Improved AR-Based GLRT Detector for Range-Spread Targets in Compound-Gaussian Clutter without Secondary Data.pdfVIP

18O JOURNALOFELECTRONICSCIENCEAND TECHNOLOG VOL．9，NO．2，JUNE2010 ImprovedAR—BasedGLRTDetectorfor Range-·SpreadTargetsinCompound--Gaussian ClutterwithoutSecondaryData Xiao—FeiShuai，Ling—JiangKong，andJian—YuYang Abstrac —Theproblem ofadaptiveradardetection heavy performance loss simply owing to the lack of incompound．Gaussianclutterwithoutsecondarydatais sufficientsecondarydata． consideredinthispaper．Inmostpracticalapplications， On theotherhand，theauthorsofworks[4]，[8】， thenumberoftrainingdataislimited．Toovercomethe [10]-[14】modeledhteclutterasanautoregressive(AR) lackoftrainingdata，anautoregressive(AR)-process- Gaussian processand derivedadaptivedetectorswihtout based covariancematrixestimatorisproposed．Then， secondary datavectors．In particular,ofrtheproblem of wi th the estimated covariance matrix the one．step detecting point．1ike targets in hte Gaussian distributed generalized likelihood ratio test(GLRT、detectoris interference． Kay has exploited AR model in his designed without training data． Finally, detection generalizedlikelihoodratiotestrGLRT detectorunderhte performanceofourproposeddetectorisassessed． assumptionofknowingtarget ，andin[10]anAR—based ／ndex 删 —Autoregressive mode1． compound generalized likelihood ratio (GLR) (ARGLR．based) detectorwasderivedofrhtecaseofaatrgetsignalkn own Gaussian， distributed targets，generalized likelihood up to a scaling factor．Moreover,ofr hte problem of ratiotest． detectingrange．spreadtargets，theARGLR-baseddetectors