基于特征的Gabor小波人臉識(shí)別算法改進(jìn).pdf

1、江蘇大學(xué)碩士學(xué)位論文基于特征的Gab小波人臉識(shí)別算法改進(jìn)姓名：俞燕申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：李正明20100612江蘇大學(xué)碩士學(xué)位論文ABSTRACTFaceRecognition，whichhasbeenstudiedformorethan30years，isoneofthemostchallengingresearchesinpatternrecognitionandmachinevision，itcanb

2、ewidelyappliedinthepublicsecurityinformationsecurityandhuman—computerinteractionGenerallyspeaking，thefacerecognitionsystemconsistsoffacedetection，featurepiontlocation，imagepre—processing，featureextractionandfacerecogniti

3、onThethesisstudiesthecorrelativeissuesbasedontheformerworks，themainpointsareasfollows：(1)First，introducebackground，significance，developmenthistoryandtheresearchprogressabroadandathomeinfacerecognitionSecond，buildafacerec

4、ognitionsystemtodiscussthemajorprocessofbuildingaworkofthevarioussectorsofthetechnologyrequired，thealgorithmisintroducedandstudied(2)Animprovedalgorithmforfacerecognitionbasedontheclassical2DGaborwaveletinfeatureextracti

5、onphaseisproposedFirstlyusingsubblocksamplingandSFFSalgorithmsselectsamplingpointoffeatureextractionand40featurematriceswhicharereconstructedwiththesamescaleandthesamedirections2DGaborwavelettransformresultsareobtainedSe

6、condlythedimensionalityreductionanddenoisedtechniquewithPCAareappliedtoformthenewtrainingsamplesLastlySVMclassifierisconstructedandthevotedecisionstrategyisusedtodeterminetherecognitionresult(3)TheresultsofsimulationsonO

7、RLandYalefacedatabaseshowthatcomparingwiththetraditional2DGaborwaveletalgorithms，theproposedalgorithmimprovesdetectionspeedobviouslyandobtainsasatisfactoryrecognitionratiowithsmallsampleprobleminfacerecognitionAtthesamet

8、ime，theproposedalgorithmhasstrongrobustnessonparametricselectioninPCAandSVMTheadvantagesnotonlysettledownthedifficultproblemofselectingthekemelparametersinSVMpartlybutalsoexpanditsapplicationlimitsKEYWORDS：FaceRecognitio