智能視頻監(jiān)控中的運動目標檢測和跟蹤技術研究.pdf

1、浙江大學信息學院碩士學位論文智能視頻監(jiān)控中的運動目標檢測和跟蹤技術研究姓名：全曉臣申請學位級別：碩士專業(yè)：信息與通信工程指導教師：劉云海20080501浙江人學頌I二學位論文AbstractIntelligentvideosurveillance(IVS)isamulti—disciplinaryresearchinvolvingtheforefrontinareaofimageprocessingandcomputervisionIV

2、ShasimportantscientificsignificanceandbroadprospectsatthesametimeitiSfullofenormouschallengesMovingtargetdetectionandtrackingisthemostbasiccoretechnologiesofIVSwhichnotonlyarethecornerstoneofthefollowingadvancedprocessin

3、gandapplication，suchastargetclassification，behavioranalysis，eventdetection，behaviorrecognition，thevideoimagecompressionandsemanticindexandSOon，butalsoarethekeyforautomation，intelligentandreal—timeapplicationofavideomonit

4、oringsystemHOWtofindallalgorithmwhichcandetectandtracktargetsaccuratelyrapidlyandstable，whichalsoCandealwiththecomplexenvironmentalchanges，iSstillabigissuetoberesolvedInthispaperbasedonthebackgroundofachievingintelligent

5、videosurveillancesystem，theauthorfocusontheresearchofmovingtargetdetectionandtracking，whicharekeytechnologiesforIVSThemainworkofthispaperisfollowing：Theauthorsummarizedthemovingtargetdetectionalgorithms，andhadanin—deptha

6、nalysisonthesubtractbackgroundmethodbasedonmixedGaussianmodel，anddescribedthebasictheoryandthesimplificationforconcreteapplicationTheexperimentalshowedthee伍ciencyofthemethodindealingwithmultimodalbackgroundNexttheauthoro

7、utlinedthealgorithmforshadowdetection，andhadadetailedanalysisoftheshadowdetectionalgorithmbasedonHSVcolorspaceThentheauthorhadanin—depthanalysisoftheparticlefilteranddescribedthebasicalgorithmframework，underwhichtheparti

8、clefiltercouldbeusedfortrackingMorecuesontheintegrationofthetrackingmethodwasanalyzedIntheframeworkoftheparticlefilteratrackingmethodbasedonadaptiveintegrationofcolorcueandcomercuewasproposedExperimentalresultsshowedthat

9、theproposedmethodismorestablethanthemethodswhichonlyusingasinglecue，andCaneffectivelyhandletheproblemofsuddenlychangeoflightandthebackgroundinterfulsionTheauthorhadananalysisontheocclusionproblemoftracking，outlinedthesta

10、teoftheartofthemethodstosolvetheocclusionproblemAnoveladaptivemultiplecuefusionbasedalgorithmWasproposedforantiocclusionsobjecttracking，whichwasintheframeworkofparticlefilterBinaryshapetemplateandcolorhistogrambasedonker

11、nelellipsewereusedascuesfortracking，thevisibilityofallobjectwasintroducedtoevaluatethereliabilityofthecolorinformationunderocclusions，accordingtothevisibilitythetwocueswerefusedadaptivelytoobtainthemeasurementmodelExperi

12、mentresultsshowedthattheproposedalgorithmWasmuchmorerobusttoocclusionsproblem，comparedwiththetrackingmethodbasedonsinglecuekeywords：targetdetection，mixedgaussmodel，shadowdetection，targettracking，particlefilter，multi—cuef