疲劳驾驶检测中人眼实时定位与跟踪  

Eye Location and Tracking for Fatigue Driving Detection

作　　者：杨智勇[1] 蒲亦非[1] 巩晓倩 周激流[1] 

机构地区：[1]四川大学计算机学院,四川成都610065

出　　处：《微电子学与计算机》2014年第8期115-118,122,共5页Microelectronics & Computer

基　　金：国家自然科学基金(60972131);四川省科技支撑计划项目(2012RZ0005);法中科学与应用基金(FFCSA);四川大学青年教师科研启动基金(2012SCU11036);四川省重大科技成果转化项目(2012SC0043);四川省科技创新苗子工程资助项目(20132032)

摘　　要：针对疲劳驾驶检测中现存的一些难点,提出了一种新颖的驾驶员眼睛定位与跟踪方法.首先采用Gamma校正和图像自商算法对红外光源采集到的驾驶员行车图像进行预处理,进而采用基于Haar-like特征的AdaBoost算法初定位驾驶员眼眉区域,并在此基础上采用眼眉区域15点ASM(Active Shape Models)模型来准确定位驾驶员的眼睛区域,最后采用了Unscented卡尔曼滤波算法完成驾驶员眼睛的实时跟踪.实验结果表明,该算法不仅可以提高对光照、姿态变化以及驾驶员眼睛运动的强非线性问题的鲁棒性,同时对驾驶员配戴眼镜的情况也能得到较为理想的处理结果.In view of some of the difficulties existing in fatigue driving detection,this paper proposes a novel method for drivers' eyes location and tracking.Firstly we uses Gamma and SQI(Self Quotient Image)algorithm to process images,which obtained by infrared light,then locate the brow area by AdaBoost algorithm,and on this basis,with the brow area 15 points ASM(Active Shape Models)model to accurately locate driver's eye area,finally the unscented kalman filter algorithm is adopted to drivers' eyes tracking.The experimental results show that the algorithm introduced can not only improve the robustness of the illumination,pose transformation,and driver's eye's strongly nonlinear movement,but also get ideal reault,even if the driver wearing glasses.

关 键 词：ADABOOST ASM 人眼跟踪 UNSCENTED卡尔曼滤波 疲劳驾驶 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]