Summary: | SeaiceedgemapsderivedfromSentinel-1SARdualpolarisationEWimagesusingaSupportVectorMachine(SVM)algorithm.ThisalgorithmisbasedonaSVMapproach,andinadditionusestexturecalculationandprincipalcomponentanalysis(PCA)toclassifyseaicetypes(Korosovetal.,2016).Themainstepsofthealgorithmsinclude:(1)pre-processingoftherawSARdata,(2)calculationoftexturefeatures,(3)unsupervisedpre-classificationoftheimageusingPCAandk-meansclusteranalysistoreducethenumberoficeclasses,(4)expertre-classificationoftheimageintothepre-calculatedclasses,(5)trainingoftheSVMusinginputfromthepreviousstep,and(6)classifyingthefullimageintothereducednumberofclassesusingthetrainedSVM.Togenerateaniceedgeproduct,theSVMalgorithmisusedwithonlytwoclasses:seaiceandopenwater. Korosov,A.,N.Zakhvatkina,A.Vesman,A.Mushta,andS.Muckenhuber,SeaiceclassificationalgorithmforSentinel-1images,PosteratESALivingPlanetSymposium2016,Prague,CzechRepublic,9-13may,2016.
|