Occlusion　recognition　plays　a　vital　role　in　smart　manufacturing　and　industrial　automation.　This　paper　proposes　an　enhanced　YOLOv8-based　model　designed　to　accurately　recognize　and　locate　occluded　objects　in　industrial　environments,　assisting　robots　in　precise　grasping　tasks.　The　model　integrates　the　AKConv　module　and　the　Effciou　loss　function,　which　improve　feature　extraction　and　bounding　box　regression,　especially　in　complex　and　highly　occluded　scenarios.　Experimental　results　confirm　that　the　modified　YOLOv8s_Effciou　model　achieves　a　balanced　performance　across　different　levels　of　occlusion,　showing　notable　robustness　in　challenging　environments　with　high　occlusion.　　©　2024　IEEE.