交通驾驶员脸疲劳驾驶行为优化图像识别  

A Detection Method of Fatigue Driving Behavior for Traffic Drivers Based on Face Image Recognition

作　　者：宫法明[1,2] 

机构地区：[1]北京科技大学计算机与通信工程学院,北京100083 [2]中国石油大学(华东)计算机与通信工程学院,山东青岛266580

出　　处：《计算机仿真》2015年第11期199-202,共4页Computer Simulation

基　　金：国家科技重大专项课题(2011ZX05011-001)

摘　　要：在对驾驶员驾驶图像进行识别的过程中,是根据眼睛等表现出来的明显疲惫特征进行识别的,但是,当前的模式识别方法只以眼睛作为判断目标,仅仅考虑了眼睛开闭唯一的判断因素,无法描述复杂的疲态特征,导致判断过程单一,精度不高。提出利用脸部行为预测技术的驾驶员疲劳驾驶检测方法,首先采用AdaBoost算法对驾驶员脸部进行检测,详细分析整个检测过程。其次求出驾驶员脸部区域图垂直方向的梯度矩阵,并对梯度矩阵进行水平投影,通过驾驶员脸部的结构特征获取眼睛在图像中的相对位置,依据距离对眼睛开闭进行确定。然后根据PERCLOS(单位时间内眼睛闭合时间所占的百分率)测量原理求出驾驶员眼睛各个状态的参数,最后通过各指标和既定阈值的关系来判断驾驶员是否处于疲劳驾驶状态。仿真结果表明,改进方法具有较高的检测精度。A driver fatigue detection method based on facial behavior prediction technology is presented. Firstly, the driver＇s face is tested by using AdaBoost algorithm, and the whole detection process is analyzed. Secondly, the vertical gradient matrix of drivers face area is figured out, the horizontal projection of the gradient matrix is derived, the relative position of eyes in the image is obtained via the structure features of the face, and the open and close behaviors of eyes are determined based on the distance. And then the state parameters of drivers＇ eyes are obtained according to the measuring principle of PERCLOS （per unit time eyes closed time percentage）. Finally, through the relationship of indexes and established threshold, it is to determine whether the driver is in a state of fatigue driving. The simulation results show that the proposed method has high precision.

关 键 词：脸部行为 驾驶员 疲劳驾驶 检测 

分 类 号：R318[医药卫生—生物医学工程]