硕士论文-复杂背景下目标搜索跟踪方法研究及DSP实现

华中科技大学硕士学位论文复杂背景下目标搜索跟踪方法研究及DSP实现姓名：阎青申请学位级别：硕士专业：模式识别与智能系统指导教师：张桂林20050510IFPGA()TMS320C6202BDSPDSP+FPGAFPGADSPDSPDSPCDSPIIAbstractVideoTargetDetectionandTrackingSystemplaysagreatroleinthefieldofdigitalimageprocessing,andiswidelyusedincivilandmilitaryaffairs,suchasimagevisionpresentiment,navigation,trafficmanageandguidedmissile.Conventionalmethodstendtofailinpresenceofirregularitiesandill-conditionsbackground,butamodelbyutilizinggradientinformationintheformoforientationcodessearchingandmatchingisrobustagainstsuchirregularities.Themajorworksofthisthesisincluding,1.reviewtheConventionalmethodsontheobjectsearchingandtracking.2.introduceandanalyzetheorientationcodemethod.3.introduceanimageprocessingsystemwhichchosenahigh-performancefloating-pointDSP-TMS320VC6202BasthecentralprocessorproducedbyTexasInstrumentInc.ThehardwaredesignofthesystemandtherealizationoforientationcodewithFPGAareexpatiatedindetail.4.objectsearchingandtrackingsoftwaredesignmethodbasedonDSPareexpatiated.BasedontheConventionalobjectsearchingandtrackingmethods,emphasisintroduceandanalyzetheorientationcodemethodindetail,andthedefine,histogram,characteranddissimilaritymeasureonorientationcodemethodisgiven,andthentherobustnessofthisarithmeticundertheclutterbackgroundsuchaspartialocclusion,partiallyshading,backgroundvariations,targetdeformationoranycombinationoftheseirregularitiescausemisregistrationisanalyzed.ExperimentalresultsforrealworldscenescomparedwithConventionalarithmeticdemonstratetheadvantageofthisarithmetic.Throughexperiments,itisknowthatthismethodisbetterthanthatofclassiccorrelationalgorithm.Andthen,introducethesystemhardwareplatformandexpatiatethemethodofhowtoobtaintheOrientationCodewithFPGAindetail.Finally,thepaperdiscusstheDSPimplementationofthissystem.Insoftwaredesignfield,thegeneralmethodsandprogrammingflowsinDSPsoftwaredesignanddebugareanalyzed.Theimageprocessingsoftwarewithorientationcodealgorithm,correlationalgorithmcontrastalgorithmdesignflowchartindetailisgiven.Inaddition,thepaperIIIanalyzesinspecialsomepivotaltechniquesandproblemsduringthesoftwaredesign.Experimentalresultsdemonstratethealgorithmrealizationofautomaticdispatcherandhasverygreatupgradingpotentiality.Keywords:SearchingTrackingorientationcodemethodcorrelationDSP2005______20052005111.1:1:();2:;3:;4TMS320C6202BDSP1.2[1]2,[2]1-11212345334PC51-21-21.3DSPDSP422.12.2TMS30C6202B[3][4][5];;:1111.(,)Xc=(,)MNxyMNxyxfxyfxy====∑∑∑∑1111.