The　complex　background　and　small　size　of　targets　pose　challenges　for　ship　detection　in　SAR　(Synthetic　Aperture　Radar)　images.　Currently,　deep　learning-based　methods　have　been　widely　used　for　SAR　ship　detection.　However,　the　neural　networks　involved　in　these　methods　have　a　large　number　of　parameters,　which　hinders　their　practical　applications　on　resource-constrained　platforms.　This　paper　proposes　a　lightweight　SAR　detection　algorithm　specifically　designed　for　small　ship　targets　in　SAR　images.　Firstly,　we　adopt　a　decoupling　strategy　to　reconstruct　the　backbone　network,　effectively　reducing　the　number　of　parameters.　Secondly,　a　new　regression　loss　function　MPDIoU　is　proposed　to　enhance　the　detection　accuracy　of　small　ship　bounding　boxes　and　improve　the　regression　speed　of　the　model.　Finally,　we　conduct　experiments　on　the　SSDD　and　HRSID　datasets,　and　the　results　show　that　our　proposed　method　achieves　improvement　in　mAP　(mean　average　precision)　by　97.8%　and　89.2%　respectively,　while　the　model　weight　parameters　is　only　72.5%　of　the　standard　YOLOv8n.　©　The　Institution　of　Engineering　&　Technology　2023.