文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > IT/计算机 > Moving Object Detection for Real-Time Applications
侵权投诉

Moving Object Detection for Real-Time Applications

MovingObjectDetectionforReal-TimeApplications

LuciaMaddalenaNationalResearchCouncilInstituteforHigh-PerformanceComputingandNetworking

ViaP.Castellino111,80131Naples,Italy

lucia.maddalena@na.icar.cnr.it

Abstract

Detectionofmovingobjectsinvideostreamsisthe rstrelevantstepofinformationextractioninmanycomputervisionapplications.Asidefromtheintrinsicusefulnessofbeingabletosegmentvideostreamsintomovingandback-groundcomponents,detectingmovingobjectsprovidesafo-cusofattentionforrecognition,classi cation,andactivityanalysis,makingtheselaterstepsmoreef cient.

Wepresentsomeextensionstothemethodformovingob-jectdetectionpresentedin[4].Ourmaincontributionsarerelatedtothepre-processingofintermediateresults(tran-siencemaps),aimedatenhancingtheaccuracyofdetec-tionresults,andtotheparallelizationofsomeofthemostcomputationallyintensivestepsusingSSE2instructions,inordertoenhanceef ciencyandallowforreal-timeapplica-tions.

1.Introduction

Detectionofmovingobjectsinvideostreamsisthe rstrelevantstepofinformationextractioninmanycomputervisionapplications,includingtraf cmonitoring,automatedremotevideosurveillance,andpeopletracking[3].Asidefromtheintrinsicusefulnessofbeingabletosegmentvideostreamsintomovingandbackgroundcomponents,detect-ingmovingobjectsprovidesafocusofattentionforrecog-nition,classi cation,andactivityanalysis,makingtheselaterstepsmoreef cient,sinceonlymovingpixelsneedbeconsidered[2].Theproblemisknowntobesigni cantanddif cult[8].Conventionalapproachestomovingobjectde-tectionincludetemporaldifferencing[7],backgroundsub-traction[8],andoptical ow[1].

Temporaldifferencingtakesintoaccountdifferencesinconsecutivesequenceframes,whichallowtodiscernstaticobjects(havingnulldifferences)frommovingobjects(hav-ingnon-nulldifferences).Thisapproachisveryadaptiveto

14th International Conference on Image Analysis and Processing (ICIAP 2007)

Moving Object Detection for Real-Time Applications

0-7695-2877-5/07 $25.00 © 2007

AlfredoPetrosino

UniversityofNaplesParthenopeDepartmentofAppliedScienceViaA.DeGasperi5,80133Naples,Italyalfredo.petrosino@uniparthenope.it

dynamicenvironments,butitisstrictlydependentontheve-locityofmovingobjectsinthesceneanditissubjecttotheforegroundapertureproblem.Incontrast,optical owtech-niquesaimatcomputinganapproximationtothe2Dmo-tion eld(projectionofthe3Dvelocitiesofsurfacepointsontotheimagingsurface)fromspatio-temporalpatternsofimageintensity[1].Theycanbeusedtodetectindepen-dentlymovingobjectsinthepresenceofcameramotion,butmostoptical owcomputationmethodsarecomputa-tionallycomplex,andcannotbeappliedtofull-framevideostreamsinreal-timewithoutspecializedhardware.

Surelybackgroundsubtractionisthemostcommonandef cientmethodtotackletheproblem(e.g.[6]).Itisbasedonthecomparisonofthecurrentsequenceframewithareferencebackground,includinginformationonthescenewithoutmovingobjects.Itisindependentonthevelocityofmovingobjectsanditisnotsubjecttotheforegroundaper-tureproblem,butitisextremelysensitivetodynamicscenechangesduetolightingandextraneousevents.Althoughtheseareusuallydetected,theyleavebehindholeswherethenewlyexposedbackgroundimagerydiffersfromtheknownbackgroundmodel(ghosts).Whilethebackgroundmodeleventuallyadaptstotheseholes,theygeneratefalsealarmsforashortperiodoftime.

Amongdifferentapproaches,theoneproposedin[4]al-lowsdisambiguationofmovingobjectsthatstopforawhile,areoccludedbyotherobjects,andthenresumemotion.Layereddetectionisbasedontwoprocesses:pixelanaly-sisandregionanalysis.Pixelanalysisdetermineswhetherapixelisstationaryortransientovertime,whileregionanalysisdetectsstationaryregionsofstationarypixelscor-respondingtostoppedobjects.Weadoptedthelayeredapproach,includingapre-processingoftransiencemapsaimedatsuppressingshadowsandreducingnoise,andclus-teringnon-backgroundpixelsusingregiongrowing.

Moreover,theneedforreal-timesystemsimposesverylowcomputationtimes.Wefocusedonmostcomputation-allyintensivestepsoftheproposedapproach,obtainingpar-

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email