基于UKF非线性人眼跟踪的驾驶员疲劳检测  

Driver Fatigue Detection Based on Unscented Kalman Filter and Eye Tracking

作　　者：张祖涛[1,2] 张家树[1] 

机构地区：[1]西南交通大学机械学院,四川成都610031 [2]西南交通大学信息科学与技术学院,四川成都610031

出　　处：《西南交通大学学报》2008年第6期697-702,共6页Journal of Southwest Jiaotong University

基　　金：教育部新世纪优秀人才支持计划资助项目(NCET-05-0794);西南交通大学机械工程学院青年教师基金(MYF0806)的资助

摘　　要：为解决驾驶员疲劳检测算法中头部快速移动、人眼非线性跟踪以及实际疲劳表情的识别问题,提出了一种新的基于UKF眼跟踪算法的驾驶员疲劳检测方法.根据近似非线性函数的概率分布比近似其函数更容易的原则,利用UT无迹变换,选择一组确定的Sigma点集逼近驾驶员人眼运动状态的后验概率密度函数,进行人眼非线性跟踪.在驾驶员人眼非线性跟踪基础上,通过计算PERCLOS值,进行现实驾驶条件下驾驶员疲劳的跟踪检测.实验结果表明,该方法不仅可以增强对驾驶员头部旋转、快速移动以及光照变换的鲁棒性,而且可以比传统的Kalm an滤波算法提供更精确的计算估计.In order to resolve the problem of fast head moving, nonlinear eye tracking and facial fatigue expression detection, a new scheme of driver fatigue detection was proposed based on unscented Kalman filter （UKF） and eye tracking. Owing to the intuition that it is easier to approximate a probability distribution than to approximate an arbitrary nonlinear function or transformation, nonlinear eye tracking can be achieved using unscented transformation （UT） by adopting a set of deterministic sigma points to match the posterior probability density function for eye movement. Driver fatigue can be detected by calculating PERCLOS （percentage of eyelid closure over the pupil over time） under a realistic driving condition after nonlinear eye tracking. The experimental results show that the proposed scheme can not only improve the robustnesses of the head rotating and fast head movement of a driver and the interference of external illuminations, but also provide more accurate estimation than the traditional Kalman filter.

关 键 词：眼跟踪 疲劳检测 UKF滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]