基于TMS320C6711DSP的實(shí)時(shí)人臉識別系統(tǒng)研究及實(shí)現(xiàn).pdf

1、上海大學(xué)碩士學(xué)位論文基于TMS320C6711DSP的實(shí)時(shí)人臉識別系統(tǒng)研究及實(shí)現(xiàn)姓名：閆娟申請學(xué)位級別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：孫鑫程武山20070201上海大學(xué)碩士學(xué)位論文ABSTRACTFacerecognitionisbasedOllitsowndatabaseofinformationtoidentifythefundamentalinter部texistsinthefaceofstaticimagesorsequ

2、encesofimagesFaceComparedwithotherbiologicalcharacteristics，theapplicationoffacerecognitionsystemsarenoncontactcontinuouslyandrealtimen血istomakeitmoredirectfriendlyandeasy1V[oreoverdifferentfromthetraditionalfacekeycardf

3、orgeryand出e丘isnoteasyanditwillbeevenhardertobelostandthereforesafetyhigherreliabilityFacerecognitiontechnologymamredhasatremendousprospectinthesafetycertificationsystemsmature，intelligentmanmachineinterface，videoconferen

4、cingcreditcardverific撕onandsoonThereforefacerecognitionhasnowbecomeahotfieldincomputervisionpatternrecognitionandartificialintelligenceresearchThispaperstudiesthemanagementofrealtimefacerecognitionsystemaimedattheentrarl

5、ce，inordertosatisfytherequirementsoftheprocessinganddevicesizeDSPisselectedbythesystemtodevelopanddesignnlethreemainparts—positioningoffacerecognitionarefacedetectionandtreatmentfacialfeatureextraction，andfacerecognition

6、matchingtheAnaLysmandResearchAccordingtothesystem’srequirement，thereasonablearithmeticandschemeareprovidedwhicharetestedandvalidatebyexperimentationviathesystemthefollowingsarcthemainjobandinnovationofthearticle：Thefast，

7、aimatthecharacteristicofthesystem，realtime，dealingwithamountofdataandthecomplexoffacerecognition，accordingtothespecialtyofDSPcomposeaseriesofvideocollectionandmanagementsystemissetupbasedonthechipOfTMS320C6711DSEthesyste

8、musingCCDc鋤eratocollectdatausingICETEKC671IⅡKMimageprocessorboardtocollectandprocesstheimage，whichcomposedthehardwareplatformoffacerecognitionsystemThesecond，tosatisfytherequirementofrealtimeandveracityofthesystem，ainnov

9、ativearithmeticarebroughtoutwhichisconsistofdifferenceframeofthemovingvideowiththecomplexionmodel，thisarithmeticisusedtoroughlyinspectthefaceareaseparatethecomplexionareofmovingpartialfromthebackgroundThethird，analyzeand

10、transacttheroughinspectionareaoffacebasedollmathematicsmorphologic，eliminatesimilarcomplexionareaandnobhumanfacearea，putforwardainnovatehumaneyesorientationmethodtodetecttheeye’slocation，accordingtothehumanapparatusdistr

11、ibutingrulestobuilduphumanfacearearestrictionconditiontolocatethefacewellandtrulythenthefacerecognitionorientationisfinishedfromrotIghtoaccurateThefinallyproposedaameliorated2Danal如calraethnd(2DPCA)toextractfacecharacter

12、setupafaceimagelibraryusingthesystemwhichisdiscussedinthispaperandvalidateandcomparetheselfbuildfaceimagelibraryandtheORLlibraryby2DPCATheCandDSPassemblelanguageareusedtoprogramarithmeticinthehardwareplatformofthesystemO

13、ver20humanfacesinmylabindifferentconditionswereinspectedandlocated，andsetupthefaceimagelibraryusingtheresultsAndvalidatethe2DPCAbycomparingtheselfbuildfaceimagelibraryandtheORLlibraryobtainedsatisfiedexperimentoutcome，se