基于双眼定位与状态判决疲劳检测算法  

Driver's fatigue detection according to double eyes positioning and status judgment

作　　者：唐美霞[1] 何勇[2] TANG Mei-xia1, HE Yong2(1. School of Information Engineering, Nanning College for Vocational Technology, Nanning 530008, China;2. School of Computer Sciences and Engineering, Hunan University of Science and Technology, Yueyang 414000, Chin)

机构地区：[1]南宁职业技术学院信息工程学院,广西南宁530008 [2]湖南科技大学计算机科学与工程学院,湖南岳阳414000

出　　处：《计算机工程与设计》2018年第6期1750-1754,1787,共6页Computer Engineering and Design

基　　金：2017年度广西高校中青年教师基础能力提升基金项目(2017KY1032)

摘　　要：提出一种基于计算机视觉的驾驶员疲劳状态检测方法,依据驾驶员双眼状态来辨别驾驶员是否疲劳。采用一种时空约束的Adaboost方法,快速检测驾驶员视频中的面部区域;在先验知识确定的可能眼睛区域,采用Haar-like特征和Adaboost分类器快速定位双眼区域;采用卷积神经网络的LeNet5网络架构,训练眼睛状态分类器和检测双眼状态,依据双眼是否闭合的特性判别驾驶员是否疲劳。实验结果表明,该方法能够可靠检测驾驶员的疲劳状态,检测效率高。A driver’s fatigue state detection method based on computer vision was presented,according to the state of driver’s eyes to identify whether the driver is tired or not.A space-time constraint Adaboost method was used to quickly detect the face area from the driver’s video.The Haar-like feature and the Adaboost classifier were used to quickly locate the regions of two eyes in the possible eye regions determined by apriori knowledge.The LeNet5 network architecture of convolution neural network was used to train the eye state classifier and detect the eye state.Whether the driver was tired or not was determined depending on whether the eyes were closed or not.Experimental results show that the method can reliably detect the fatigue state of the driver and the detection efficiency is high.

关 键 词：大数据 疲劳检测 计算机视觉 卷积神经网络 HAAR-LIKE特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]