数据同化框架下基于差分进化的遥感图像融合(3)

G61.671420.81997.679031.39990.1539B

62.698020.75847.454531.8117

0.2254

当式(5)中的a1,a2,···,an取不同的组合,会

有不同的结果,总之是那些大权重所对应的属性指标会得到较大的提高,而小权重对应的属性指标会得到较小的提高,有时小权重所对应的指标还会有所下降.各个属性指标的权重大小不仅要考虑各属性指标对后续处理影响的程度,还要考虑到各个属性指标值的大小,也就是各个属性指标值要进行归一化处理,不同的归一化处理方法对最终结果也会有一定的影响.

4结论与展望

本文的方法在一定程度上解决了现有融合模型的建立不够客观,参数选取随意性大,不易根据后续处理目的对融合规则进行自适应调整的缺点,也解决了不同融合方法的优点不易综合的问题,能够更容易地获取合适的结果图像.

选取合适的融合方法作为数据同化系统的模型算子和观测算子是一个难点和重点,合适的模型算子和预测算子有利于获取好的预测数据和观测数据,而好的预测数据和观测数据有助于最终合适数据的获取[16 17].如何在机理上和实验上分析各度量指标的相互关系,综合考虑多个属性对图像后续处理的协同影响,选取合适的指标权重来构造合适的目标函数,选择合适的优化方法以充分利用所选模型算子和观测算子的各自优点,以及满足特定的后续应用需要,是下一步研究的重点.

References

1Gonzalez-AudicanaM,SaletaJL,CatalanRG,GarciaR.Fusionofmultispectralandpanchromaticimagesusingim-provedIHSandPCAmergersbasedonwaveletdecomposi-tion.IEEETransactionsonGeoscienceandRemoteSensing,2004,42(6):1291 1299

2WuJ,LiuJ,TianJW,YinBK.Wavelet-basedremotesensingimagefusionwithPCAandfeatureproduct.In:Proceedingsofthe2006InternationalConferenceonMecha-tronicsandAutomation.Luoyang,China:IEEE,2006.2053 2057

3ChenMi,XuanJian-Hui,LiDe-Ren,QinQian-Qing,JiaYong-Hong.Imagefusionalgorithmbasedonindependentcomponentanalysis.Opto-ElectronicEngineering,2007,34(6):82 87

(陈蜜,伭剑辉,李德仁,秦前清,贾永红.独立分量分析的图像融合算法.光电工程,2007,34(6):82 87)

4MiaoQi-Guang,WangBao-Shu.Multi-sensorimagefusionbasedonimprovedLaplacianpyramidtransform.ActaOp-ticaSinica,2007,27(9):1605 1610

(苗启广,王宝树.基于改进的拉普拉斯金字塔变换的图像融合方法.光学学报,2007,27(9):1605 1610)

5TangL,ZhaoZG.Multiresolutionimagefusionbasedonthewavelet-basedcontourlettransform.In:Proceedingsofthe10thInternationalConferenceonInformationFusion.Quebec,Canada:IEEE,2007.184 189

6WangYue-Shan.Dataassimilation—itscause,itsmeaningandmainprocedures.MarineForecasts,1999,16(1):11 20(王跃山.数据同化—它的缘起、含义和主要方法.海洋预报,1999,16(1):11 20)

alos

7BachH,MauserW.Methodsandexamplesforremotesensingdataassimilationinlandsurfaceprocessmodel-ing.IEEETransactionsonGeoscienceandRemoteSensing,2003,41(7):1629 16378HuangChun-Lin,LiXin.Areviewoflanddataassimilationsystem.RemoteSensingTechnologyandApplication,2004,19(5):424 430

(黄春林,李新.陆面数据同化系统的研究综述.遥感技术与应用,2004,19(5):424 430)9LiX,ToshioK,putersandGeosciences,2004,30(3):239 24810ChenMi.ResearchonMethodsandApplicationsofImage

