基于膚色模型與神經(jīng)網(wǎng)絡(luò)集成的人臉檢測(cè).pdf

1、上海大學(xué)碩士學(xué)位論文基于膚色模型與神經(jīng)網(wǎng)絡(luò)集成的人臉檢測(cè)姓名：杜曉鳳申請(qǐng)學(xué)位級(jí)別：碩士專(zhuān)業(yè)：計(jì)算機(jī)應(yīng)用指導(dǎo)教師：丁友東20060101L海大學(xué)壩二L學(xué)位論文ABSTRACTFacedetectionistheprocessofsearchinghumanfacesinanimagetodecidetheirmlmbers，positions，sizesAsthepreviousstepoffacerecognition，facedete

2、ctionisrelatedtothe療eldsofpattemrecogni“on，imageprocessing，physjoIogy，cognitivescienceetcOVerthepasttenyearsfacedetectionhasbeenthoroughlystudiedincomputerVisionresearchforitsinterestingapplications，suchasfacerecognition

3、system，surVeillancesystemandmachineimerfaceManyalgorithnlshavebeengotthrou曲decade’syearsofstudyButthereexistsstilllotsofdifncultiesinthenledsuchasf≥Iceexpression，illumination，trainingspeedanddetectionspeedToday，moreandmo

4、reimagesarec010rimages，andthec010rfea“IreoffaceskinisanimportantfeatureInthisthesis，thefundamentaltheoriesandthekeytechnologiesoffastfacedetectionfromcolorimagesaremainlystudiedTheresearchworkinthisthesisincludesthefollo

5、wingcomentsFirstly，Dif詫rentskinmodelsa11dseveralc010rspacesarestudied，a11dthenwemakesanimprovementofskincolormodelbasedonk—meanscluster，andwealsoproposedanoVelmethodofthresholdselectionExperimentsshowtheycangetbetterreSL

6、IltsSecondly，facerepresemationssuchasPCA，edgefcatures；HaarandHaarlikefeaturesaremainlystudied，andarepresentationbasedonHaarfeaturesisdesignedbycomparingexperimemSubsequently，somemainstreamclassmersarediscussed，andattenti

7、onsarenlainlypaidonNNE(neuralnetworkensemble)class試erAnovelmethodnamedF10atBagisproposed，whichselectanoptimumsubsetofindividualnetworkstoconstituteanensembleAsamethodoffeaturesselectionbasedonFloatingSearch，F(xiàn)loatBagcange

8、tmoreeHbctiVenetworksthannornlalBagging；whileithascapabilityofparallelcomputingasadeVelopmentofBaggingFinally，afacedetectionsystemforcolorimageisproposedandimplementedbycombiningskindetectionandFloatBagclassifierExperime