發(fā)布時(shí)間：2019-05-29 03:31

【摘要】：駕駛員疲勞駕駛是引發(fā)交通事故的一個(gè)重要原因。為了減少疲勞駕駛引發(fā)的交通事故,可以設(shè)法在駕駛員進(jìn)入疲勞狀態(tài)時(shí)及時(shí)給駕駛員提醒。為了達(dá)到此目的需要一套實(shí)時(shí)準(zhǔn)確的疲勞駕駛監(jiān)測系統(tǒng)。人們已提出了許多疲勞駕駛監(jiān)測方法。在這各種方法中,基于圖像處理的監(jiān)測算法是重要的一類。但是因?yàn)槿四槺旧淼膹?fù)雜性,以及外部環(huán)境的復(fù)雜性,使得各種基于圖像處理的算法的實(shí)時(shí)性和魯棒性等仍然有大問題。人臉關(guān)鍵點(diǎn)定位算法是人臉相關(guān)的圖像處理任務(wù)常用的基礎(chǔ)算法。該類算法在人臉識(shí)別和表情識(shí)別中已有很多應(yīng)用,而在疲勞駕駛監(jiān)測中應(yīng)用較少且應(yīng)用不夠充分。因此,本文將以人臉關(guān)鍵點(diǎn)定位算法為重點(diǎn),同時(shí)研究并實(shí)現(xiàn)疲勞駕駛監(jiān)測系統(tǒng)核心算法的其他三個(gè)子模塊。本文主要完成了如下工作:(1)構(gòu)建建模和測試所需的數(shù)據(jù)庫。因?yàn)樗械木哂腥斯?biāo)注的人臉數(shù)據(jù)庫都缺少含有疲勞相關(guān)的面部信息的人臉圖像,所以我們通過整合多個(gè)現(xiàn)有數(shù)據(jù)庫并加入額外采集的人臉圖像制作了針對(duì)疲勞駕駛監(jiān)測的數(shù)據(jù)庫。(2)研究了混合模型算法用于疲勞相關(guān)信息獲取。本文介紹了ASM、AAM、STASM、CLM四種主流人臉關(guān)鍵點(diǎn)定位算法的基本原理。主要從人臉關(guān)鍵點(diǎn)定位和人臉局部狀態(tài)信息獲取兩個(gè)角度出發(fā)對(duì)這四種算法進(jìn)行了對(duì)比實(shí)驗(yàn),對(duì)比了四種算法在不同人臉點(diǎn)集上的定位效果,并分析總結(jié)出各種算法的性能特點(diǎn)。以此實(shí)驗(yàn)的基礎(chǔ)上,給出了混合人臉關(guān)鍵點(diǎn)定位算法,并對(duì)可行性進(jìn)行了進(jìn)一步的實(shí)驗(yàn)分析。(3)設(shè)計(jì)疲勞駕駛監(jiān)測系統(tǒng)核心算法的四個(gè)子模塊:人臉檢測、圖像增強(qiáng)、人臉關(guān)鍵點(diǎn)定位和疲勞判定。人臉檢測模塊以基于Ada Boost的人臉檢測算法為核心。圖像增強(qiáng)模塊以去除光照干擾為主要目的。人臉關(guān)鍵點(diǎn)定位模塊以混合定位算法為基礎(chǔ)。疲勞判定模塊以眼部為例,采用了PERCLOS疲勞判定準(zhǔn)則。(4)針對(duì)疲勞駕駛監(jiān)測的需要和特性,優(yōu)化各個(gè)子模塊的性能,最后將各個(gè)子模塊組合為一個(gè)整體。特別是通過充分利用關(guān)鍵點(diǎn)定位算法的跟蹤能力,在保證算法準(zhǔn)確性的前提下,提高了算法的速度。我們實(shí)現(xiàn)了整個(gè)核心算法,進(jìn)行了模擬測試。從實(shí)驗(yàn)結(jié)果看,該系統(tǒng)具有較好的性能。
[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.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：U495;U463.6;TP391.41

【相似文獻(xiàn)】

相關(guān)期刊論文 前5條

1 衣曉,何友,關(guān)欣;基于不同局部模型的聯(lián)合濾波算法研究[J];中國慣性技術(shù)學(xué)報(bào);2002年05期

2 趙啟林,孫寶俊,江克斌;非局部損傷模型的客觀性研究[J];工程力學(xué);2003年05期

3 李慶良;雷虎民;;一種基于即時(shí)學(xué)習(xí)的多模型在線建模方法[J];四川大學(xué)學(xué)報(bào)(工程科學(xué)版);2010年01期

4 歐陽永金;;廈門市仙岳路上跨金尚路立交橋牛腿應(yīng)力的有限元分析[J];鐵道科學(xué)與工程學(xué)報(bào);2006年01期

5 ;[J];;年期

相關(guān)會(huì)議論文 前2條

1 李孟國;;工程潮流數(shù)值模擬中局部模型及其應(yīng)用[A];第五屆全國水動(dòng)力學(xué)學(xué)術(shù)會(huì)議暨第十五屆全國水動(dòng)力學(xué)研討會(huì)文集[C];2001年

2 孫偉平;馮慶;向金海;余勝生;;基于局部模型的視頻目標(biāo)跟蹤[A];第八屆和諧人機(jī)環(huán)境聯(lián)合學(xué)術(shù)會(huì)議（HHME2012)論文集NCMT[C];2012年

相關(guān)重要報(bào)紙文章 前1條

1 劉自國;“三條黃河”全面進(jìn)入實(shí)施階段[N];中國水利報(bào);2003年

相關(guān)碩士學(xué)位論文 前1條

1 董慶然;疲勞駕駛監(jiān)測系統(tǒng)核心算法的研究與實(shí)現(xiàn)[D];電子科技大學(xué);2015年

，

本文編號(hào)：2487603