Insulator　detection　in　aerial　images　plays　a　crucial　role　in　the　inspection　of　transmission　lines,　as　it　is　vital　for　maintaining　the　safety　and　sustainability　of　the　power　system.　This　paper　presents　an　enhanced　algorithm　for　detecting　insulators　in　transmission　lines　using　the　RetinaNet　model.　Our　approach　first　utilizes　ResNet101　instead　of　ResNet50　to　extract　more　comprehensive　feature　information.　Additionally,　we　replaced　the　original　L1Loss　with　the　Distance-IoU　(DIoU)　loss　function　to　improve　the　accuracy　of　bounding　boxes.　Experimental　results　demonstrate　that　our　method　achieves　a　1.2%　increase　in　mean　average　precision　(mAP)　and　a　1.7%　increase　in　average　recall　(AR)　compared　to　the　original　RetinaNet　algorithm.　These　findings　indicate　that　our　enhanced　approach　effectively　improves　the　performance　of　insulator　detection.　©　2024　IEEE.