Detection of polymeric insulator defects based on YOLOv8

This work proposes using the You Only Look Once (YOLO) architecture for defect detection in suspension polymeric insulators from images. Firstly, an image dataset of 138 kV polymeric suspension insulators was developed. Then, the dataset was augmented by horizontal and vertical flips to increase the training data. Afterward, YOLOv8n and YOLOv8s architectures were trained and used to detect insulator defects from the captured images. The developed models can detect four insulator defects: exposed rod, end fitting corrosion, cracks, and damaged sheds. Finally, the trained models were evaluated based on precision, recall, mean Average Precision (mAP), and confusion matrix. The results showed that YOLOv8s presented satisfactory results and potential for deploying and experimentation on UAVs to monitor polymeric insulators.