針對空域目標的視頻跟蹤技術研究.pdf

1、南京理工大學碩士學位論文針對空域目標的視頻跟蹤技術研究姓名：毛瑞申請學位級別：碩士專業(yè)：控制理論與控制工程指導教師：盛安冬20080615碩士論文AbstractVisualtrackinghaslongbeenoneofthehotspotsindomesticandforeignresearchIthasbeenwidelyappliedinnavigationsecuritysurveillance，medicineandscie

2、ntificresearchetcEspeciallytheresearchonvideotrackingforspatialobjecthasearlierbecomethekeytechnologyinmilitaryfieldTheaccuracyandstabilityoftrackingeffectdependonthedesignofalgorithmstoagreatextentSofartherehavebeenlots

3、ofreferencesputtingforwardoptimizedalgorithmontargetdetectionHowevertoimprovethetrackingperformancefurtherwillbedifficultsimplyrelyingontargetdetectionWhileorganicallycombiningtargetdetectionwitheffectivetrackingorfilter

4、algorithmwillprobablyhelptealizestableobjecttrackingTherefore，aimedatvisualtrackingforspatialobject，thispaperwillmakeresearchonbothobjectdetectionalgorithmandstateestimationFirstlycirclingaroundobjectdetectionalgorithm，t

5、hisarticledesignedtrackingwave—doorandrealizedgrayimagebinarizationbyassociatingthewavedoordimensionwiththemethodofsectioningthreshold，thenthenearestneighborassociationalgorithmwasproposedtoidentifythetruetargetfromthose

6、deceptiousobjectscausedbybinarizationmethodandtheresultsofemulationalexperimentdemonstrateditsvalidityNext，inthetrackingprocess，standardKalmanfilteralgorithmwasusedfirstlytoforecastandtrackthespatialtarget，thenthecorrela

7、tionbroughtbymeasureerrorwasanalyzed，furthercontinuedwiththeKalmanfilteralgorithmwhichconsideredcorrelativemeasurenoisewasproposedtoestimateeffectivelythemotionalfeatureofairspacetargetandthealgorithmnotonlyachievedthego

8、alofaccuratetrackingbutalsoenhancedtherobustnessoftrackingsystemFinallymanydifferenttargetsampleswere鼬ckedandtestedbasedonCVmodel。Singermodel，”PresentHmodelandIMMmodelSOthatmostsuitabletargettrackingmodelandselfadaptivef

9、ilteralgorithmforvisualtrackingcouldbeacquiredbyanalyzingandcomparingestimatedprecisionforthesametargetundereachdifferentmodelTheoreticalanalysisandnumericalsimulationhaveprovedthatKalmanfilteralgorithmconsideringnoisede

10、correlationhadobviouspredominance011theforecasteITOISthantraditionalKalmanfilteralgorithmandtheaccuracyofvisualtrackingwasimprovedagaindependonIMMmodelThroughthiswaythispapermayprovideusefulreferenceforfuturevisualtracki