Transmission Line Insulator Defect Detection Based on Enhanced YOLOv5n

To meet the inspection requirements for insulators on power transmission lines, this article proposes an improved algorithm based on YOLOv5n. First, the FasterNet Block is employed to replace the Bottleneck in the C3 module of YOLOv5n, reducing redundant computations and memory accesses. Then, the Dyhead (Dynamic Head) module is introduced to integrate scale-aware, spatial-aware, and task-aware attention mechanisms simultaneously, enhancing the detection capability of insulator defects in complex environments. Finally, the Normalized Wasserstein Distance (NWD) is introduced into the loss function calculation to mitigate the sensitivity of IoU-based metrics to small target position deviations, thus improving the detection capability for small-sized insulator defects. Experimental results show that compared to the original YOLOv5n, the insulator defect Average Precision (AP) improved by 1.9%, mAP improved by 1.8%, and parameters and FLOPs were reduced by 8.5% and 4.9%, respectively. Compared to YOLOv7-tiny, the insulator defect AP increased by 1.7%, while parameters and FLOPs were only 26.7% and 29.5% of YOLOv7-tiny. The proposed algorithm allows for more accurate detection of insulator defects with lower model complexity.