人脸检测与跟踪在疲劳驾车中的研究应用

VIP免费

3.0 侯斌 2024-11-19 4 4 3.48MB 70 页 15积分

侵权投诉

摘要

近年来，交通事故发生率急剧增长、数量急剧增多，安全驾车问题俨然成

为人们生活中不容小视的一部分，而由于疲劳驾车引发的交通事故及伤亡人员

数量不计其数，疲劳驾车已成为车辆事故的主要原因之一，因此研究出实时高

效的疲劳驾车检测系统对于保护人们的生命钱财安全及减少车辆事故发生率有

着重大现实意义。

本论文在大量通读国内外文献的基础上，根据实际驾驶环境，开发出一种

基于 OpenCV 的疲劳驾车检测系统，该系统在 Visual Studio2013 平台上借助 Intel

的开源图像处理开发包 OpenCV 开发出来，系统采用罗技摄像头对实验中的人

脸进行实时脸部图像抓拍，采用光照补偿技术、人脸检测技术、眼部定位技术、

眼部跟踪等技术，并借助疲劳状态识别方法判定实验者的疲劳状态。

疲劳驾车检测系统由人脸的检测、眼部的定位、眼部的跟踪、疲劳的判定

四大模块组成。人脸的检测模块提出了一种改进的光照补偿算法对摄像头所采

集的视频帧图像进行光照的补偿预处理并以此消除光照干扰，通过融合改进的

高斯肤色模型及非线性彩色变换肤色模型提出一种 Double 肤色模型用以检测

人脸图像，并对融合算法的收敛性加以验证，证实其用于人脸检测的可靠性。

眼部的定位模块通过粗定位大致定位到视频帧中的人眼，在对其进行细定

位前先是进行眼部的去噪、滤波处理，消除噪声干扰，细定位采用水平直方投

影图定位，对视频中动态眼部定位时不是用定位方法对每一帧视频进行定位，

而是采用先定位初始帧眼睛图像，将其作为模板，再根据跟踪算法对其进行跟

踪来定位后面视频帧中的眼部位置，这样可大大提高疲劳驾车检测系统的眼部

定位速度。

眼部的跟踪模块提出一种 Mean shift 与Kalman 滤波器结合的方法，通过

Kalman 滤波器先预测到眼部图像将其作为 Mean shift 的起始点，再用其在搜索

区域内跟模板颜色特征相匹配，直至搜索到最匹配的目标为止。采用眼部跟踪

技术可很大地提高疲劳驾车检测系统的实时性和检测的快速性。

疲劳的判定模块用眨眼一次的时间间隔跟改进的 PERCLOS 原理两种疲劳

参数来判定眼睛的开与闭状态，通过系统疲劳检测界面实时获取眼睛的视频图

像，通过疲劳检测算法及事先设定的疲劳报警阈值来控制系统是否报警，并通

过界面显示出具体疲劳的时间。

为检验本论文疲劳驾车检测系统的可靠性，采用多组实验对其进行验证，

验证系统中的人脸检测模块的算法在复杂光照与背景下的单人脸、多人脸、视

频人脸图像中的检测效率及算法的鲁棒性及精确性。通过多个不同实验者在佩

戴及不佩戴眼镜下下的眼部定位与跟踪实验验证了眼部定位算法及跟踪算法的

有效性。为验证疲劳驾车检测系统的疲劳判定模块的可靠性，对视频中的 100

帧人脸图像的眼部闭合时间进行记录，根据与事先设定的疲劳预警阈值时间比

较，若超过阈值则发出疲劳警告。

关键词：疲劳驾车 Double 肤色模型水平直方投影眼部跟踪

OpenCV 疲劳报警

ABSTRACT

In recent years, with the rapid development of social economy and the

improvement of people's living standards, there is a sharp increase in the incidence

of traffic accidents and a sharp increase in the number of traffic accidents, safe

driving issue has become a part of people's lives that can’t be overlooked. There is

countless number of accidents and casualties caused by fatigue driving, There is no

doubt that fatigue driving has become the main reason of accidents ,therefore, there

is great practical significance to come up with efficient real-time fatigue driving

detection system for projecting the safety of human's lives and property and reducing

the rate of accident.

