基于膚色和AdaBoost算法的多角度多姿態(tài)人臉檢測(cè).pdf

1、北京交通大學(xué)碩士學(xué)位論文基于膚色和AdaBoost算法的多角度多姿態(tài)人臉檢測(cè)姓名：王一偉申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：電子與通信工程指導(dǎo)教師：郝曉莉201106=i￡立交適厶堂童些亟堂僮論塞△旦墨!＆△g!ABSTRACTABSTRACT：FacedetectionhassignificantlyadVancedinthelastdecadesEspeciallyinupri曲t舶ntalfacedetection，currentworksh

2、aVebeendeVelopedthatareabletoachieVehigIldetectionrate柚dquickspeedwhichhas饑abledtheircommercialapplicationsHoweVer，merearestillmanydeficienciesinmulti—Viewfacedetectionduetothediversityofthemulti—viewfaceSThispaperinVest

3、igatesthemulti—ViewfacedetectionmethodbasedonAdaBoostFacedetectionmethodandskincolordetectionmethodTwofacedetectionmethodsareproposedbasedonskincolordetectionandAdaB00stalgorithnl，whichcanreducethefalseala衄sa11dimproVeth

4、edetectionrateFacedetectionbasedonAdaBoostalgorithmwhichproposedbyViolahasbeenwidelyreco印izedasthemoste伍cientmethodBasedonthisalgorithm，IaddnewHaarfeaturestotrainmulti—Viewfaces鋤plesSkincolordetectioniseasiertoimplementa

5、IldmuchfasterThispaperproposestwokindsofmethodswhichcombineskincolorinfo肌ationwithAdaBoostBecausethesemethodcaIlfastexcludenonfacere垂on，itcanimpr0VeitsdetectionrateandShortendetetctiontimeThemainresearchworkandcontributi

6、onsofthispaperareasfollws：(1)BasedontheAdaBoostalgorithm，addnewHaarfeatureStodesc^bethemulti—viewfaces鋤plesmoreaccurate(2)Selectmorethanl0000imagesfo鋤CAS—PEALRlfacedatabaseofChineseAcademyofScienceSandAm嘶caJlFERETdatabas

7、eThencutalltheimagest020木20Theseimagescomposeamulti—Viewfacesampledatabase(3)UsingtheAdaBoostalgonthmasmecorealgoritllIIl，t】rainmemulti—viewfacesampleS，combinemeskincolordetectiontoeXcludemuc_hnonfacere百oll，finallyachieV

8、ebetterresuhIIltheendof伽spaper’IdolotsofeXp嘶mentstoproVethatmenewalgoritllIllcanreducethefalsealanlls鏟eatlywhileimproVethedetectionrateorshorteIlmedetectiontimeKEYWORDS：塒【ulti—viewface(1etection；AdaBooSt；slCincolorse舯ent