鄰近多目標(biāo)場(chǎng)景下MIMO雷達(dá)檢測(cè)數(shù)據(jù)處理算法研究鄰近多目標(biāo)場(chǎng)景下MIMO雷達(dá)檢測(cè)數(shù)據(jù)處理算法研究

摘要：

MIMO雷達(dá)是一種多輸入多輸出雷達(dá)系統(tǒng)，可以利用天線陣列快速獲取地面目標(biāo)信息，并提供較高的分辨率和檢測(cè)性能。然而，在鄰近多目標(biāo)場(chǎng)景下，傳統(tǒng)的MIMO雷達(dá)數(shù)據(jù)處理算法通常存在一些問題，如交叉干擾、信號(hào)混疊等。針對(duì)這些問題，本文提出了一種改進(jìn)的鄰近多目標(biāo)場(chǎng)景下MIMO雷達(dá)數(shù)據(jù)處理算法。首先，使用半監(jiān)督學(xué)習(xí)技術(shù)來減少交叉干擾的影響，并利用最小二乘法進(jìn)行信號(hào)分離。其次，采用弱化總變差（TV）約束的方法來優(yōu)化多目標(biāo)信號(hào)的稀疏表示，并利用非局部均值（NLM）去除信號(hào)混疊。最后，利用仿真實(shí)驗(yàn)和實(shí)測(cè)數(shù)據(jù)進(jìn)行算法測(cè)試，結(jié)果表明，該算法能夠有效地提高鄰近多目標(biāo)場(chǎng)景下MIMO雷達(dá)的檢測(cè)性能。

關(guān)鍵詞：

MIMO雷達(dá)、多目標(biāo)場(chǎng)景、交叉干擾、信號(hào)混疊、稀疏表示。

Abstract：

MIMOradarisamultiple-inputmultiple-outputradarsystemthatcanquicklyacquiregroundtargetinformationusingantennaarrayandprovidehighresolutionanddetectionperformance.However,intheadjacentmultiple-targetscenes,traditionalMIMOradardataprocessingalgorithmsusuallyhavesomeproblems,suchascrossinterferenceandsignalaliasing.Inordertosolvetheseproblems,thispaperproposesanimprovedMIMOradardataprocessingalgorithmforadjacentmultiple-targetscenes.Firstly,semi-supervisedlearningisusedtoreducetheimpactofcrossinterference,andtheleastsquaresmethodisusedtoseparatethesignals.Secondly,thetotalvariation(TV)constraintisweakenedtooptimizethesparserepresentationofmultipletargetsignals,andthenon-localmeans(NLM)methodisusedtoremovesignalaliasing.Finally,simulationexperimentsandactualmeasurementsareusedtotestthealgorithm,andtheresultsshowthattheproposedalgorithmcaneffectivelyimprovethedetectionperformanceofMIMOradarinadjacentmultiple-targetscenes.

Keywords:

MIMOradar,multiple-targetscenes,crossinterference,signalaliasing,sparserepresentationMultiple-targetscenesareacommonoccurrenceinmodernMIMOradarsystems.However,thesescenesposesignificantchallengestosignaldetectionduetothepresenceofcross-interferenceandsignalaliasing.Cross-interferenceariseswhentheechoesfromonetargetinterferewiththeechoesfromanother.Signalaliasing,ontheotherhand,iscausedbyoverlappingofthesignalsfromdifferenttargetsinthefrequencydomain,whichresultsinincorrectlocalizationofthetargets.

Toovercomethesechallenges,weproposeanewalgorithmbasedonsparserepresentationandnon-localmeans.Themainideabehindourapproachistoweakentheinterferencebetweenthetargetsignalsbyexploitingthesparsityofthescene.Specifically,weuseasparserepresentationmodeltoseparatethesignalsfromdifferenttargets,whichallowsustoreducethecross-interference.Thesparsitymodelassumesthatthesignalscanberepresentedasalinearcombinationofasmallnumberofbasisfunctionsordictionaryelements,whichiswell-suitedforMIMOradarsystems.

Inadditiontosparsity-basedprocessing,wealsousethenon-localmeans(NLM)methodtoaddresstheissueofsignalaliasing.TheNLMmethodisapopulardenoisingtechniquethatremovesnoisebyaveragingsimilarpatchesinanimage.Inourcase,weapplytheNLMmethodtotheMIMOradarsignalstoremovethealiasingcausedbyoverlappingofthesignalsfromdifferenttargets.

Toevaluatetheeffectivenessofourproposedalgorithm,weconductedsimulationexperimentsandactualmeasurements.TheresultsshowthatouralgorithmcaneffectivelyimprovethedetectionperformanceofMIMOradarinadjacentmultiple-targetscenes.Specifically,ouralgorithmachievesbettertargetlocalizationandhighersignal-to-interference-plus-noise-ratio(SINR)comparedtoexistingmethods.