On the basis of reading a large number of domestic and foreign articles,

according to the in fact, this paper develops a driver fatigue detection system based

on OpenCV, the system is developed on Visual Studio2013 with the assist of open

source image processing development package OpenCV, the human faces of

experiment will be captured by Luo ji camera, the paper judges the fatigue status of

people using illumination compensation technology ,human face detection

technology, human eyes location technology ,eyes tracking technology and fatigue

status recognition technology.

The fatigue driving detection system consists of four modules- face detection

module, eyes location module, eyes tracking module and fatigue judgment module.

For human face detection module, this paper proposing an improved illumination

compensation algorithm to do the illumination compensation pretreatment for the

video image acquired from webcam to eliminate light interference, proposing a

Double skin model which is integrated by the improved Gaussian skin color model

and nonlinear color transformation skin color model, verifying the astringency of the

integrated algorithm and verifying the reliability of it for detecting human face.

For eyes location module, first through a coarse positioning to generally locate

the human eyes in the video ,do removing noise and filtering process for the eyes to

eliminate noise before the fine location of the eyes. Fine location of the eyes by the

horizontal histograms projection, locate the first frame eyes image and make it as a

template for the later eyes frame images, as the later eyes frames images use the eyes

tracking algorithm to locate rather than use eyes location method like the first frame

image, the tracking algorithm can greatly improve the location speed of the fatigue

driving detection system.

For eyes tracking module, proposing Mean shift combined with the Kalman

filter method, first predicted the eye image by the Kalman filter and use it as a

starting point for Mean shift, and then use Mean shift to search the most matched

color feature to the template in the search area , until find the best matched target.

For Fatigue judgment module, judging the status of eyes by using two fatigue

parameters of the interval of blink for one time and the improved PERCLOS

principle, acquiring the eyes video images by the interface of the fatigue detection

system, use the fatigue algorithm and the threshold of alarm to control whether raise

the alarm, and show the fatigue time.

To verify the reliability of the fatigue driving detection system, this paper

verifies it by multiple sets of experiments, verifying the detection efficient,

robustness and accuracy of algorithm of human face detection module in the system

under the complex illumination and background, the multiple sets of experiments are

single human face, many human faces and video face image, verifying the efficient

of eyes location and eyes tracking algorithm through the different experimenter who

wear glasses and wear no glasses. To verify the reliability of the module of fatigue

judgment, we record the eyes close time of 100 frames human face images,

comparing the time with the preset threshold of fatigue alarm time, if it is greater

than the threshold, giving the alarm of fatigue.

Key words：Fatigue driving, Double skin color module, Horizontal

histograms projection, Eyes tracking, OpenCV, Fatigue alarm

中文摘要

ABSTRACT

第一章绪论................................................................................................................ 1

1.1 课题研究背景及意义...................................................................................... 1

1.1.1 引言........................................................................................................ 1

1.1.2 疲劳驾车的定义及产生原因................................................................. 2

1.2 疲劳驾车检测技术发展概述及国内外研究现状........................................... 3

1.2.1 疲劳驾车检测技术发展概述................................................................ 3

1.2.2 疲劳驾车检测技术国外研究现状......................................................... 4

1.2.3 疲劳驾车检测技术国内研究现状........................................................ 5

1.3 论文的主要内容及创新点.............................................................................. 6

1.3.1 论文的主要内容.................................................................................... 6

1.3.2 论文的创新点........................................................................................ 6

1.4 论文的结构安排.............................................................................................. 7

第二章人脸的检测算法研究与仿真.......................................................................... 8

2.1 人脸的检测概述.............................................................................................. 8

