疲劳驾驶监测系统核心算法的研究与实现
发布时间：2019-05-29 03:31
【摘要】：驾驶员疲劳驾驶是引发交通事故的一个重要原因。为了减少疲劳驾驶引发的交通事故,可以设法在驾驶员进入疲劳状态时及时给驾驶员提醒。为了达到此目的需要一套实时准确的疲劳驾驶监测系统。人们已提出了许多疲劳驾驶监测方法。在这各种方法中,基于图像处理的监测算法是重要的一类。但是因为人脸本身的复杂性,以及外部环境的复杂性,使得各种基于图像处理的算法的实时性和鲁棒性等仍然有大问题。人脸关键点定位算法是人脸相关的图像处理任务常用的基础算法。该类算法在人脸识别和表情识别中已有很多应用,而在疲劳驾驶监测中应用较少且应用不够充分。因此,本文将以人脸关键点定位算法为重点,同时研究并实现疲劳驾驶监测系统核心算法的其他三个子模块。本文主要完成了如下工作:(1)构建建模和测试所需的数据库。因为所有的具有人工标注的人脸数据库都缺少含有疲劳相关的面部信息的人脸图像,所以我们通过整合多个现有数据库并加入额外采集的人脸图像制作了针对疲劳驾驶监测的数据库。(2)研究了混合模型算法用于疲劳相关信息获取。本文介绍了ASM、AAM、STASM、CLM四种主流人脸关键点定位算法的基本原理。主要从人脸关键点定位和人脸局部状态信息获取两个角度出发对这四种算法进行了对比实验,对比了四种算法在不同人脸点集上的定位效果,并分析总结出各种算法的性能特点。以此实验的基础上,给出了混合人脸关键点定位算法,并对可行性进行了进一步的实验分析。(3)设计疲劳驾驶监测系统核心算法的四个子模块:人脸检测、图像增强、人脸关键点定位和疲劳判定。人脸检测模块以基于Ada Boost的人脸检测算法为核心。图像增强模块以去除光照干扰为主要目的。人脸关键点定位模块以混合定位算法为基础。疲劳判定模块以眼部为例,采用了PERCLOS疲劳判定准则。(4)针对疲劳驾驶监测的需要和特性,优化各个子模块的性能,最后将各个子模块组合为一个整体。特别是通过充分利用关键点定位算法的跟踪能力,在保证算法准确性的前提下,提高了算法的速度。我们实现了整个核心算法,进行了模拟测试。从实验结果看,该系统具有较好的性能。
[Abstract]:Driver fatigue driving is an important cause of traffic accidents. In order to reduce the traffic accidents caused by fatigue driving, we can try to remind the drivers when they enter the fatigue state. In order to achieve this goal, a real-time and accurate fatigue driving monitoring system is needed. Many fatigue driving monitoring methods have been put forward. Among these methods, the monitoring algorithm based on image processing is an important one. However, due to the complexity of face itself and the complexity of external environment, there are still great problems in the real-time and robustness of various algorithms based on image processing. Face key point location algorithm is a common basic algorithm for face related image processing tasks. This kind of algorithm has been widely used in face recognition and expression recognition, but it is less used and insufficient in fatigue driving monitoring. Therefore, this paper will focus on the face key point location algorithm, and study and implement the other three sub-modules of the core algorithm of fatigue driving monitoring system. The main work of this paper is as follows: (1) build the database needed for modeling and testing. Because all face databases with manual tagging lack face images that contain fatigue-related facial information, Therefore, we make a database for fatigue driving monitoring by integrating several existing databases and adding additional face images. (2) the hybrid model algorithm is studied to obtain fatigue related information. This paper introduces the basic principles of four mainstream face key point location algorithms in ASM,AAM,STASM,CLM. This paper mainly compares the four algorithms from two angles of face key point location and face local state information acquisition, and compares the localization effects of the four algorithms on different face point sets. The performance characteristics of various algorithms are analyzed and summarized. On the basis of this experiment, the hybrid face key point location algorithm is given, and the feasibility is further analyzed. (3) four sub-modules of the core algorithm of fatigue driving monitoring system are designed: face detection, image enhancement, Face key points location and fatigue decision. The face detection module takes the face detection algorithm based on Ada Boost as the core. The main purpose of image enhancement module is to remove light interference. The face key point location module is based on the hybrid localization algorithm. Taking the eye as an example, the PERCLOS fatigue criterion is adopted in the fatigue judgment module. (4) according to the needs and characteristics of fatigue driving monitoring, the performance of each sub-module is optimized, and finally, each sub-module is combined into a whole. Especially, by making full use of the tracking ability of the key point location algorithm, the speed of the algorithm is improved on the premise of ensuring the accuracy of the algorithm. We implement the whole core algorithm and carry on the simulation test. The experimental results show that the system has good performance.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;U463.6;TP391.41
