摘要
基于机器视觉的驾驶人疲劳检测系统,通过对驾驶人眼睛动作的分析实现对驾驶人疲劳状态的估计。眼睛区域的准确定位是保证疲劳检测精度的前提条件。然而,实际行车过程中,驾驶人头部姿态随机、快速变化会造成眼睛区域定位精度的严重下降。该文在基于主动形状模型(ASM)算法实现驾驶人眼睛区域粗定位的基础上,针对ASM模型在实际检测过程中的姿态适应性较低与定位精度不高的问题,提出局部ASM模型来增强ASM算法的姿态适应性;进一步引入平均合成精确滤波器(ASEF)算法与ASM算法相结合的思路提高对眼睛区域的定位精度;同时,提出单、双眼相结合的ASEF算法来提高眼睛虹膜中心定位的鲁棒性。实验结果表明:该算法对于驾驶人头部姿态变化具有较强的适应性,能够实现眼睛区域的准确定位。
Driver drowsiness estimates can be realized by analyses of the drivers＇eye movements based on a machine vision system.However,the system requires accurate eye region recognition in the driver＇s facial image.Random,rapid changes of the head posture complicate locating the eye region in real driving scenarios.The active shape model（ASM）can be used to coarsely locate the eye region.This study uses a local ASM model to enhance the head posture adaptability of the ASM algorithm.Then,the average of synthetic exact filters（ASEF）algorithm and the ASM are combined to improve the eye region location precision.A single eye ASEF and a double eyes ASEF are integrated to more robustly identify the iris center location.Tests show that the algorithm has strong head posture adaptability and can robustly and accurately identify the eye region location.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2014年第6期756-762,共7页
Journal of Tsinghua University(Science and Technology)
基金
交通运输部信息化科技项目(2012-364-835-110)
关键词
疲劳驾驶
机器视觉
眼睛定位
driver drowsiness
machine vision
eye location
