成都理工大学2014届学士学位论文(设计)基于2DPCA的人脸识别算法研究摘要人脸识别技术是对图像和视频中的人脸进行检测和定位的一门模式识别技术,包含位置、大小、个数和形态等人脸图像的所有信息。由于近年来计算机技术的飞速发展,为人脸识别技术的广泛应用提供了可能,所以图像处理技术被广泛应用了各种领域。该技术具有广阔的前景,如今已有大量的研究人员专注于人脸识别技术的开发。本文的主要工作内容如下:1)介绍了人脸识别技术的基础知识,包括该技术的应用、背景、研究方向以及目前研究该技术的困难,并对人脸识别系统的运行过程以及运行平台作了简单的介绍。2)预处理工作是在原始0RL人脸库上进行的。在图像的预处理阶段,经过了图象的颜色处理,图像的几何归一化,图像的均衡化和图象的灰度归一化四个过程。所有人脸图像通过上述处理后,就可以在一定程度上减小光照、背景等一些外在因素的不利影响。3)介绍了目前主流的一些人脸检测算法,本文采用并详细叙述了Adaboost人脸检测算法。Adaboost算法首先需要创建人脸图像的训练样本,再通过对样本的训练,得到的级联分类器就可以对人脸进行检测。4)本文介绍了基于PCA算法的人脸特征点提取,并在PCA算法的基础上应用了改进型的2DPCA算法,对两者的性能进行了对比,得出后者的准确度和实时性均大于前者,最后将Adaboost人脸检测算法和2DPCA算法结合,不仅能大幅度降低识别时间,而且还相互补充,有效的提高了识别率。关键词:人脸识别2DPCA特征提取人脸检测成都理工大学2014届学士学位论文(设计)2DPCAFaceRecognitionAlgorithmBasedonTheResearchAbstract:Facerecognitionisatechnologytodetectandlocatehumanfaceinanimageorvideostreams,Includinglocation,size,shape,numberandotherinformationofhumanfaceinanimageorvideostreams.Duetotherapiddevelopmentofcomputeroperationspeedmakestheimageprocessingtechnologyhasbeenwidelyappliedinmanyfieldsinrecentyears.Thispaper'sworkhasthefollowingseveralaspects:1)Explainedthebackground,researchscopeandmethodoffacerecognition,andintroducedthetheoreticalmethodoffacerecognitionfieldingeneral.2)ThepretreatmentsworkisbasedontheoriginalORLfacedatabase.Intheimagepreprocessingstage,therearethecoloroftheimageprocessing,imagegeometricnormalization,imageequalizationandimagegrayscalenormalizationfourparts.Afterunitedprocessing,thefaceimageisstandard,whichcaneliminatetheadverseeffectsofsomeexternalfactors.3)Allkindsoffacedetectionalgorithmisintroduced,anddetaileddescribingtheAdaboostalgorithmforfacedetection.ThroughtheAdaboostalgorithmtocreateatrainingsample,thenTrainingthesamplesoffaceimage,andobtainingthecascadeclassifiertodetecthumanface.4)ThispaperintroducesthefacialfeaturepointsextractionbasedonPCA,and2DPCAisusedonthebasisofthePCAasaimprovedalgorithm.Performanceiscomparedbetweenthetwo,itisconcludsthattherealtimeandaccuracyofthelatterisgreaterthantheformer.FinallytheAdaboostfacedetectionalgorithmand2DPCAarecombined,whichnotonlycangreatlyreducetherecognitiontime,butalsocomplementeachother,effectivelyimprovetherecognitionrate.Keywords:Facerecognition2DPCAFeatureextractionFacedetection成都理工大学2014届学士学位论文(设计)目录第1章前言........................................................11.1人脸识别的应用和研究背景....................................11.2人脸识别技术的研究方向......................................21.3研究的现状与存在的困难......................................31.4本文大概安排................................................4第2章人脸识别系统及软件平台的配置................................42.1人脸识别系统概况............................................42.1.1获取人脸图像信息.......................................52.1.2检测定位...............................................52.1.3图像的预处理...........................................52.1.4特征提取...............................................62.1.5图像的匹配与识别.......................................62.2OpenCV......................................................62.2.1OpenCV简介............................................62.2.2OpenCV的系统配置......................................72.3Matlab与图像处理............................................8第3章图像的检测定位...............................................83.1引言........................................................83.2人脸检测的方法..............................................83.3Adaboost算法................................................93.3.1Haar特征.............................................103.3.2积分图................................................103.3.4级联分类器............................................11第4章图像的预处理................................................134.1引言.......................................................134.2人脸图像库.................................................134.3人脸预处理算法.............................................144.3.1颜色处理..............................................144.3.2几何归一化............................................154.3.3直方图均衡化..........................................164.3.4灰度归一化............................................184.4本章小结...................................................19第5章图像的特征提取与识别........................................195.1引言.......................................................195.2图像特征提取方法...........................................205.2.1基于几何特征的方法....................................205.2.2基于统计的方法........................................205.2.3弹性图匹配(elasticgraphmatching)....................215.2.4神经网络方法..........................................215.2.5支持向量机(SVM)方法...................................225.3距离分类器的选择...........................................225.4PCA算法的人脸识别..........................................24成都理工大学2014届学士学位论文(设计)5.5二维主成分分析(2DPCA)....................................255.5.12DPCA人脸识别算法....................................255.5.2特征提取..............................................275.5.3分类方法..............................................275.5.4基于2DPCA的图像重构..................................285.6实验分析...................................................28第6章总结与展望.....................................