SAR圖像船舶目標(biāo)Transformer輕量化檢測(cè)方法分析與設(shè)計(jì)

打開文本圖片集

中圖分類號(hào)：TP391 文獻(xiàn)標(biāo)志碼：A 文章編號(hào)：2095-2945（2025）24-0007-07

Abstract：ShiptargetdetectioninSARimagesisofgreatsignificanceinmaritimesearchandrescue，aswellasin environmentalmonitoring.Inrecentyears，Transformer-baseddetectionnetworkshavedevelopedrapidlyinthefieldofSAR targetdetection.However，thiskindofSARshipdetectionnetworkstllhastwoprominentproblems：dificultyindetecting fshoreshipsandhighcalculationcomplexity.Thispaperconductsanin-depthanalysisoftheseproblems.Aimingatthe problemof highfalsealarmsandmanymissedmissesincoastal shipdetection，thenetwork structureisreconstructed，a deformableatentionmechanismisusedinthebackbonenetworkpart，andadata-relatedofsetgeneratorisintroduced；a contour-guidedshapeenhancementmoduleisusedinthenecknetworktosuppesstheinterferenceofbackground clutter generatedbycoasts，islandsorwavesontheimage.Inordertosolvetheproblemofcomputingefciencythispaperaddsa dynamicnetworksparsitymoduletothenetworkfrontendtograduallypruneredundantinformationindiferentlayers. Experimentalresultsshowthattheimproveddetectionnetworknotonlyimprovesthedetectionefectofofshoreships，butalso reduces the computing complexity，and has practical application value.

Keywords：targetdetection;syntheticapertureradar （SAR）；Transformer，contourenhancement;deformableattention

SAR船舶檢測(cè)在海上交通監(jiān)控、非法捕撈監(jiān)測(cè)等領(lǐng)域具有重要價(jià)值，尤其在惡劣天氣和復(fù)雜環(huán)境下提供了全天候的遠(yuǎn)程監(jiān)測(cè)能力。（剩余12227字）