(,)Yc=(,)MNxyMNxyyfxyfxy====∑∑∑∑(2.1)XcYcMN2.3[6][7][8]5:::1()2[9][10][11]()[12],3:6[13]ABSabsoluleBalanceSearch1[14]ABS:∑∑==++−×=MmNnjnimFnmTNMjiMSE112)),(),((1),((2.2)∑∑==++−×=MmNnjnimFnmTNMjiMAD11|),(),(|1),((2.3)11(,)(,)MNmnPDCijFij===∑∑(2.4)1|(,)(,)|(,)0:otherwiseTmnFminjtFij:−++≤=tt2[15]NC(Normalized7Correlation)GsMsNsGrMrNrMsMrNsNruvMsNsGs),(vup21121111]),([)],(),([]),()],(),([),(sMiNjsrMiNjrssrMiNjrGjiGvuGvjuiGGjiGvuGvjuiGvup−−++−×−++=∑∑∑∑∑∑======(2.5)1111221111(,)(,)(,)(,)(,)MNsrMNijMNMNijsrijijGuvGiujvpuvGuvGiujv========++=++∑∑∑∑∑∑∑∑(2.6)),(vuGrsG),(vuGrsGNCNCNCSSDA)3.XY48()f(m,n)r0r1...rk-1,:P(ri)=i/.i0,1,2…k-1,)1(10=∑−=kiirPP(r),k:drrPrmkk)(∫=(2.7)kdrrPrrkk)()(∫−=µ(2.8)01mmr=k:10kkkµηµ+=(2.9):2121ηη=f2132ηηη=f2243ηη=f(2.10):9}{minkkiee=(2.11)1332211||/()kjkjjjjefff===−∑∑k=1,2,...,S,Sjfjkfki5[16][17]Hu[18]777Zernike[19][20]6[20][21]),(yxa),(yxb),(vuA),(vuB,:),(),(00yyxxayxb−−=(2.12):)(200),(),(vyuxjevuAvuB+−=π(2.13):10*00(,)(,)(,)exp(2())|(,)||(,)|TAuvBuvQuvjuxvyAuvBuvπ==+(2.14)),(00yx[22]:)cossin,sincos(),(0000000yyxxyxayxb−+−−+=θθθθ(2.15):)cossin,sincos(),(0000)(200θθθθπvuvuAevuBvyuxj+−+=+−(2.16):)cossin,sincos(),(),,(θθθθθvuvuAvuBvuQR+−+=(2.17)),(vuA0θθ=[23][24]Hausdorff[25][26][27][28][29],2.4()11.[30]()640x48064x6430100(,)(,)(,)UVuvRxyMuvSxuyv===−++∑∑MSumSSum00(,)UVMuvSumMuv===∑∑00(,)UVSuvSumSuv===∑∑(2.18)(,)MSRxySumSum≥−MSSumSum−minR(,)RxyminR2[31]SSDASequentialSimilarityDetectionAlgorithBarneaHFSilverman1972[32]12SSDASSDA:(,,,)|((,))((,)(,))|kkkkkkijmnTmnTFminjFijε=−−++−(2.19)∑∑==×=MmNnnmTNMT11),(1∑∑==++×=MmNnkkjnimFNMjiF11),(1),(TkSSDA:11(,){|min[(,,,)]}rkkhrMNkIijrijmnTε≤≤≤==≥∑2.20)I(i,j)(i,j)1.2,A,B,CA,BTC2-2hT2-313SSDA2)1(+−MNm),(jiI(ji)ji,hTSSDA),(),(),(),(|),(),(|nmTjnimFTjiFTjiFnmTjnimF⊕++=+=−++(2.21):⊕(2)),(),(),(11nmTjnimFjiDMmNn⊕++=∑∑==(2.22)HammingDSSDA3SequentialHierarchicalSceneMatchingR.Y.WongE.L.HallSSDA1422KKLTSSDA2.51533.13-1abcd3-116[33][34]3.2,3-2[35][36]21321005025108035453-217f(x,y),xffx∂∇=∂yffy∂∇=∂(,)(1,)xffxyfxy∇=−−(3.1)(,)(,1)yffxyfxy∇=−−(3.2)3-3f(x-1,y-1)f(x,y-1)f(x+1,y-1)f(x-1,y)f(x,y)f(x+1,y)f(x-1,y-1)f(x,y-1)f(x+1,y-1)3-3SobelSOBEL121101000202121101Sobel(1,1)2(,1)(1,1)(1,1)2(,1)(1,1)xfxyfxyfxyfxyfxyfxy∆=−++++++−−−−−−+−(3.3)(1,1)2(1,)(1,1)(1,1)2(1,)(1,1)yfxyfxyfxyfxyfxyfxy∆=−−+−+−+−+−−+−++(3.4)1tan/yxθ−=∆∆,θN(=2π/θ∆)θ∆18,::xyijijffCLotherθθ∇+∇Γ=∆(3.5)Γ,ijCLΓΓ3-401234567891011121314150/8/43/8/25/83/47/89/85/411/83/213/87/415/83-4KirschSobelSobelsobel16π/8,Γ10L2558bit3-53-6193-53-63.3TI)),(),,((1,Im,Im,jijidMnSmOOTnn∑=(3.6),mnIOTOM(,)mn(.)dmin{,(,)max/2elseabNababNdabE−−−−(3.7)Em