資源描述：
《Driver Eye State Classification Based on Cooccurrence Matrix of Oriented Gradients》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、HindawiPublishingCorporationAdvancesinMechanicalEngineeringArticleID707106ResearchArticleDriverEyeStateClassificationBasedonCooccurrenceMatrixofOrientedGradientsBoZhang,WenjunWang,andBoChengStateKeyLaboratoryofAutomotiveSafetyandEnergy,TsinghuaUniversity,Beijing100084,ChinaCorre
2、spondenceshouldbeaddressedtoBoCheng;chengbo@tsinghua.edu.cnReceived19August2014;Accepted4November2014AcademicEditor:HongweiGuoCopyright?BoZhangetal.ThisisanopenaccessarticledistributedundertheCreativeCommonsAttributionLicense,whichpermitsunrestricteduse,distribution,andreproduct
3、ioninanymedium,providedtheoriginalworkisproperlycited.Accuratedetectionofdriver’seyestatebycomputervisioniscriticaltodriverdrowsinessmonitoring.Thehistogramoforientedgradients(HOG)iscommonlyusedasdescriptivefeatureofeyeimageforstateclassification.However,HOGoftensuffersfromtheli
4、mitoflocalgradientinformation.ThispaperproposesanewHOG-likefeatureofeyeimage,calledcooccurrencematrixoforientedgradients(CMOG),forthepurposeofmoreeffectivelyclassifyingtheeyestate.Byintroducingthecooccurrencematrix,theCMOGenhancestheabilityofdescribingglobalgradientinformationof
5、eyeimages.TheZJUeyeblinkdatabaseisusedasthebaselineimagesforperformancecomparison.TheclassificationresultsshowthattheaccuracyofCMOGreachesupto95.9%incomparisonwith91.9%byHOGunderthisdatabase.1.Introductionimportantfactorinthiskindofmethods.Manytypesoffeaturesfromeyeimageshavebee
6、nproposedbynowadaysThedrivereyestate,thatis,openingandclosing,istheresearchers,suchasHOG(HistogramsofOrientedGradi-mostsalientfacialexpressionrelatedtodriverdrowsiness.ents)[4],LBP(LocalBinaryPatterns)[5],GaborwaveletsEyestateclassificationbasedoncomputervisionplaysan[6],Eigeney
7、e[7],andASEF(AverageofSyntheticExactimportantroleinthefieldofdrowsinessmonitoring.ItisaFilters)[8].TakingLBP,forexample,itusesthedifferencechallengingtasktodetectdrowsinessfromeyeimagesowingbetweenpixelsinalocalscaletorepresenttheimagewhichtovariablefacialexpression,randomillumi
8、nation,andheadisinsensitivetoilluminationchangi
