资源描述：
《Extended target tracking using a Gaussian Mixture PHD filter》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、1ExtendedTargetTrackingusingaGaussian-MixturePHDfilterKarlGranstrom,¨Member,IEEE,ChristianLundquist,andUmutOrguner,Member,IEEEAbstract—ThispaperpresentsaGaussian-mixtureimplemen-BaumetalhavepresentedtherandomhypersurfacemodeltationofthePHDfilterfortrackingextendedtargets.The[4],anexten
2、dedtargetmodelwhichhasbeenusedtotrackexactfilterrequiresprocessingofallpossiblemeasurementsetelliptictargets[5],aswellasmoregeneralshapes[6].Apartitions,whichisgenerallyinfeasibletoimplement.Amethoddifferentapproachtoelliptictargetmodelingistherandomisproposedforlimitingthenumberofcon
3、sideredpartitionsandpossiblealternativesarediscussed.TheimplementationisusedmatrixframeworkbyKoch[7].Thetargetkinematicalstatesonsimulateddataandinexperimentswithreallaserdata,andaremodeledusingaGaussiandistribution,whiletheellip-theadvantageofthefilterisillustrated.Suitableremediesar
4、esoidaltargetextensionismodeledusinganinverseWishartgiventohandlespatiallyclosetargetsandtargetocclusion.distribution.UsingrandommatricestotrackgrouptargetsIndexTerms—Targettracking,extendedtarget,PHDfilter,underkinematicalconstraintsisdiscussedin[8].Modificationsrandomset,Gaussian-mix
5、ture,lasersensor.andimprovementstotheGaussian-inverseWishartmodelof[7]havebeensuggestedin[9],andthemodel[7]hasbeeninte-I.INTRODUCTIONgratedintoaProbabilisticMulti-HypothesisTracking(PMHT)Inmostmulti-targettrackingapplicationsitisassumedthatframeworkin[10].Acomparisonofrandommatricesa
6、ndtheeachtargetproducesatmostonemeasurementpertimestep.randomhypersurfacemodelundersingletargetassumptionThisistrueforthecaseswhenthedistancebetweentheisgivenin[11].Measurementsoftargetdown-rangeextenttargetandthesensorislargeincomparisontothetarget’sareusedtoaidtrackretentionin[12].
7、Otherapproachestosize.Inothercaseshowever,thetargetsizemaybesuchestimatingthetargetextensionsaregivenin[13]–[15].thatmultipleresolutioncellsofthesensorareoccupiedbyUsingtherigorousfinitesetstatistics(FISST),Mahlerhasthetarget.Targetsthatpotentiallygiverisetomorethanpioneeredtherecenta
8、dvancesinthe
