交通监控视频中的异常事件检测

《交通监控视频中的异常事件检测》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、北京邮电人学硕：f：学位论文摘要关键词：虚拟线圈，时间窗ISl，ISM，前景分割，均值位移，感兴趣区域，中心点．包络模型一u，北京邮电大学硕上学位论义abstractABNoRM队LITIESDETECTIoNINTRAFFICMoNITOlUNGSCENESABSTRACTAbnormalitiesdetectioniSacriticalresearchfieldofIntellegentTransportationSystem(ITS)．ItwillserveasanimportantcomponentofITSwh

2、enitbecomesapplicable．TheVirtualLoopSystemisbeingwidelyusednow,butabnormalitiesdetectiontechniquebasedontheanalysisofmovingobjects’activitieshasnotbeenavailableyet．Themarketdemandforthiskindofdetectionsystemisenlarging．ThispapergivesadetailedresearchoftheVirtualL

3、oopSystemfortrafileflowmonitoring，andtheIrregularTrajectoriesDetectionSystemaswell．AsimpleVirtualLoopSystemhasbeensetuptocountthetrafficflowandroughlyevaluatethevehiclespeed．TheformoftheVirtualLoopsisthreesuccessiveVirtualLinesforeachvehiclechannel．It’Slikeaneval

4、uationgateplaced"ontheroad．Thissystemiscomputationefficientanddoesnotinvolvecomplexreasoning．ItcancountthetrafjfiCflowaccurately．Thispaperpresentsavirtuallines’statesanalysisalgorithmbasedontimewindowtoimprovethesystemperformance．Itholdsthecomputeefficiencyofthes

5、ystem，anddoesnotintroducetoomuchcomplexity．ThispapergivesanuanceddiscussionoftheIrregularTrajectoriesDetectionSystem，andcompletelypresentsthealgorithmsofthefourcomponents，includingISM(illuminationsignificancemeasure)extraction，foregroundprocessing，objectstracking

6、andtrajectoriesanalysis．ThefirsttwocomponentscontaintheISMthresholdingmethodforextractionoftheforegroundpoints．Itkeepsthecomputationefficiencyandcanimmediatelyrespondtothesuddenlightchangesofthescene．SoitCanmeetthedemandsofreal．timemonitoring．Inordertoruleoutthef

7、alse．positiveforegroundpointsandpartsoftheshadowpoints，analgorithmforadjustingthebounding-boxoftheobjectshas北京邮电人学硕}：学位论文abstractbeenproposed．Itcangetbetterbounding-box，usuallymoreaccurate．ThispaperchoosestheMeanShiftalgorithmforobjectstracking．Atlast，theprincipl

8、esoftrajectoriesanalysisarepresented，includingthePointsofInterestandtheconceptofCentroid—Envelop．ThispaperproposesaPOI(PointsofInterest)AutomatedExtractionAlgo