移動對象軌道聚類算法的研究.pdf

1、南京航空航天大學(xué)碩士學(xué)位論文移動對象軌道聚類算法的研究姓名：王明濤申請學(xué)位級別：碩士專業(yè)：計算機科學(xué)與技術(shù)指導(dǎo)教師：皮德常201012移動對象軌道聚類算法的研究IIAbstractWiththedevelopmentofcomputertechnologyaswellascontinuousimprovementofmovingobjecttrackingtechnologymassivetrajectydataiscollected.

2、Itbecomesurgenttoextracthiddeninfmationfromthesedatathrougheffectiveanalysiswhichleadstothegenerationdevelopmentofmovingobjecttrajectyclusteringtechnology.Trajectyclusteringofmovingobjectcanfindobjectswiththesamesimilarm

3、ovementbehaviextracttheircommonfeaturepatternusingwhichtoanalyzepredictthefuturemovingbehaviofmovingobjects.Fthepurposeoffindingcommonsubtrajectyofmovingobjectsinthispaperresearchexplationarecarriedoninviewsoftheweakness

4、ofTRACLUS.ThescopeofthisresearchliesinsomeburningissueswhichinvolveinenhancingTRACLUS’sexecutingefficiencyshieldingitsparametersensitivitysoon.Themaincontributionsofthispaperaresummarizedasfollows:Firstlyexistingspatiali

5、ndexstructurescannotbedirectlyusedtosupptefficientneighbhoodqueryoflinesegmentinTRACLUS’slinesegmentclusteringphasemakingthetimecomplexityoflineclusteringalgithmisO(N2)whereNisthetotalnumberoflinesegmentsaftertrajectypar

6、titioneditrequireslargevolumeofmemysupptneedsalotofIOcostsasNincreases.AimtothisissuereferencelineisintroducedtotrajectyclusteringRLTCbasedonthereferencelineispresentedwhichusesacertainnumberofreferencelinestorepresentth

7、espatialacteristicsofalineclusteringapproximately.Theclusteringofthelinesegmentscanbeimplementedbytheclassificationofreferencelines.TheexperimentalresultsdemonstratethatRLTCmaintainsthetrajectyclusteringresultofTRACLUSwh

8、ileimprovestheefficiency.SecondlyaimingtotheproblemthatTRACLUSissensitivetotheinputparameterssuchlikeεMinLnsatrajectyclusteringalgithmnamedOPSTCisproposedwhichcanshieldclusteringalgithm’sparametersensitivity.OPSTCdefines

9、thecedistancereachabilitydistanceoflinesegmentaccdingtowhichgeneratesprocessingsequenceoflinesegmentwhichactuallycontainstheinfmationthatequalstothelinesegmentclusteringsbasedonthedensityfromabroadinputtingrangeofparamet

10、ers.Thenitconstructstherechabilityplotoflinesetsaccdingtothesequencewithwhichextractsthefinalclusteringstructuresafterproperdistanceisedfextraction.TheexperimentalresultsrevealthatOPSTCreducesTRACLUS’sdependencysensitivi