<p>To address the challenges of misdetection and missed detection caused by the diversity of insulator types and the complexity of defect textures in power transmission lines, and to meet the demands of collaborative inspection, we propose a real-time detection algorithm for insulator defects based on YOLOv8. First, considering the characteristics of the dataset sample sizes, we designed a lightweight OMAL-Neck structure that optimizes feature fusion, enhancing the utilization of feature information and improving detection performance for medium and large targets. Second, to address the issue of large parameter and computation requirements in the YOLOv8 detection head, we designed a lightweight and efficient detection head. This redesigned detection head incorporates PConv, further accelerating model inference speed. Lastly, to counteract the decline in detection accuracy due to model lightweighting, we integrated the C2f module with DySnakeConv, enhancing the feature extraction capability for tubular structures and complex textures, thereby preventing information loss. Experimental results demonstrate that compared to the baseline YOLOv8s, the proposed model increases FPS from 44 to 78 frames/s, reduces the number of parameters and computational complexity by 27 and 38%, respectively, and improves the mAP by 1.7%. The improved model offers significant advantages in both detection accuracy and real-time performance, enabling rapid and precise identification of insulators and their defects, thereby improving the efficiency of power line inspections.</p>