SR-YOLO defect detection algorithm for transmission line insulator small target

In this paper, a new SR-YOLO defect detection algorithm for insulator small target defect detection is proposed in view of the characteristics of small size and easy to be ignored of transmission line insulators. Based on YOLOv10 model, the algorithm aims to reduce false detection and missed detection in the process of detecting small insulator targets. Firstly, Swin Transformer is used to replace the backbone network of YOLOv10 to enhance the recognition ability of small target features. Then, the RepGFPN module is introduced to optimize the neck network of YOLOv10, which further improves the performance of small target detection and reduces missed detection. Finally, shape-IOU is combined to improve the loss function and improve the accuracy of bounding box positioning. The optimized YOLOv10 model with SR-YOLO algorithm achieved an average accuracy of 94.0% in its experimental phase. This is 2.9 percentage points better than the previous model, which helps improve the effectiveness of detecting small target defect abnormalities in insulators.