您好,欢迎访问三七文档
当前位置:首页 > 商业/管理/HR > 管理学资料 > 基于隔帧差分区域光流法的运动目标检测
:2008-03-24.:(60742005);;(2008GZS0032);(GJJ09183).,,(,330063):,,,,;,,,,,,:;;:TP391:A:1001-5868(2009)02-0300-05MovingTargetsDetectionBasedonRegionalOpticalFlowofDiscontinuousFrameDENGHui2bin,XIONGBang2shu,OUQiao2feng(KeyLab.ofNondestructiveTest,MinistryofEducation,NanchangHangkongUniversity,Nanchang330063,CHN)Abstract:Combiningwiththehighspeedofframe2differencemethodandthehighaccuracyoftheopticalflowmethod,amovingtargetdetectionalgorithmbasedonregionalopticalflowofdiscontinuousframe2differenceisproposed.Afteracquiringtheregionconsistingofmovingtargetsbyusingthediscontinuousframe2differencemethod,onlyitsopticalflowiscalculated.Thecalculatedregionisreduced,thereforeitimprovedthedetectionspeed.Theopticalflowofthelargergradientpointsiscalculatedwiththebasicequation,whilethesmalleronesarecalculatedwiththesmoothconstraintequation,whichcanimprovetheaccuracyofthemovingtargetsdetection.Inaddition,throughimportinginertiafactor,theiterativenumbersoftheopticalflowequationisreducedwhilethedetectionspeedisimproved.Experimentalresultsprovetheeffectivenessoftheproposedalgorithm.Keywords:targetdetection;temporaldifference;localopticalflow0,,,,,[1][1][2][3],,,,,;,,,003SEMICONDUCTOROPTOELECTRONICSVol.30No.2Apr.2009,,,,,,,,,,,,,,,1,,,:(1);(2);(3);(4)1.1,,,,,,55,,,,,[7],,,,:(1)(i,j)W(i,j),5,W(i,j-1)W(i-1,j),1;(2),,:m(1),m(2),,m(5),,m;(3)W(i,j),W(i,j-1)W(i-1,j),W(i,j-1),m(6),W(i-1,j),M(1),M(2),,M(5),;(4),(3),(5);(5),(2),1,,,,11()[7]33120126553006020771176168421.2,,,,,10320094302:,,,:(1),,,;(2),n,fin(x,y),fi(x,y),fin(x,y)n=1,,n=2,1,n=3,2,,,;,;(3)FDb(x,y)FDf(x,y),T,,,(1)[4]Otsu,T;(4)FDb(x,y)FDf(x,y)En(x,y),(2)FDb(x,y)=1|fi(x,y)-fi-n(x,y)|T0|fi(x,y)-fi-n(x,y)|TFDf(x,y)=1|fi+n(x,y)-fi(x,y)|T0|fi+n(x,y)-fi(x,y)|T(1)En(x,y)=FDb(x,y)FDf(x,y)(2)2(a)(b)(c)(d)1234,11,,,,,21.3,,,,Et(x,y):9Et9xuxy+9Et9yxy+9Et9t=0(3),9Et9x,9Et9y,9Et9tE(x,y,t)xytE(x,y,t)t(x,y),(4)(3)(uxy,xy),,,HornSchunk[5],,(5)9Et9x=14{[E(x+1,y,t)-E(x,y,t)]+[E(x+1,y+1,t)-E(x,y+1,t)]+[E(x+1,y,t+1)-E(x,y,t+1)]+[E(x+1,y+1,t+1)-E(x,y+1,t+1)]}9Et9y=14{[E(x,y+1,t)-E(x,y,t)]+[E(x+1,y+1,t)-E(x+1,y,t)]+[E(x,y+1,t+1)-E(x,y,t+1)]+[E(x+1,y+1,t+1)-E(x+1,y,t+1)]}9Et9t=14{[E(x,y,t+1)-E(x,y,t)]+[E(x+1,y,t+1)-E(x+1,y,t)]+[E(x,y+1,t+1)-E(x,y+1,t)]