本文编号：2487603
[Abstract]:Driver fatigue driving is an important cause of traffic accidents. In order to reduce the traffic accidents caused by fatigue driving, we can try to remind the drivers when they enter the fatigue state. In order to achieve this goal, a real-time and accurate fatigue driving monitoring system is needed. Many fatigue driving monitoring methods have been put forward. Among these methods, the monitoring algorithm based on image processing is an important one. However, due to the complexity of face itself and the complexity of external environment, there are still great problems in the real-time and robustness of various algorithms based on image processing. Face key point location algorithm is a common basic algorithm for face related image processing tasks. This kind of algorithm has been widely used in face recognition and expression recognition, but it is less used and insufficient in fatigue driving monitoring. Therefore, this paper will focus on the face key point location algorithm, and study and implement the other three sub-modules of the core algorithm of fatigue driving monitoring system. The main work of this paper is as follows: (1) build the database needed for modeling and testing. Because all face databases with manual tagging lack face images that contain fatigue-related facial information, Therefore, we make a database for fatigue driving monitoring by integrating several existing databases and adding additional face images. (2) the hybrid model algorithm is studied to obtain fatigue related information. This paper introduces the basic principles of four mainstream face key point location algorithms in ASM,AAM,STASM,CLM. This paper mainly compares the four algorithms from two angles of face key point location and face local state information acquisition, and compares the localization effects of the four algorithms on different face point sets. The performance characteristics of various algorithms are analyzed and summarized. On the basis of this experiment, the hybrid face key point location algorithm is given, and the feasibility is further analyzed. (3) four sub-modules of the core algorithm of fatigue driving monitoring system are designed: face detection, image enhancement, Face key points location and fatigue decision. The face detection module takes the face detection algorithm based on Ada Boost as the core. The main purpose of image enhancement module is to remove light interference. The face key point location module is based on the hybrid localization algorithm. Taking the eye as an example, the PERCLOS fatigue criterion is adopted in the fatigue judgment module. (4) according to the needs and characteristics of fatigue driving monitoring, the performance of each sub-module is optimized, and finally, each sub-module is combined into a whole. Especially, by making full use of the tracking ability of the key point location algorithm, the speed of the algorithm is improved on the premise of ensuring the accuracy of the algorithm. We implement the whole core algorithm and carry on the simulation test. The experimental results show that the system has good performance.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;U463.6;TP391.41
【相似文献】
相关期刊论文 前5条
1 衣晓,何友,关欣;基于不同局部模型的联合滤波算法研究[J];中国惯性技术学报;2002年05期
2 赵启林,孙宝俊,江克斌;非局部损伤模型的客观性研究[J];工程力学;2003年05期
3 李庆良;雷虎民;;一种基于即时学习的多模型在线建模方法[J];四川大学学报(工程科学版);2010年01期
4 欧阳永金;;厦门市仙岳路上跨金尚路立交桥牛腿应力的有限元分析[J];铁道科学与工程学报;2006年01期
5 ;[J];;年期
相关会议论文 前2条
1 李孟国;;工程潮流数值模拟中局部模型及其应用[A];第五届全国水动力学学术会议暨第十五届全国水动力学研讨会文集[C];2001年
2 孙伟平;冯庆;向金海;余胜生;;基于局部模型的视频目标跟踪[A];第八届和谐人机环境联合学术会议（HHME2012)论文集NCMT[C];2012年
相关重要报纸文章 前1条
1 刘自国;“三条黄河”全面进入实施阶段[N];中国水利报;2003年
相关硕士学位论文 前1条
1 董庆然;疲劳驾驶监测系统核心算法的研究与实现[D];电子科技大学;2015年
，本文编号：2487603
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2487603.html
