Image-basedgaitanalysisasameansofbiometricidentificationhasattractedmuchresearchattention.Mostoftheexistingmethodsfocusonhumanidentification,postureanalysisandmovementtracking.Therehavebeenfewinvestigationsonmeasuringthecarriedloadbasedonthecarrier'sgaitcharacteristicsbyautomaticimageprocessing.Nevertheless,thismeasurementisveryusefulinanumberofapplications,suchasthestudyofthecarriedloadontheposturaldevelopmentofchildrenandadolescence.Inthispaper,weinves-tigatehowtoautomaticallyestimatethecarriedweightfromasequenceofimages.Wepresentamethodtoextracthumangaitsilhouettebasedonanobservationthathumanstendtominimizetheenergyduringmotion.Weputeseveralanglesofbodyleaninganddeter-minetherelationshipofthecarriedweight,theleaninganglesandthecentroidlocationaccordingtoahumankiicstudy.Ourweightdeterminationmethodhasbeenverifiedsuccessfullybyexperiments.
KuiZhang,YingMu(ImageProcessingCentre,BeihangUniversity,Beijing100083,China)
NingYao,RobertJ.Sclabassi,MinguiSun(LaboratoryforComputationalNeuroscience,DepartmentsofNeurosurgery,Electrical&ComputerEngineering,andBioengineering,UniversityofPittsburgh,Pittsburgh15213,USA)
第2篇：Image registration based on matrix perturbation analysis using spectral graph
Wepresentanovelperspectiveoncharacterizingthespectralcorrespondencebetweennodesoftheweightedgraphwithapplicationtoimageregistration.Itisbasedonmatrixperturbationanalysisonthespectralgraph.Thecontributionmaybedividedintothreeparts.Firstly,theperturbationmatrixisobtainedbyperturbingthematrixofgraphmodel.Secondly,anorthogonalmatrixisobtainedbasedonanoptimalparameter,whichcanbettercapturecorrespondencefeatures.Thirdly,theoptimalmatchingmatrixisproposedbyadjustingsignsoforthogonalmatrixforimageregistration.Experimentsonbothsyntheticimagesandreal-worldimagesdemonstratetheeffectivenessandaccuracyoftheproposedmethod.
ZhengTian(SchoolofScience,NorthwesternPolytechnicalUniversity,Xi'an710072,China;StateKeyLaboratoryofRemoteSensingScience,InstituteofRemoteSensingApplications,ChineseAcademyofSciences,Beijing100101,China)
第3篇：Image registration based on matrix perturbation analysis using spectral graph
Wepresentanovelperspectiveoncharacterizingthespectralcorrespondencebetweennodesoftheweightedgraphwithapplicationtoimageregistration.Itisbasedonmatrixperturbationanalysisonthespectralgraph.Thecontributionmaybedividedintothreeparts.Firstly,theperturbationmatrixisobtainedbyperturbingthematrixofgraphmodel.Secondly,anorthogonalmatrixisobtainedbasedonanoptimalparameter,whichcanbettercapturecorrespondencefeatures.Thirdly,theoptimalmatchingmatrixisproposedbyadjustingsignsoforthogonalmatrixforimageregistration.Experimentsonbothsyntheticimagesandreal-worldimagesdemonstratetheeffectivenessandaccuracyoftheproposedmethod.
ZhengTian(SchoolofScience,NorthwesternPolytechnicalUniversity,Xi'an710072,China;StateKeyLaboratoryofRemoteSensingScience,InstituteofRemoteSensingApplications,ChineseAcademyofSciences,Beijing100101,China)