灰色预测和混沌PSO的红外小目标检测

第6卷第2期Vo.l6.220114CAAITransactionsonIntelligentSystemsApr.2011do:i10.3969/.jissn.16734785.2011.02.006PSO尹丹艳1,吴一全1,21南京航空航天大学信息科学与技术学院,江苏南京210016;2光电控制技术重点实验室,河南洛阳471000):,PSO.GM(1.1),,;(particleswarmoptimization,PSO),.:,.:;;;;:TP18;TN911.73:A:16734785(2011)02012606ThedetectionofasmallinfraredtargetbasedongraypredictionandchaoticPSOYINDanyan1,WUYiquan1,2(1SchoolofInformationScienceandTechnology,NanjingUniversityofAeronauticsandAstronautics,Nanjing210016,China;2ScienceandTechnologyonElectroopticControlLaboratory,Luoyang471009,China)Abstract:Byanalyzingthecharacteristicsofasmalltargetandthebackgroundofaninfraredimage,adetectionmethodofaninfraredsmalltargetbasedongraypredictionandchaoticparticleswarmoptimization(PSO)wasproposed.Firs,ttheGM(1,1)modelofgraysystemtheorywasadoptedtopredicttheinfraredimagebackgroundinatimedomain.Thesubtractionofthesourceimageminusthepredictedimagegavetheresidualimage.Asaresul,tthebackgroundwassuppressedandthetargetwasenhanced.Then,atwodimensionalhistogramobliquesegmentationmethodbasedonchaoticPSOandfuzzymaximumentropywaspresented.Theresidualimagewassegmentedbythismethod,leadingtothedetectionofthesmalltarge.tTheexperimentalresultsshowthattheproposedmethodcansignificantlyincreasethedetectionprobabilityofaninfraredtargettoachievelongrangesmalltargetdetection.Keywords:detectionofinfraredsmalltarge;tbackgroundsuppression;grayprediction;fuzzymaximumentropy;chaoticparticleswarmoptimization:20100421.:(60872065);(20105152026);(KFKT2010B17).:.Emai:[e-mail].,,,.,,,,.,,,.,.,,,,.,,..,,[110]..(MF)(MedF)Wiener(WF).,,.,,,..(signalnoiseratio,SNR)(signalclutterratio,SCR),,.:,[11].SNRSCRSNRSCR,;,.[1214],,3:1),4;2),,;3),,.,.,,4.,,,:,,.,.GM(1,1),,;PSO[1516],.1基于灰色预测的红外图像背景时域预测1.1GM(1,1)GM(1,1),,,,.x(0)={x(0)(1),x(0)(2),,x(0)(n)}.(1)(1)x(1)={x(1)(1),x(1)(2),,x(1)(n)}.:x(1)(k)=ki=1x(0)(i).GM(1,1):x(0)(k)+az(1)(k)=b,k=1,2,,n.(2):ab;z(1)(k)=x(1)(k)+(1-)x(1)(k-1)x(0)(k),.,0.5.,:ab=(BTB)-1BTY.:BT=-z(1)(2)-z(1)(3)-z(1)(n)111,Y=[x(0)(2)x(0)(3)x(0)(n)]T.ab:a=nk=2z(1)(k)nk=2x(0)(k)-(n-1)nk=2z(1)(k)x(0)(k)(n-1)nk=2[z(1)(k)]2-(nk=2z(1)(k))2,(3)b=nk=2x(0)(k)nk=2[z(1)(k)]2-nk=2z(1)(k)nk=2z(1)(k)x(0)(k)(n-1)nk=2[z(1)(k)]2-(nk=2z(1)(k))2.(4)ab(2),x(0)GM(1,1):x(0)(k)=1-0.5a1+0.5ak-2b-ax(0)(1)1+0.5a,k=n+1,n+2,(5)1.2:,,,,.,1272,:PSO;(),,,.:,,.,.n.MN,x(0)={x(0)(1),x(0)(2),,x(0)(n)},=1,2,,MN.x(0)(k)(k=1,2,,n)k.:1)x(0)={x(0)(1),x(0)(2),,x(0)(n)};2)x(1)={x(1)(1),x(1)(2),,x(1)(n)};3)z(1)={z(1)(2),z(1)(3),,z(1)(n)};4)(3)(4)ab;5)(5)n+1,n+2,;6)n+1n+2n+1n+2,n+1n+2.n+1n+2.,1)~6),ab.2基于混沌PSO模糊最大熵的残差图像分割:,,,,.h(f,g)=r(f,g)MN,0f,gL-1.r(f,g)f,ssg,MN,L.,,Oo(f,g)(6),Bb(f,g)(7),f,ssgOB,l=f+g,cdo(f,g)+b(f,g)=1,(c,d).o(f,g)=0,2[(l-c)/(d-c)]2,1-2[(l-d)/(d-c)]2,1,0lc;cl(c+d)/2;(c+d)/2ld;dlL-1.