AIMS-YOLOv11n: Small target defect detection model for power line insulators

Abstract

With the rapid development of smart grid construction, safe and stable operation of transmission lines has become critical to ensuring power supply reliability. Addressing issues such as missed detections, false positives, and poor real-time performance in transmission line insulator defect detection, this paper proposes AIMS-YOLOv11n—a small-object defect detection model for transmission line insulators based on an improved YOLOv11n model. First, the backbone network incorporates average pooling downsampling (ADown) to enhance detection performance while reducing model parameters and computational complexity. Second, the C3k2_iRMB_Cascaded module is restructured to improve spatial information capture across scales, significantly boosting detection capabilities for objects of varying sizes. Third, a parallel MLCAttention mechanism is integrated post-feature fusion to enhance small object representation. Finally, the Focaler-SIoU loss function is employed to address the limitations of CIoU loss in handling small-scale objects. Experimental results on a public insulator detection dataset demonstrate that compared to the YOLOv11n model, the proposed AIMS-YOLOv11n achieves significant improvements in both detection performance and lightweight optimization: its mAP50% reaches 91.4%, a 3.9% increase over the original model, while mAP50%–95% improves by 4.5%, effectively addressing the accuracy limitations in insulator defect detection. Simultaneously, the model achieves lightweight optimization with a 24.2% reduction in parameters and a 17.5% decrease in computational complexity. These results demonstrate that the AIMS-YOLOv11n model effectively meets the accuracy requirements for transmission line insulator inspections while satisfying lightweight deployment demands at the edge.