利用高斯混合模型的SAR图像目标CFAR检测新方法.pdfVIP

14 1 V o l. 14, N o. 1 2009 1 Journa l o f Im age and G raphics Jan. , 2009 SAR CFAR 张 军 田 昊 黄英君 ( , 4 10073 SAR ( CFAR ( SAR , CFAR , CFAR SAR , , CFAR , , CFAR , CFAR , , SAR : TP391 : A : 1006-8961( 2009 01- 0019-06 A NovelCFAR Algorithm forDetecting Targets in SAR ImagesUsingGaussianM ixtureModel ZHANG Jun, T IAN H ao, HUANG Y ing-Jun (S ch ool of Inf orma tion Sy stem and M anagem en ,t N a tional University of D ef en se Techn ology, Chang sha 4 10073) Abstract C lutters in SAR ( synthetic aperture radar mi age is variou s and com plex1 So d ifferent k inds o f d istribution m ode ls shou ld be u sed for CFAR ( constant fa lse alarm rate based targ et detection of a irbo rne SAR mi ag es, w h ich increases the d iff icu lties and comp lex ities o f autom atic targe t detection1 G aussian m ix ture m ode l ( GMM is an ex tension to G au ssian probab ility density func tion1 T heo re tica lly, it can approx mi a te any distr ibution smoo th ly1 In th is paper, c lutter d istr ibu tions o f SAR mi ag e are estmi ated u sing the GMM 1 A nd a novel CFAR thresho ld expression w as propo sed1 D escrib ing different c lutter d istributions w ith one un iform model, the proposed CFAR detec tor is m ore un iversa l since it is less dependent on c lutte r distributions1 T he expe rmi ental results show that, ev en though the c lutter d istributions are unknow n, the propo sed technique still has a n ice perform ance1 eywords GMM ( G au s