英文翻译运用于机场多普勒天气雷达的移动杂波频谱滤波器.docx

英文翻译运用于机场多普勒天气雷达的移动杂波频谱滤波器.docx

《英文翻译运用于机场多普勒天气雷达的移动杂波频谱滤波器.docx》由会员分享，可在线阅读，更多相关《英文翻译运用于机场多普勒天气雷达的移动杂波频谱滤波器.docx（19页珍藏版）》请在冰豆网上搜索。

英文翻译运用于机场多普勒天气雷达的移动杂波频谱滤波器

成都信息工程学院

毕业设计英文翻译

运用于机场多普勒天气雷达的移动杂波频谱滤波器

系别

电子工程学院

姓名

杨杰

专业

电子信息工程

班级

大气探测2班

学号

2007021206

MovingClutterSpectralFilterforTerminalDoppler

WeatherRadar

1.INTRODUCTION

Detectinglow-altitudewindshearinsupportofaviationsafetyandefficiencyistheprimarymissionoftheTerminalDopplerWeatherRadar（TDWR）.Thewind-sheardetectionperformancedependsdirectlyonthequalityofthedataproducedbytheTDWR.Attimesthedataqualitysuffersfromthepresenceofclutter.Althoughstationarygroundcluttersignalscanberemovedbyahigh-passfilter,movingcluttersuchasbirdsandroadwaytrafficcannotbeattenuatedusingthesametechniquebecausetheirsignalpowercanexistanywhereintheDopplervelocityspectrum.Furthermore,becausetheTDWRisasingle-polarizationradar,polarimetrycannotbeusedtodiscriminatethesetypesofclutterfromatmosphericsignals.

ThemovingclutterproblemisexacerbatedatWesternsiteswithdrymicroburst,becausetheirlowsignal-to-noiseratios（SNRs）aremoreeasilymaskedbyunwantedmovingclutter.ForLasVegas（LAS）,Nevada,theoffendingclutteristrafficonroadsthatareorientedalongtheradarlineofsightneartheairport.Theradarislocatedatasignificantlyhigheraltitudethanthetown,improvingthevisibilitytotheroads,andgivingLAStheworstroadclutterproblemofallTDWRsites.TheSaltLakeCity（SLC）,Utah,airportislocatedneartheGreatSaltLake,whichisthebiggestinlandstagingareaformigratingseabirdsinthecountry.It,therefore,suffersfrombirdclutter,whichnotonlycanobscurewindshearsignaturesbutcanalsomimicthemtotriggerfalsealarms.TheTDWR“dry”siteissuesarediscussedinmoredetailbyCho（2008）.

Inordertomitigatetheseproblems,wedevelopedamovingclutterspectralfilter（MCSF）.Inthispaperwedescribethealgorithmandpresentpreliminarytestresults.

2.THEPROBLEM

Figure1illustratesoneofthedifficultieswithnotfilteringoutmovingcluttersignalsfromTDWRdata.Awidespreadcoherentbirdflighteventlikethishastwoofthecharacteristicsofamicroburst-outflowofheightenedreflectivityawayfromacentralsourceandvelocitydivergencealongtheradials.Thus,amicroburstdetectorisindangerofissuingafalsealertinthiscase.

Figure2showsDopplervelocityspectravs.rangeforamicroburstevent（left）andacasewithbirdsinmanydifferentrangegates（right）.Notonlyaretherebirdsignaturesthatareisolatedinrange-Doppler,thereisafeatureinrangegates~40-50thatisquasi-continuousandlookssimilartoastrongwindshear.Thechallengeistofilterthebirdsignalsintheright-handcasebutnotthemicroburstintheleft-handcase.

Figure1ExampleofbirdclutterobservedwiththeSLCTDWRat0.5°elevation.

Leftpanelshowsreflectivity,rightpanelshowsradialvelocity.

Figure2.Dopplervelocityspectravs.rangeforamicroburstcaseobservedwiththeProgramSupportFacility（PSF）TDWRinOklahomaCity,OKat0.3°elevation（left）,andawidespreadbirdcontaminationcaseobservedwiththeSLCTDWRat0.5elevation（right）.

Batscanalsocausesimilarproblemsforradarwindsheardetectionwhentheyleavetheirroostsenmasseatdusk.

Roadandrailtrafficarealsosourcesofmovingclutter.Inthiscase,therange-azimuthcellsaffectedareknownapriori,sothecurrentmethodofdealingwiththeproblemistoperiodicallygenerateaclutterresiduemap（CREM）andthencensorthebasedatawherethereflectivitydoesnotexceedacertainamountovertheCREMreflectivity.TheLASroadclutterproblemisshowninFigure3.Notethatsomeoftheclearestroadechoeslineupwiththeradarline-of-sightradials.Thisphenomenonisalsoobservedatothersites,andwecallitthebuildingcanyoneffect.Inanurbanenvironmentroadsarelinedwithbuildings,sotrafficisoftennotvisibletotheradarunlessthebeamshootsdownalongtheroaditself.

Figure3.LASroadclutterat0.8°elevation.Leftpanelshowsreflectivity,rightpanelshowsradialvelocity.Thelargewedgeofblankreflectivitytothenortheastoftheradarisduetoterrainblockage.

Airplanes,ofcourse,arealsomovingclutter.Becausetheyareisolatedtargets,apointtargetfilterdealswiththemeffectively.Althoughthecurrentoperationalalgorithmmerelycensorsthesepoints,weplantointerpolatethebasedataacrossthesepointsinthenextmajorradardataacquisition（RDA）systemsoftwarerevision.Thefilteringalgorithmpresentedinthispapershouldbeanevenbettersolutioninthefuture.

