With the rapid development of unmanned supermarkets, commodity object detection technology has become one of the key techniques for achieving efficient and accurate settlement. This paper presents an enhanced commodity object detection method based on an optimized YOLOv8 network model, which improves detection performance through optimizing the model structure and parameter combinations. Through data augmentation and model optimization methods, the model's generalization ability in complex scenes is improved. Specifically, the Conv layer is replaced with GhostConv, and the C2f module is replaced with C2f_RepGhost, significantly reducing the model's parameter volume and achieving a 9.5% improvement over the original YOLOv8. Ablation studies and data augmentation validate the robustness and effectiveness of the proposed method. Field applications in the unmanned supermarket environment demonstrate that the proposed method excels in real-time performance, accuracy, and robustness, effectively meeting the practical application needs of unmanned supermarkets.