A PMHT approach for extended objects and object groups

《A PMHT approach for extended objects and object groups》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、I.INTRODUCTIONInrealistictrackingapplicationsthereisahighdemandfortherecognitionofextendedobjectsasAPMHTApproachforindividualunits,fortheinitiationofextendedobjecttracks,andforextendedobjecttrackmaintenance.ExtendedObjectsandObjectExtendedobjectsarecharacterizedbyarelativelylargeandoftenstronglyfl

2、uctuatingnumberofsensorGroupsreportsoriginatedbytheindividualscatteringcentersthatarepartofoneandthesameobject.Inthiscontext,weusuallycannotassumethatinsubsequentobjectilluminationsthesamescatteringcenterswillalwaysberesponsibleforthemeasurements.TheMONIKAWIENEKEindividualsensorreportscantherefore

3、nolongerWOLFGANGKOCH,Fellow,IEEEbetreatedinanalogytopointsourcemeasurementsFraunhoferFKIEproducedbyagroupofwell-separatedobjects.In[1]aBayesianapproachtoextendedobjecttrackingusingrandommatricesispresented.WithinConventionaltrackingalgorithmsrelyontheassumptionthisapproach,ellipsoidalobjectextents

4、aremodeledthatthetargetsofinterestarepointsourceobjects.However,byrandommatricesandtreatedasadditionalstateinrealisticscenariosthepointsourceassumptionisoftennotvariablestobeestimated.However,theproposedtrackingmethoddidnotincludeasolutionforsuitableandestimatingtheobjectextentbecomesacrucialaspec

5、t.dataassignmentconflictstypicallyoccurringinRecently,aBayesianapproachtoextendedobjecttrackingusingmulti-objectscenarios.Therefore,wenowpresentrandommatriceshasbeenproposed.Withinthisapproach,themulti-objectextensionofourapproach.Weellipsoidalobjectextensionsaremodeledbyrandommatricesandderiveane

6、wkindofprobabilisticmulti-hypothesistreatedasadditionalstatevariablestobeestimated.However,tracking(PMHT)thatsimultaneouslyestimatestheonlyasingle-objectsolutionhasbeenpresentedsofar.Inthisellipsoidalshapeandthekinematicsofeachobjectworkwepresentthemulti-objectextentofthisapproach.Weusingexpectati

7、on-maximization(EM).Bothellipsoidsderiveanewvariantofprobabilisticmulti-hypothesistrackingandkinematicstatesareiterativelyoptimizedby(PMHT)thatsimultaneouslyestimatestheellipsoidalshapeandspecificKalmanfilterform