Overview of the Research on Improved YOLO Model in Insulator Defect Detection

Abstract

People’s everyday lives depend on electricity, and insulators are essential to the reliable and safe functioning of high-voltage transmission lines. Therefore, defect detection of insulators has always been a concern for many research teams and researchers. The You Only Look Once(YOLO) series algorithm has always been the most widely used algorithm in the research of insulator defect detection. This paper seeks to examine the future directions for enhancing the YOLO model, evaluate the research problems in the field of insulator defect detection, and methodically study the pertinent theories for doing so. First, the paper outlines the main research progress in the field of insulator defect detection. Then focus on discussing the basic structure, dataset, improvement research, and experimental results of the three generations of YOLOv3, YOLOv5, and YOLOv8 models. Based on the above review, the paper proposes directions for enhancing the dataset of insulator defects and provides insights on improving the model for insulator defect detection. In addition, this article also points out the difficulties in collecting datasets in current research, providing research suggestions for future researchers.