资源描述:
《Tracking Object In Video Sequence Using Active Contour Models and Unscended Kalman Filter.pdf》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、20117thInternationalWorkshoponSystems,SignalProcessingandtheirApplications(WOSSPA)TRACKINGOBJECTSINVIDEOSEQUENCEUSINGACTIVECONTOURMODELSANDUNSCENTEDKALMANFILTER112ZahirMessaoudi,AbdelazizOuldaliandMouradOussalah1ElectronicsandOptronicsLaboratory,EcoleMilitairePolytechnique,Algiers,Algeria2EECEDep
2、artment,UniversityofBirmingham,Birmingham,UnitedKingdomTheyproposeanewmodelbasedontheassociationofABSTRACTGACandC-Vforimagesegmentation.ThisnewmodelusesthestatisticalinformationinsideandoutsidethecontourinordertoconstructaregionbasedsignedInthispaper,weproposeanewassociationofactivepressureforce(
3、SPF).ThepurposeofusingtheSPFistocontourmodel(ACM)withtheunscentedKalmanfiltercontrolthedirectionoftheevolutionofthecurve.Also,(UKF)totrackdeformableobjectsinavideosequence.in[5]theauthorsproposeanewselectivebinaryandTheproposedapproachisbasedontheuseoftheGaussianfilteringregularizedlevelset(SBGFR
4、LS),selectivebinaryandGaussianfilteringregularizationinsteadofthetraditionallevelset(TLS),toimplementlevelsetassociatedtotheUKF(ACM-SBGFRLS-UKF)thenewmodel.TheSBGFRLSpresentsmanyadvantagesinsteadofthetraditionallevelset(TLS)associatedtothesuchasavoidingthecalculationofthesigneddistanceUKF(ACM-TLS
5、-UKF).Infact,inthepresentwork,wefunction(SDF)anditsre-initializationasintheTLScaseexploitthevariousadvantagesthattheSBGFRLSoffers[5].Inaddition,theSBGFRLSismoreefficientthanthecomparedtotheTLSwhichsuffers,fromthesensibilityTLSandalsohasthepropertyoflocalorglobaltoinitialsconditionsandnoise,tothei
6、mpossibilitytosegmentationselectionwhichensuresalowcomputingselectpartialorglobalsegmentationandalsofromthetime[5].In[3]theauthorsproposetousetheassociationcomplexityoftheapproach.Finally,acomparisonstudyispresented,throughoutseveralnumericalsimulations,oftheunscentedKalmanfilter(UKF)andtheC-Vmod
7、elofthisnewassociationapproachACM-SBGFRLS-UKFimplementedwiththeTLS(ACM-TLS-UKF)tosolvetheagainsttheACM-TLS-UKFfortrackingdeformableproblemoftrackingdeformableobjectsinvideoobjectsinavideosequence.sequence.Inthisapproac