视频监控中基于行人跟踪的摄像机自动标定

整理文档很辛苦,赏杯茶钱您下走!

免费阅读已结束,点击下载阅读编辑剩下 ...

阅读已结束,您可以下载文档离线阅读编辑

资源描述

©1994-2009ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.()JTsinghuaUniv(Sci&Tech),20094942009,Vol.49,No.4w33http:ööqhxbw.chinajournal.net.cn,,,,(,100084):2008201208:(60573167);(2006AA01Z118):(1983),(),,:,,E2mail:yangshq@mail.tsinghua.edu.cn:,,,,RANSAC,:,:;;;;;RANSAC:TP391.4:A:100020054(2009)0420586204Cameraauto-calibrationbyvideosequenceofsinglewalkingpedestrianinsurveillancevideoCUIYibo,TANXiaoyu,CUIPeng,SUNLifeng,YANGShiqiang(DepartmentofComputerScienceandTechnology,TsinghuaUniversity,Beijing100084,China)Abstract:Asinglecamerabasedauto2calibrationmethodispresentedtoobtainthecameraparametersfromavideosurveillancesequence.Thehorizontalvanishinglineandtheverticalvanishingpointarecomputedfromtheextractedhead2footpositionsofthepedestriancontours,withalltheparametersthendeducedfromtheprincipalpointandthevanishingpoint.ThefindresultisoptimizedbasedontheRANSACmethod.Testsshowthatthismethodisrobusttovariouscircumstancesandpreciseenoughtobeappliedtootherresearchfields.Keywords:surveillancevideo;calibration;head2footposition;vanishingpoint;vanishingline;RANSAC,[1],[2],2,[3]Bayesian,[4],,;,,-[5],[4],,2:1),,;2),11.1,pinhole,:©1994-2009ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.{u,v,1}TµPõ{X,Y,Z,1}T.(1)P533,:P=fsuP0fõavP001intrinsic(I33O31)õR33O31O131rotationI33T31O131translation,(2)intrinsicrotationtranslationa=1,s=0,1.21.2.1VYVLnWh={Ph1,,Phn}Wf={Pf1,,Pfn}PhPf,VY;PhiPhjLhPfiPfjLfPxVL,[5](MLE)VYVL:Px8P(k)xP(l)x,2L(3)(dist(Pi,L)PiL),LVLL=argminPi8dist(Pi,L).(3)1.2.2VYVL:Y=0,XZ,XZVLVY,FengjunLv[4]Caprile[7],XZVXVZPp[8],VXVZVYPp[7],(4)(H,[4])HC(Y):HHC=1-d(x,c)d(x,v)(d(x,c)d(x,v)).(4)(xo,yo),(xoyo1)TµPõ(0001)T(2):(TXTYTZ)Tµ(AõR)-1õ(xoyo1)T.(5)TY=HC,2TXTZ(5)22.1,2:;PhPf[9]kerneldensityestimator[6],,{(x1,y0)(xn,y0)},,Ph:xtop-pt=1nni=1xi,ytop-pt=y0.(6){(a1,b)(am,b)},Pf:xbottom-pt=12(aleft-pt+aright-pt),ybottom-pt=b.(7)PhPfC,CF,2.2,,RANSAC,F:1)FN(3N5),F,FVY;2)FPfPC,,3C(3),{Ci,Cj,Ck}=argmaxPfiCi,PfjCj,PfkCkijk(dist(Pfi,Pc)+dist(Pfj,Pc)+dist(Pfk,Pc)).(8)VLCPxF(8)11F3)Eerr(3);785,:©1994-2009ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.)1),1)3)T,Eerr2C,LhLf,VLPx,VL(8),F,2C,PxVL3,(2)PP,F{(uh,vh),(uf,vf)},,YYf0,XfZf1µp11p13p14p21p23p24p31p33p34-1ufvf1.(9),XZYH,(Xf,H,Zf)(1)(uh,vh)(uh,vh)dFCdEerr,,Eerr=CiF(uhi-uhi)2+(vhi-vhi)2.(10)44400300,VC++IntelOPENCV,21()(),21öö()öm1469.2313.772.6332497.2313.582.7323430.3114.232.597,:3(uf,vf),(9)(Xf,0,Zf),H(Xh,H,Zh),(uh,vh)(uf,vf)vh-vf,h,3,3H,(4),,ahb4,PH(9)(Xf,0,Zf),XhXfZhZf,Xh=XfZh=Zf,uhvh1=m11Xf+m13Zf+m14m12m21Xf+m23Zf+m24m22m31Xf+m33Zf+m34m32(XXH)T.(11)H,Ph(uh,vh,1))Phs=(uh-uh)2+(vh-vh)2,H5(H1H1H,H2H2H):885()2009,49(4)©1994-2009ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.=Xf,Z=Zf(Xf,H,Zf),(uh,vh)H,(Xf,H,Zf)(uh,vh)(s),H2(6,):5cm,,,565,RANSAC,;,,:,,,,(References)[1]ZhangZ.Aflexiblenewtechniqueforcameracalibration[J].IEEETransonPatternAnalysisandMachineIntelligence,2000,11(11):1330-1334.[2]KongD,GrayD,TaoHai.Aviewpointinvariantapproachforcrowdcounting[C]ööIEEEProcInternationalConferenceonPatternRecognition.HongKong,2006.[3]KrahnstoeverN,MendoncaPRS.Bayesianauto2calibrationforsurveillance[C]ööProceedingsofthe10thIEEEInternationalConferenceonComputerVision:Beijing,China,2005.[4]LvF,ZhaoT,NevatiaR.Self2calibrationofacamerafromvideoofawalkinghuman[C]ööIEEEProcInternationalConferenceonPatternRecognition:Quebec,Canada,2002:562-567.[5]LvF,ZhaoT,NevatiaR.Cameracalibrationfromvideoofawalkinghuman[J].IEEETransonPatternAnalysisandMachineIntelligence,2006,9(28):1513-1518.[6]ElgammalA,HarwoodD,DavisL.Non2parametricmodelforbackgroundsubtraction[C]ööFRAME2RATEWorkshop.IEEEpress,1999:1151-1163[7]CaprileB,TorreV.Usingvanishingpointsforcameracalibration[J].InternationalJournalofComputerVision,1990(4):127-140.[8]CipollaR,DrummondT,RobertsonD.Cameracalibrationfromvanishingpointsinimagesofarchitecturalscenes[C]ööProceedingsofBritishMachineVisionConference.UniversityofNottingham,1999(2):382-391.[9]MittalA,ParagiosN.Motion2basedbackgroundsubtractionusingadaptivekerneldensityestimation[C]ööIEEEProcComputerSocietyConferenceonComputerVisionandPatternRecognition.WashingtonDC,2004:302-309.985,:

1 / 4
下载文档,编辑使用

©2015-2020 m.777doc.com 三七文档.

备案号:鲁ICP备2024069028号-1 客服联系 QQ:2149211541

×
保存成功