(6)b(f,g)=1-(f,g).(7)QH(Q|O)=L-1f=0L-1g=0o(f,g)hfgp(O)logo(f,g)hfgp(O).:p(O)=L-1f=0L-1g=0o(f,g)hfg.QH(Q|B)=-L-1f=0L-1g=0b(f,g)hfgp(B)log(f,g)hfgp(B).:p(B)=L-1f=0L-1g=0b(f,g)hfg.:H=H(Q|O)+H(Q|B)=-L-1f=0L-1g=0o(f,g)hfgp(O)logo(f,g)hfgp(O)+b(f,g)hfgp(B)logb(f,g)hfgp(B).(8)(c,d)(8).n,iXi=(Xi1,Xi2,,Xin)Vi=(Vi1,Vi2,,Vin).,pfiti;.,.t,ipbesti(t),,gbest(t),.t+1,.1286Vi(t+1)=wVi(t)+c1r1[pbesti(t)-Xi(t)]+c2r2[gbest(t)-Xi(t)].vi(t+1),Xi(t+1)=Xi(t)+Vi(t+1).:t;c1=c2=2;r1r2(0,1);ww=wmax-t(wmax-wmin)/tmax,tmax,wmaxwmin,wmax=0.9,wmin=0.1.,[Vmin,Vmax],Vmin=-Vmax=-5,.Logistick+1j=kj(1-kj),k=1,2,j(0,1),j0.25,0.5,0.75.,pbesti(t)=[X1X2XjXn]:pc=Xj,min+k+1j(Xj,max-Xj,min),Xk+1j=(1-g)Xkj+gpc.:g,Xj,g=1-[(-1)/].:,u,u=2.gbest,(c,d)(c,d)*.(c,d)*o(f,g),o(f,g)0.5(f,g)O,B.,,,.3实验结果与分析,LSSVM[17].:=.(a)(4686)(b)LSSVM(4686)(c)(4686)(d)LSSVM(4686)(e)(4686)1Fig.1Resultsofinfraredsmalltargetdetectionmethods300,LSSVM240,80%;279,93%.1(a),,,,,,.1(b)LSSVM,,,.1(c),(b),,LSSVM1292,:PSO,(c).,.1(d)LSSVM(b),46,86,.1(e),2,...:SNR=(Gt-Gb)/b.:Gt,Gb,b.:GSNR-SNRO/SNRI.SNRO,SNRI.SNRI,:GSNR=(SNRO+|SNRI|)/|SNRI|.21.12Table1ComparisonofthetwomethodsSNRISNROGSNR46LSSVM0.42443.0971169.50247.2976399.393086LSSVM0.30090.9771108.88973.2473361.8800,,.,,,,.1,,.4结束语,GM(1,1),;,;,.LSSVM,,.:[1],,,.Gabor[J].,2009,38(4):737741.QINHanlin,LIUShangqian,ZHOUHuixin,eta.lBackgroundsuppressionfordinsmalltargetwithGaborkernelnonlocalmeans[J].InfraredandLaserEngineering,2009,38(4):737741.[2]WUWe,iPENGJiaxiong.Algorithmofinfraredbackgroundsuppressionandsmalltargetdetection[J].InfraredLaserEngineering(S10072276),2003,32(2):148152.[3],,,.MRF[J].,2009,28(6):440444.WANGDabao,LIUShangqian,KOUXiaoming,eta.lInfraredbackgroundcluttersuppressionalgorithmofadaptiveregularizationbasedonMRF[J].JInfraredandMillimeterWaves,2009,28(6):440444.[4],,,.[J].:,2009,36(6):10751078.LIFan,LIUShangqian,HONGMing,eta.lDiminfraredtargetsdetectionbasedonbackgroundprediction[J].JournalofXidianUniversity:NaturalscienceEdition,2009,36(6):10751078.[5]SONIT,ZEIDLERJR,KUWH.Performanceevaluationof2Dadaptivepredictionfiltersfordetectionofsmallobjectsinimagedata[J].IEEETransactionsonImageProcessing,1993,2(3):327340.[6]MICHIELSENK,DeRH.Integralgeometrymorphologicalimageanalysis[J].PhysicsReports,2001,347:461538.[7],.[J].,2008,35(4):1216.WUYiquan,LUOZijuan.Me