基于人脸对齐和多特征融合的人脸识别方法研究-计算机应用技术专业论文.docx

研究生优秀毕业论文Abstract硕士学位论文AbstractWithmerapiddeVelopmentofcomputersciencea11dne觚orkco舢叭micationtecllIlology，personsbegaIltousebiologicalcharactersofhumallforidentification．Biometricreco印itionincludesf．acerecognition，Voicerecognition，finge印rint andpalmprintrecognition，gaitrecognitionaJldsoon．Comparedwithotherbiometrics，f．aCerecognitionhaSwide applicationsinsociaIpublicsecurityanddailylife．facerecognitionhastheadvantageoffeature obtainedeasya11dlittletimecost，so“hasawideapIplicateprospectinsocialpublicsecudtyanddaily1ife．Acompletefacerecognitionsystemincludesthefollowingprocesssteps：facedetection，facealigIunent，f-eatureextl．actionandclaSsification．Amongthem，faceali殍Ⅲlenta11dfea_tureextractionarethetwokeysteps．So，、VevdUresearchthese帆，oproblems．Themaincontributionsofthisdissertationinclude：(1)ASMisthemostrepresentatiVefaceali舀1IIlentmethodbaSedonstatisticalmodel．HoweVert11e f．eaturepointsearch rangeofASMislimited．Tloovercomemisproblem，weproposeaIlimproVedf-aCeali印mentmethodnamedmeilnprovedASMbasedonLBP．1Hsmethodextendsthesearchregiontosearcht11efeanlrepointsa11dusescodingofLBPinsteadofgrayV羽uetobuildthelocaltexturemodel．The experimental resultsdemon蚰natethatttleimproVedASMbaSedonLBPperfonnbetterthanASM．(2)W色mailllyinVeStigatere鲈essionbaSedmemodsforfaceali卸啦ent，includinglinearregressionbasedmethoda11dshapere铲essionbaSedmethod．Wbfirstlystudythedenselinearregressionmemod，sparselinearregressionmethodaIld如llyspaurselinearregressionmetllod．Thus，、^，eevaluateddifI、erentexperimentstoveri母media’erencesofa_bovet11reemethods．Comparedwithre伊essionbaSedmethods，explicitShapere留essionbaSedm甜lodhaSnoneedoffixedshapemodelp娥IrlletersaIldemployscaScadedre铲essortoupdatet11eimtiaJf-aceshapeforlocatingf．aciallandmarks．ShaperegressionbausedmeⅡlodemploysthelinearcombillation ofs锄plestocons仃ajnmesh印erathermanusiIlgPCAtocons臼面nthat．ExperimentaJresuhsdemonst】嘣ethatexplicitsh印eregressionbasedmetllodac王lievesbetterperfo咖a11cethan1inearregressionbasedmethods．(3)WremainlyinVestigate也elocalfeatIlreextmcti