摘要
Homologousfeaturepointextractionisakeyproblemintheopticalandsyntheticapertureradar(SAR)imageregistrationforislands.Anewfeaturepointextractionmethodusingathresholdshrinkoperatorandnon-subsampledwavelettransform(TSO-NSWT)foropticalandSARimageregistrationwasproposed.Moreover,thematchingforthisproposedfeaturewasdifferentfromthetraditionalfeaturematchingstrategiesandwasperformedusingasimilaritymeasurecomputedfromneighborhoodcirclesinlow-frequencybands.Then,anumberofreliablymatchedcoupleswithevendistributionswereobtained,whichassuredtheaccuracyoftheregistration.ApplicationoftheproposedalgorithmtoSPOT-5(multi-spectral)andYG-1(SAR)imagesshowedthatalargenumberofaccuratelymatchedcouplescouldbeidentified.Additionally,bothoftherootmeansquareerror(RMSE)patternsoftheregistrationparameterscomputedbasedontheTSO-NSWTalgorithmandtraditionalNSWTalgorithmwereanalyzedandcompared,whichfurtherdemonstratedtheeffectivenessoftheproposedalgorithm.Thealgorithmcansupplythecrucialstepforislandimageregistrationandislandrecognition.
出版日期
2014年05月15日(中国期刊网平台首次上网日期,不代表论文的发表时间)