第 44卷 第 1期 国 防 科 技 大 学 学 报 Vol.4noiseinterferenceinremotesensingmultimodalimagematching,
aremotesensingmultimodalimagematchingalgorithmbasedonPCA(principalcomponentsanalysis)andHOG(histogramoforientedgradients)
wasproposed.ThisalgorithmusesHOGtoextractthecommonfeaturesofgeometricstructurebetweenimages,whichcaneffectivelyovercomethe
problemofnonlineargrayscaledistortionofmultimodalimages.Besides,afastmultiscalePCA algorithm wasproposedtoenhancethelocal
gradientdirectioninHOG,sothatitcanaccuratelyextractthestructuralfeaturesoftheimageundertheconditionofstrongnoiseinterference.In
ordertoimprovethecalculationspeedofthealgorithm,theintegratedimagemethodwasusedtoreducethecomputationalcomplexityofthefeature
extractionprocess,andthefastFouriertransformwasusedtoachieveahighlyefficientmatchingsearch.Theexperimentusedavarietyofremote
sensingmultimodalimages(includingvisiblelightimages,syntheticapertureradarimages,andinfraredimages)toverifythematchingalgorithm.
Theresultsshowthat,comparedwithexistingalgorithms,theproposedalgorithmsignificantlyimprovesthematchingperformance.
Keywords:remotesensingimages;multimodalimagematching;principalcomponentanalysis;histogram oforientedgradients;structure
featuredescription
由于成像的物理机制不同,异源图像可反映出
多尺度PCA-HOG遥感异源图像匹配算法 来自淘豆网m.daumloan.com转载请标明出处.