Finally,thereareothermovingtargetsthatappearinTDWRdatasuchasseaclutterandthespinningbladesofawindpowerturbine.Thelatterisexpectedtobeanever-growingsourceofclutterthatisstillinsearchofasolution.Theworkpresentedinthispaperdoesnotspecificallyaddressthesephenomena,asthefocushasbeenonthedataqualityissuesatSLCandLAS.

3.MOVINGCLUTTERSPECTRALFILTER

Stationaryclutterfiltersaretypicallyappliedononerange-azimuthdwelldataatatime.Recognizingthatmorecontextualinformationisnecessarytoidentifymovingclutter,techniquesweredevelopedthatutilizemultiplerangegatesofdata.ThepointtargetfiltersemployedonboththeTDWRandtheWeatherSurveil-lanceRadar-1988Doppler（WSR-88D,morecommonlyknownasNEXRAD）lookfortargetswithreflectivitymuchhigherthantheneighboringrangegates.Thistypeoffiltercanremoveaircraftandisolatedaviansignals,buttheweatherreturnsinthesamecellsarealsolost.

Amoremodernapproachfiltersdatainthetwo-dimensional（2D）,Dopplervelocityspectrumvs.range,domain（e.g.,Sasaoka2003;Meymaris2007）.Becausemovingclutterspectralsignalstendtobespectrallycompactanddiscontinuousinrange,thesetechniqueswouldlookforarange-continuoussignal（weather）anddiscardotherspectralcomponents（clutter）.（Ofcourse,stationarygroundcluttercanalsobecontinuousinrange,soitwouldhavetobefilteredoutfirst.）Twodrawbackstothisapproacharethatitiscomputationallyintensiverelativetotraditionalweatherradarsignalprocessing,andthattheoutputbasedataaresmoothedinrange.

WiththedevelopmentofanewRDAsystem（Choetal.2005）,whichiscurrentlyrunningoperationallyatLASandSLC,thereisconsiderablesparecomputationalcapacityfortheimplementationoffuturealgorithms.Thescalablearchitectureeasilyallowsforadditional（and/orupgraded）centralprocessingunits（CPUs）tobeinstalledtomeetfurthercomputationaldemand.Withregardtotheseconddrawbackoftheexisting2Dspectralprocessingapproachformovingclutterfiltering,spatialsmoothingofthedatarunscountertotheTDWR’smissionofdetectingwindshear.ThisisthereasonwhywedecidedtodevelopourownmovingclutterspectralfilterfortheTDWR.

First,wewillgiveabriefoutlineoftheMCSFalgorithm,andthendelveintomoredetails.

Tosetuptheprocessingenvironment,weformaDopplervelocitypowerspectrumateachrangegatewiththeusualwindow-and-DFTapproach.Thesespectraarestoredina2Dmatrix（Figure2,right）.Thefollowingstepsarethenfollowed.

1）Removestationarygroundclutterwiththespectralfilterofchoice（Figure5）.

2）Removepositivepoweranomaliesalongrangeateachspectralbininamannersimilartoapointtargetfilter（Figure6）.Cluttersignalsthatarenotcontinuousinrangeareremovedinthisway.

3）Aftersuitabledatamassaging,determinethenumberandlocationofspectralmodes（statisticallysignificantpeaks）ateachrangegate.Ameasurecalledthenormalizedcircularexcessmass（NCEM）isusedtowinnowout"insignificant"peaksinthespectra.

4）Selectonemodeateachrangegatesuchthatthepathconnectingthemodeshastheshortestpossibleoveralldistance（Figure7）.Thisprocessfavorsretaininggloballycontinuous（inrange）signalsoveronlylocallycontinuousfeatures.Theassumptionisthatatmosphericsignalswillbecontinuousoverlongerrangesthanmovingclutterfeatures.

Figure5.StationarygroundclutterisremovedFigure6.Isolatedspectralpeaksareremoved

fromthespectrashowninFigure2,right.fromthespectrashowninFigure5.

5）Formodesotherthantheonesselectedbytheshortestconnectingpath,reducethenumberof"extra"modesusingastricterNCEMthreshold.Thisstepisneededtoreducetheincidenceofthese"extra"modesbeingpartoftheatmosphericspectrum.

6）Anyleftover"extra"modesaredeemedtobemovingclutterandareremovedbyfillinginwiththecomputedspectral"noise"floor（Figure8）.

7）Carryonwiththerestoftheusualsignalprocessingforbasedatageneration.

Letusnowelaborateoneachstep.（Note:

Theterm"circular"usedinthispaperreferstothefactthattheDopplervelocityspectrumaliases,orwrapsaround,attheendpoints.）Thereisnothingspecialaboutthefirststepofstationaryclutterfiltering.Oneofvariousdifferenttechniquescanbeused.However,notethattheGaussianmodeladaptiveprocessing（GMAP）filter（SiggiaandPassarelli2004）assumestheabsenceofspectralsignalotherthanastationarygroundclutterpeakandaGaussianweathercomponent.Therefore,thepresenceofmovingcluttersignalsdistortstheweathermomentestimationandcanspuriouslywidenthe"weather"spectrumwithinthefilledgroundcluttergap.We,thus,recommendagainsttheuseofGaussianfittingandclutter-gapfillinginthesuspectedpresenceofmovingclutter.

Theisolatedspectralpeakfilterinstep2issimilartotheTDWRpointtargetfilter,exceptthatitrunsalongeachspectralbinvs.range,andthenumberofneighborsconsidered