Insulator defect detection in UAV inspection images of transmission lines is hampered by challenges, including complex background clutter and significant variations in object scale. This paper proposes a novel YOLOv12-based method for detecting insulator defects. To effectively enhance the model’s ability to capture irregular breakage edges of insulators, the C3k2-WTConv module is designed, which expands the model’s receptive field through multi-frequency feature fusion. Furthermore, to address missed and false detections of small targets and improve feature extraction performance in complex backgrounds, an attention module named SEAM is introduced into the detection head. Extensive experiments on a self-constructed insulator defect dataset verify the effectiveness of the proposed approach, showing consistent improvements over the baseline in detection precision and robustness. The findings provide valuable insights for advancing intelligent UAV-assisted inspection of power transmission infrastructure.
 

Insulator defect detection; Improved YOLOv12; Complex background; Data augmentation; Image Recognition