外相关研究的基础上,分析对比各种疲劳检测方法,深入研究了基于视觉的非接触式、实时的驾驶疲劳检测方法。以视频图像信息提取和处理理论为基础,研究探讨了基于视觉的疲劳驾驶检测的关键技术,包括人脸检测、眼睛定位与跟踪、眼睛特征提取与状态识别、疲劳状态分析。首先采用了实时性较好的基于肤色检测的人脸定位方法,将肤色与非肤色区域进行分离,较为准确地定位人脸区域;在关键环节眼睛定位中,选择了积分投影函数与混合投影函数相结合的方法,精确定位了眼睛中心,同时将Kalman滤波与MeanShift算法相结合的眼睛跟踪方法融合到实验中,实现了对序列视频中头部旋转或倾斜情况下的连续眼睛检测和定位;在此基础上对眼睛进行特征提取,主要分析了决定眼睛状态的关键特征,包括虹膜、眼角、上眼睑,并对传统眼睛模型进行了修正,利用上眼睑与下眼睑的高度差来判断眼睛的状态;最后选取目前公认最有效的疲劳分析方法PERCLOS法,结合眨眼频率对疲劳状态进行分析。基于上述研究所实现的疲劳检测软件系统,能够实时地检测驾驶者的眼睛状态并进行疲劳识别,在实验室环境下获得了较好的实验效果。更多还原

【Abstract】 With the increasing number of motor vehicles, the problem of traffic safety has become more and more concerned by the society. Research has shown that fatigued driving is one of the most dangerous factors which cause death in traffic accidents. Therefore, a system for monitoring the driver’s level of vigilance and alerting the driver when he is fatigued is very important to prevent traffic accidents .Based on the researches at home and abroad, various methods of fatigue detection are analysed and compared. A non-contact, real-time method based on vision is studied in-depth. Key technologies of the method are presented, including face detection, eyes location and tracking, feature detection and fatigue recogonition.First of all, a real-time algorithm of face detection based on color is used to separate the color region from non-color region. So the face region is located accurately. Then, in the key process of eyes location, grayness projection algorithm is selected. The integral projection function and hybrid projection function are combined to acquire the eye’s center accurately. And a real-time eye tracking algorithm combined with Kalman Filter and Mean Shift is integrated into the experiment to obtain a good effect of eyes tracking.On the above basis, some features of eyes are detected, including the irises, the eye corners and the upper eyelid. And a new eye model is proposed in the system. According to the model, the height of upper eyelid is detected to show the eyes’ status.Finally, the PERCLOS method which is accepted as the most effective method of fatigue analysis currently is selected. The Eye blink frequency is also picked up to analyse the eyes’ status.The software system based on the above technologies can detect the driver’s eyes and recognize his fatigue in real time and the experiment result is very good in the laboratory environment.更多还原