三维人脸数据获取与识别技术分析

《三维人脸数据获取与识别技术分析》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、AbstractAsanimportantcomponentofbiologicalfeaturerecognition,facerecognitionhasbeenoneofthemostpopulartopicsinthefieldofpaRemrecognitionandartificialintelligence，withtheadvantagesofinitiative：convenience，non-invasiveandconcealmen_t’ete．TheFacerecognitionalgorithmsw

2、hicharebasedonthe2Dimages，areoftengreatlyinfluencedbylight,posture，facialexpressionsandotherfactors．Sofacerecognitionalgorithmswhicharebasedon3Ddata,areanincreasingconcerns．Thispaperfocusesonthestudyofthequicklyaccruingof3Dfacedatawithgoodexpressionabilitybybinocul

3、arstereovisiontechnology．Onthisbasis，westudythePCAfacerecognitionapproach(basedonthedepthmap)andrecogni'tionapproach(basedongeometricfeatures)．Furthermore，wecombinedthesetwomethodsforfacerecognition．Themaincontributionsofourworkareasfollows．(1)3Dfacepointclouddataa

4、cquisition．First,wemake戳ofepipolarrectificationalgorithm,whichisbasedonrectificationerror,todoepipolarrectificationonbinocularimagesofhumanfaces．Consideringthefactthatfaceisaspecialstereomatchingscene，weresearchseveralstereomatchingmethods，whichincludingthefastregi

5、onstereomatchingmethod,improvedRanktransformstereomatching(basedonmutualinformation)anddisparitygrowingmethod．Furthermore，weintroducepyramidslxuctureformatching，璐啦differentstereomatchingmethodtoeachlayeroffaceimage，toquicklygetthehigher-accuracydensedisparitymap．Fi

6、nally，withthecKmeraparametersachievedbythepreviouscameracalibrationstep，wereconstructthe3Dfacepointclouddata埘mgoodexpressionability，andthe3Dfacedatabasenamed3DFACE-XMUisupdated．(2)Aresearchon3Dfacerecognitionintegrativetechniquebasedonpo衄cloud．Takingglobalandlocalf

7、eaturesintoaccount，wecollectPCAstatisticaldepthfeaturesandgeometricfeaturesof3Dfacerespectivelytoguiderecognition．WeusethreeintegrativetechniquesbetweenPCAfacerecognitionand3DgeometricfeaturesIIIrecognition,whichaletheintegrativeapproachofselectingtheresultsofPCAus

8、inggeometricfeatures，theintegrativeapproachbasedoncriticaldecisionandtheintegrativeapproachbasedOllvotescore．(3)Developa3Dfacerecognitionintegrat