Underwater　robots　based　on　machine　vision　encounter　issues　such　as　limited　computational　resources　and　low　real-time　monitoring　accuracy　when　performing　submarine　cable　tracking.　To　address　these　challenges,　we　propose　a　lightweight　submarine　cable　object　detection　algorithm,　YOLOv8n-LSC.　In　the　backbone　network,　the　standard　convolution　(Conv)　is　replaced　by　the　receptive　field　attention　convolution　(RFAConv),　and　the　SENetV2　structure　is　introduced　into　the　C2f　module　to　build　C2f_SENetV2,　enhancing　the　backbone　network＇s　ability　to　extract　global　features.　In　the　neck　network,　the　Cross-scale　Feature　Fusion　Module　(CCFM)　structure　is　combined　with　the　VOV-GSCSP　module,　significantly　reducing　model　parameters　and　computational　complexity　while　maintaining　detection　accuracy.　Finally,　the　traditional　upsampling　module　is　replaced　with　a　Dysample　to　improve　image　processing　efficiency.　Experimental　results　show　that　the　proposed　YOLOv8n-LSC　algorithm　significantly　improves　detection　accuracy　and　reduces　the　number　of　model　parameters　compared　to　conventional　detection　algorithms　on　the　self-constructed　Submarine　Cable　Image　Dataset　(SMCD).　This　approach　effectively　balances　performance　and　lightweight　design,　making　it　well-suited　for　deployment　in　underwater　target　detection　devices　with　limited　computational　resources.　©　2025　SPIE.