资源描述：
《3 SAR AND OPTICAL IMAGES REGISTRATION USING SHAPE CONTEXT》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、SARANDOPTICALIMAGESREGISTRATIONUSINGSHAPECONTEXT1,213LeiHuangIEEEStudentMember,ZhenLi,RuiZhang1CenterforEarthObservation&DigitalEarth,ChineseAcademyofSciences,China2GraduateUniversityofChineseAcademyofSciences,China3DepartmentofGeographyandGeoinformationScience,CollegeofScienceˈGeorgeMasonUnivers
2、ity4400UniversityDrive,Fairfax,VA22030,USAABSTRACTsensing[3].PatchfeaturesareusedinSpotandERS-1imagesmatching[4].Inurbanareas,linefeaturescanbeImageregistrationistheprocessofoverlayingtwoorextractedfromroadsandbematchedwithModifiedmoreimagesofthesamescenetakenatdifferenttimes,IteratedHoughTransfo
3、rm[5].In[6],spatialrelationsfromdifferentviewpoints,and/orbydifferentsensors.andorganicallyfeaturesimilaritywerecombinedasaAnovelfeature-basedmulti-sensorimageregistrationmatrix,anditsglobalmaximumwasassumedtobesystemisdeveloped.Thesystemconsistsoftwonewreachedwhentwoimagesmatchedwellwitheachothe
4、r.points:first,edgefeaturesareextractedfromimages,andGenerally,thefeaturematchingstrategiesmentionedthefeaturesaredilatedtosuppresssomecertainkindsofaboverelyonstrongfeatures,suchasspecialpoints,noisearisingfromgroves;second,thepreprocessedislands,closedlakesorstraightlines.Inthispaper,anfeatures
5、arematchedusingtheimprovedshapecontext.effectivemethodtoenhanceedgefeaturesisemployed,Theshapecontexthasbeenfoundtoberobustinhandandabroadlysuitablealgorithmforregistrationnamedwrittendigitandobjectrecognition,andnowitisshapecontextisintroduced,evenforweakfeatureareas.introducedintoremotesensingi
6、magematchingafterThestudyareasareflatfarmland,inSichuanProvincesomeadjustments.ThedevelopedsystemissuccessfullyofChina,andtheusualfeatureisthefieldridge.ThedataappliedtoregisterairborneopticalandC-bandSARcomefromairborneopticalandmulti-bandSARimages.imagesinourexperiments,andtheresultsdemonstrate
7、itsrobustnessandaccuracy.2.METHODSIndexTerms:Imageregistration,edgefeatures,featureFeature-basedmethodsbasicallyconsistoffoursteps:dilation,shapecontext,airborne.featureextraction,featurematching,transformmodelestimati