2.1.1 基于统计学习的人脸的检测方法........................................................ 9

2.1.2 基于面部特征的人脸的检测方法...................................................... 10

2.1.3 基于模板匹配的人脸的检测方法...................................................... 10

2.1.4 基于肤色的人脸的检测方法............................................................... 10

2.2 基于肤色模型的人脸的检测算法................................................................ 11

2.2.1 图像的光照预处理.............................................................................. 11

2.2.2 颜色空间的选择.................................................................................. 12

2.2.3 Double 肤色模型 .................................................................................. 14

2.2.4 人脸的检测实验结果分析.................................................................. 17

2.3 本章小节........................................................................................................ 20

第三章人眼的定位算法研究与仿真.......................................................................... 21

3.1 人眼的定位算法概述.................................................................................... 21

3.2 人眼的定位算法............................................................................................ 23

3.2.1 人眼的粗定位...................................................................................... 23

3.2.2 人眼图像二值化................................................................................... 24

3.2.3 眼部图像去噪....................................................................................... 26

3.2.4 人眼的细定位...................................................................................... 27

3.3 目标跟踪算法概述........................................................................................ 29

3.3.1 Mean shift 算法 .................................................................................... 31

3.3.2 Kalman 滤波算法 ................................................................................. 34

3.3.3 基于 Mean-shift 与Kalman 滤波算法的人眼跟踪 ........................... 36

3.3.4 人眼的跟踪结果分析........................................................................... 41

3.4 本章小结........................................................................................................ 42

第四章人眼状态识别及眼部疲劳判定...................................................................... 43

4.1 人眼开闭判定................................................................................................ 43

4.2 人眼部位状态识别特征参数概述................................................................. 44

4.3 本文采用的人眼状态识别方法及多疲劳判定参数的疲劳判定................. 46

4.3.1 人眼状态识别方法.............................................................................. 46

4.3.2 多疲劳判定参数的疲劳判定.............................................................. 48

4.3.3 实验结果分析....................................................................................... 50

4.4 本章小结......................................................................................................... 51

第五章疲劳驾车检测系统的实现.............................................................................. 52

5.1 疲劳驾车检测系统软硬件配置.................................................................... 52

5.1.1 硬件配置............................................................................................... 52

5.1.2 软件配置............................................................................................... 53

5.2 疲劳驾车检测系统模块及流程图................................................................. 53

5.2.1 疲劳驾车检测系统的模块................................................................... 53

5.2.2 系统流程设计图................................................................................... 54

5.2.3 检测系统界面图.................................................................................. 54

5.3 实验结果分析................................................................................................ 57

5.4 本章小结......................................................................................................... 58

第六章总结与展望...................................................................................................... 59

6.1 总结................................................................................................................ 59

6.2 展望................................................................................................................ 59

参考文献...................................................................................................................... 61

在读期间公开发表的论文和承担科研项目及取得成果.......................................... 65

致谢.............................................................................................................................. 66

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

15 积分 4人已下载

立即下载 VIP免费下载

摘要：

摘要近年来，交通事故发生率急剧增长、数量急剧增多，安全驾车问题俨然成为人们生活中不容小视的一部分，而由于疲劳驾车引发的交通事故及伤亡人员数量不计其数，疲劳驾车已成为车辆事故的主要原因之一，因此研究出实时高效的疲劳驾车检测系统对于保护人们的生命钱财安全及减少车辆事故发生率有着重大现实意义。本论文在大量通读国内外文献的基础上，根据实际驾驶环境，开发出一种基于OpenCV的疲劳驾车检测系统，该系统在VisualStudio2013平台上借助Intel的开源图像处理开发包OpenCV开发出来，系统采用罗技摄像头对实验中的人脸进行实时脸部图像抓拍，采用光照补偿技术、人脸检测技术、眼部定位技术、眼部跟踪等...

展开>> 收起<<

人脸检测与跟踪在疲劳驾车中的研究应用.pdf

共70页,预览7页

还剩页未读，继续阅读

人脸检测与跟踪在疲劳驾车中的研究应用

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

推荐作者

热门标签

举报选择: