购物中心视频实时监控外文文献参考资料Word文档下载推荐.docx
- 文档编号:14537668
- 上传时间:2022-10-23
- 格式:DOCX
- 页数:10
- 大小:26.18KB
购物中心视频实时监控外文文献参考资料Word文档下载推荐.docx
《购物中心视频实时监控外文文献参考资料Word文档下载推荐.docx》由会员分享,可在线阅读,更多相关《购物中心视频实时监控外文文献参考资料Word文档下载推荐.docx(10页珍藏版)》请在冰豆网上搜索。
RobertoArroyo;
J.JavierYebes.,Expertvideo-surveillancesystemforreal-timedetectionofsuspiciousbehaviorsinshoppingmalls[J]Expert
Systems
withApplications,
2016,12(02):
91-105
Expertvideo-surveillancesystemforreal-timedetectionofsuspiciousbehaviorsinshoppingmalls
J.JavierYebes
Abstract
Expertvideo-surveillancesystemsareapowerfultoolappliedinvariedscenarioswiththeaimofautomatzingthedetectionofdifferentrisksituationsandhelpinghumansecurityofficerstotakeappropriatedecisionsinordertoenhancetheprotectionofassets.Inthispaper,weproposeacompleteexpertsystemfocusedonthereal-timedetectionofpotentiallysuspiciousbehaviorsinshoppingmalls.Ourvideo-surveillancemethodologycontributesseveralinnovativeproposalsthatcomposearobustapplicationwhichisabletoefficientlytrackthetrajectoriesofpeopleandtodiscoverquestionableactionsinashopcontext.Asafirststep,oursystemappliesanimagesegmentationtolocatetheforegroundobjectsinscene.Inthiscase,themosteffectivebackgroundsubtractionalgorithmsofthestateoftheartarecomparedtofindthemostsuitableforourexpertvideo-surveillanceapplication.Afterthesegmentationstage,thedetectedblobsmayrepresentfullorpartialpeoplebodies,thus,wehaveimplementedanovelblobfusiontechniquetogroupthepartialblobsintothefinalhumantargets.Then,wecontributeaninnovativetrackingalgorithmwhichisnotonlybasedonpeopletrajectoriesasthemostpartofstate-of-the-artmethods,butalsoonpeopleappearanceinocclusionsituations.Thistrackingiscarriedoutemployinganewtwo-stepmethod:
(1)thedetections-to-tracksassociationissolvedbyusingKalmanfilteringcombinedwithanown-designedcostoptimizationfortheLinearSumAssignmentProblem(LSAP);
and
(2)theocclusionmanagementisbasedonSVMkernelstocomputedistancesbetweenappearancefeaturessuchasGCH,LBPandHOG.Theapplicationofthesethreefeaturesforrecognizinghumanappearanceprovidesagreatperformancecomparedtootherdescriptiontechniques,becausecolor,textureandgradientinformationareeffectivelycombinedtoobtainarobustvisualdescriptionofpeople.Finally,theresultanttrajectoriesofpeopleobtainedinthetrackingstageareprocessedbyourexpertvideo-surveillancesystemforanalyzinghumanbehaviorsandidentifyingpotentialshoppingmallalarmsituations,asareshopentryorexitofpeople,suspiciousbehaviorssuchasloiteringandunattendedcashdesksituations.Withtheaimofevaluatingtheperformanceofsomeofthemaincontributionsofourproposal,weusethepubliclyavailableCAVIARdatasetfortestingtheproposedtrackingmethodwithasuccessnearto85%inocclusionsituations.Accordingtothisperformance,wecorroborateinthepresentedresultsthattheprecisionandefficiencyofourtrackingmethodiscomparableandslightlysuperiortothemostrecentstate-of-the-artworks.Furthermore,thealarmsgivenoffbyourapplicationareevaluatedonanaturalisticprivatedataset,whereitisevidencedthatourexpertvideo-surveillancesystemcaneffectivelydetectsuspiciousbehaviorswithalowcomputationalcostinashoppingmallcontext.
Keywords:
Video-surveillanceinshoppingmalls;
Backgroundsubtraction;
Humantracking;
1.Introduction
Inthecurrentworld,surveillancehasbecomeanessentialelementinalotofdailyactivitiestoguaranteehumansecurityandpropertyandassetsprotection.Itispresentinallkindoflocations:
banks,prisons,airports,parkinglots,petrolstations,storesandanyimaginablebusinessorenterprise.Duetothis,itexistsanincipientneedrelatedtoautomatingcertainsurveillancetasksforassistingsecurityofficersandallowthemdeveloptheirworkinamoreefficientway.
Nowadays,thevideoimagescapturedfromcamerasstrategicallylocatedaretheprincipalelementinanysurveillancesystem.Forthisreason,computervisionprocessingcanbeemployedforextractingusefuldatafromthesevideosandreasoningthisinformationoutwiththeaimofautomatingseveralvideo-surveillancetasksbygivingoffalarmswhenriskeventsaredetected.
Thispaperisfocusedonaspecificcaseofvideo-surveillance:
todetectpotentiallysuspicioushumanbehaviorsinshoppingmalls.Inthisscenario,therearesomeparticularsituationswhichmustbeanalyzed,suchasstoreentryorexit,loiteringeventsthatcanculminateinatheftorsituationswhereacashdeskisunattended,asshownin
Fig.1
Beforegivingoffthedescribedalarms,therearesomepreviousstageswhichmustbeconsideredinordertolocateandfollowthepotentiallysuspiciouspeopleinthevideos,asshownin
Fig.2.Thispreviousprocessstartsbymakinganimagesegmentationusingbackgroundsubtractiontechniques.Acomparisonamongthem
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 购物中心 视频 实时 监控 外文 文献 参考资料
![提示](https://static.bdocx.com/images/bang_tan.gif)