+[E(x+1,y+1,t+1)-E(x+1,y+1,t)]}(4)9Et9xunxy+9Et9yvnxy+9Et9t2+9uxy9x2+9uxy9y2+9vxy9x2+9vxy9y2dx=min(5)(5),203SEMICONDUCTOROPTOELECTRONICSVol.30No.2Apr.2009(3),,,,HornSchunck,,,,,(6):k(x,y)9Et9xunxy+9Et9yvnxy+9Et9t2+9uxy9x2+9uxy9y2+9vxy9x2+9vxy9y2dxdy=min(6),(x,y),(7):(x,y)=19E2t9x+9E2t9yT09E2t9x+9E2t9yT(7),T,,,(6)2,2,(x,y)(n+1)(un+1xy,vn+1xy),,,,,un+1xy=unxy-9Et9x9Et9xunxy+9Et9yvnxy+9Et9t+9Et9x2+9Et9y2+(n)(unxy-un-1xy)vn+1xy=vnxy-9Et9x9Et9xunxy+9Et9yvnxy+9Et9t+9Et9x2+9Et9y2+(n)(vnxy-vn-1xy)(8),(n),(n)=(800-3n)/810,(u0xy,v0xy)=(u-1xy,v-1xy)=(0,0),,1.4,,,,,1.4.1,,[0,255],,nRi,mii,f(x,y)mi,:|f(x,y)-mi|Ti(x,y),(x,y)Ri,Ri(x,y)f(x,y),:Ti(x,y)=1-imiT(9),mii(x,y)Ri,Ri:mi=1n+1[f(x,y)+nmi](10)i1n+1{n2i+[f(x,y)-mi]2}(11)(x,y)|f(x,y)-mi|Ti(x,y),|f(x,y)-mi|Ti(x,y),(x,y)T1.4.2,,,,4282,30320094302:,,,,;,,2,P4(1.8GHz,1GB)C++Builder6.0,3202402.13,3(a)3(b)tt+3,,3(c),,;,,3(d),2[10]2,,321(s)2(s)3(s)200/s1.7961.9221.8751.8282.90.1720.2340.2030.1561.42.2,,0.08,(8)1,T30,8,,4(a)4(b)4(c)4,43,,,,,:[1],.[J].,2007,7(11):9202927.[2],.[J].,2006,27(8):9612964.[3]HoldenEJ,OwensR.SegmentingOccludedObjectsUsingaMotionSnake[C]//AsianConferenceonComputervision(ACCV),2004:SIV23.[4],.2Otsu[J].,2005,4(10):4842488.[5]HornBKP,SchunckBG.Determiningopticalflow[J].ArtificialIntelligence,1981,17(1):1852220.[6]HaritaogluI,HarwoodD,DavisLS.Realtimesurveillanceofpeopleandtheiractivities[J].IEEETrans.onPatternAnalysisandMachineIntelligence,2000,22(8):8092830.[7],.[J].(),2004,22(3):2322235.(307)403SEMICONDUCTOROPTOELECTRONICSVol.30No.2Apr.2009,MATLABRTDX2,DSPBUFFER3202,,RTDXMATLAB,MATLAB,MATLABDSP,ETTIC6000,Simulink,5[4]56(a),6(b)DSP,6(c)MATLAB,6(b)(c),(c)6MATLABETTIC6000DSP5MATLABETTIC6000DSP,MATLABMATLABDSP,,,:[1],,.DSP[M].:,2003.[2],.DSP[M].:,2004.[3],,.TMS320C6000DSP[M].:,2001.[4],,.MATLAB[M].:,2001.:(1982-),,,,E2mail:xft_7912@163.com(304)[8],.[J].,2004,25(6):4622464.[9]LangHong.Multi2resolutionmultiple2modeltargettracking[J].IEEETrans.onAerospaceandElectronicSystems,1994,30(2):5182524.[10]XuHaifeng,YounisAA,KabukaMR.Automaticmovingobjectextractionforcontent2basedapplications[J].IEEETrans.onCircuitsandSystemsforVideoTechnol.,2004,14(6):7962812.:(1980-),,,,E2mail:panni24@126.com70320094302:MATLABDSP
本文标题:基于隔帧差分区域光流法的运动目标检测
链接地址:https://www.777doc.com/doc-637876 .html