Inconclusion,ourproposedalgorithmbasedonsparserepresentationandnon-localmeansisapromisingapproachtoaddressthechallengesofsignaldetectioninmultiple-targetscenesforMIMOradarsystems.OurapproachimprovestheperformanceandreliabilityofMIMOradarsystemsandhaspotentialapplicationsinareassuchassurveillance,mapping,andremotesensingOneareawhereourproposedalgorithmcouldbeparticularlyusefulisdisasterresponse.Insituationssuchasearthquakes,hurricanes,orfloods,traditionaldetectionmethodsmaybeinsufficientinidentifyingpotentialsurvivorsorlocatingburiedvictims.However,MIMOradarsystemsusingourproposedalgorithmcouldimprovethespeedandaccuracyofrescuesbyenablingbettertargetlocalizationandidentificationinthesescenarios.

Additionally,ouralgorithmcouldalsohaveapplicationsinautonomousvehiclesandrobotics.Byimprovingtargetdetectionandlocalizationcapabilities,autonomousvehiclescouldnavigatemoreaccurately,avoidingobstaclesandimprovingsafetyontheroad.Inthefieldofrobotics,ouralgorithmcouldenablebetterperceptionandinteractionwiththeenvironment,allowingrobotstoperformtaskssuchasmappingandexplorationmoreeffectively.

Overall,ourproposedalgorithmbasedonsparserepresentationandnon-localmeanshasthepotentialtogreatlyimprovetheperformanceandreliabilityofMIMOradarsystemsindetectingmultipletargetsincomplexenvironments.Itsapplicationsarewide-ranging,fromdisasterresponseandsurveillancetoautonomousvehiclesandrobotics.Furtherresearchanddevelopmentinthisareacouldleadtosignificantadvancementsintargetdetectionandlocalization,benefitingarangeofindustriesandsocietalneedsSparserepresentationandnon-localmeansalgorithmshavebecomeapromisingsolutionforthedetectionandlocalizationofmultipletargetsincomplexenvironmentsusingMIMOradarsystems.Thistechniquehasshownitspotentialtoimprovethereliabilityandaccuracyoftargetdetectioninavarietyofapplicationsspanningacrossdifferentindustries.

Oneofthemajorbenefitsofthesparserepresentation-basedalgorithmisitsabilitytocapitalizeontheinherentsparsityofthesignaltoaccuratelyidentifyweakandcloselyspacedtargets.Thispropertybecomesparticularlyusefulinenvironmentswheretargetsarecloselylocated,andconventionaltechniquesmaystruggletoresolvethem.Thealgorithmachievesthisbyrepresentingthesignalasalinearcombinationoffewatomsfromapre-defineddictionary.Thesparsecoefficientsarethenestimatedusinganoptimizationmethod,suchasBasisPursuitorOrthogonalMatchingPursuit.ThismethodhasbeenshowntoimprovethedetectionandlocalizationperformanceofMIMOradarsystemsinvariousstudies.

Thenon-localmeansalgorithm,ontheotherhand,exploitstheredundancyofinformationinthereceivedsignaltosuppressnoiseandenhancethesignal-to-noiseratio(SNR).Theapproachreliesontheideathattargetsthatarespatiallyclosewillhavesimilarscatteringcharacteristics.Byexploitingthissimilarity,onecanestimatethesignalatagivenlocationbyaveragingthesignalatotherlocationswithinaspecificrange.Thisaveragingprocesshelpstosuppressrandomnoiseandpreservesignalfeatures,resultinginanimprovementintheSNR.

Thecombinationofthesetwoalgorithmshasshowntoyieldexceptionalresultsindetectingmultipletargetswithhighaccuracy,eveninenvironmentswithahighlevelofnoiseandclutter.Moreover,thealgorithmsrequirelesscomputationtimethantraditionalmethods,whichcanbeadvantageousinreal-timeapplications.Thismakesthemsuitableforawiderangeofapplications,includingsurveillance,disasterresponse,autonomousvehicles,androbotics.

Fordisasterresponseandsurveillanceapplications,theuseofradarsystemsindetectingandlocalizingtargetscanprovecritical.Theabilitytoaccuratelydetectandlocatepeopleorobjectsinadisaster-strickenareacanprovidevaluableinformationtotherescueteams,facilitatingthesearchandrescueprocess.Similarly,insurveillanceapplications,theuseofradarsystemscanhelpindetectingandtrackingintrudersinrestrictedareas,enhancingsecurity.Thesparserepresentationandnon-localmeansalgorithmscanimprovetheperformanceoftheradarsystemsusedintheseapplications,leadingtobetterresults.

Intheautonomousvehicleindustry,robusttargetdetectionandtrackingarecrucialforsafenavigation.Theuseofradarsystemsinautonomousvehiclesforobjectdetectionandlocalizationhasrapidlyincreasedinrecentyears.Thecombinationofthesparserepresentationandnon-localmeansalgorithmscanfurtherenhancetheperformanceofthesesystems,improvingtheirreliabilityandaccuracyindetectingandtrackingobjects,thusleadingtosaferautonomousnavigation.

Intheroboticsindustry,theuseofradarsystemscanenablerobotstooperateincomplexenvironmentswithvaryinglightingconditions.Thecombinationofthesparserepresentationandnon-localmeansalgorithmscanimprovetheperformanceoftheradarsystems,makingtherobotsbetterequippedtodetectandlocalizeobjects,leadingtobetterdecision-makingcapabilities.

Inconclusion,thesparserepresentationandnon-localmeansalgorithmshaveshownthepotentialtogreatlyimprov