DataFusionBasedonIndependentComponentAnalysis[Ph.D.dissertation],WuhanUniversity,China,2006

(陈蜜.基于独立分量分析的影像信息融合方法与应用研究[博士学位论文],武汉大学,中国,2006)11MitianoudisN,StathakiT.Optimalcontrastforcolorim-agefusionusingICAbases.In:Proceedingsofthe11thInternationalConferenceonInformationFusion.Cologne,Germany:IEEE,2008.1 712LiuBo,WangLing,JinYi-Hui.Advancesindi erentialevo-lution.ControlandDecision,2007,22(7):721 729

(刘波,王凌,金以慧.差分进化算法研究进展.控制与决策,2007,22(7):721 729)13daCunhaAL,ZhouJP,DoMN.Thenonsubsam-pledcontourlettransform:theory,design,andapplications.IEEETransactionsonImageProcessing,2006,15(10):3089 310114LiuGui-Xi,ChenWen-Jin,YangWan-Hai.Studyonthein- uenceoffusionparametersontheperformanceofcontrastpyramiddecomposition-basedimagefusionscheme.JournalofCircuitsandSystems,2006,11(1):39 45

(刘贵喜,陈文锦,杨万海.融合参数对对比度塔形分解图像融合方法性能的影响研究.电路与系统学报,2006,11(1):39 45)15DeAA,MoschittaA,RussoF,CarboneP.Imagequality

assessment:anoverviewandsomemetrologicalconsidera-tions.In:Proceedingsofthe2007IEEEInternationalWork-shoponAdvancedMethodsforUncertaintyEstimationinMeasurement.Trento,Italy:IEEE,2007.47 5216ChenRY,LiS,YangR,QinQQ.Multi-focusimagesfusion

basedondataassimilationandgeneticalgorithm.In:Pro-ceedingsofthe2008InternationalConferenceonComputerScienceandSoftwareEngineering.Wuhan,China:IEEE,2008.249 25217ChenRY,LiS,YangR,QinQQ.Remotesensingim-agefusionbasedondataassimilationandgeneticsimulatedannealingalgorithm.In:Proceedingsofthe2008Interna-tionalSymposiumonInformationScienceandEngineering.Shanghai,China:IEEE,2008.520 524

陈荣元武汉大学测绘遥感信息工程国家重点实验室博士研究生,湖南商学院信息学院讲师.主要研究方向为图像处理、信息融合和人工智能.本文通信作者.E-mail:chenrongyuan@(CHENRong-YuanPh.D.candi-dateattheStateKeyLaboratoryofInformationEngineeringinSurveying,

MappingandRemoteSensing,WuhanUniversity.Lec-turerintheDepartmentofInformation,HunanUniversityofCommerce.Hisresearchinterestcoversimageprocess-ing,informationfusion,andarti cialintelligence.Corre-spondingauthorofthispaper.)

林立宇武汉大学测绘遥感信息工程国家重点实验室讲师.主要研究方向为图像处理、信息融合和超分辨率.E-mail:foxery@(LINLi-YuLecturerattheStateKeyLaboratoryofInformationEngi-neeringinSurveying,MappingandRe-moteSensing,WuhanUniversity.His

researchinterestcoversimageprocessing,informationfu-sion,andsuperresolution.)

王四春湖南商学院信息学院教授.主要研究方向为人工智能和智能决策.E-mail:wscabcbank@

(WANGSi-ChunProfessorattheCollegeofInformation,HunanUniver-sityofCommerce.Hisresearchinterestcoversarti cialintelligenceandintelli-gentdecision.)

秦前清武汉大学测绘遥感信息工程国家重点实验室教授.主要研究方向为信息处理和小波分析.

E-mail:qqqin@(QINQian-QingProfessorattheStateKeyLaboratoryofInformationEngineeringinSurveying,MappingandRemoteSensing,WuhanUniversity.

Hisresearchinterestcoversinformationprocessingandwavelettransform.)