Defect Detection of Transmission Line Bolts Based on Improved YOLOv5

Abstract

In addressing the challenge of striking a balance between detection speed and accuracy for detecting nail defects in the task of intelligent nail defect elimination in transmission line inspection robots, this paper analyzes existing object detection algorithms. We propose an improved small-object detection algorithm based on YOLOv5. Firstly, the rationale for selecting the YOLOv5 algorithm is explained, and an overview of the traditional YOLOv5 algorithm is provided. Subsequently, we establish a dataset of nail images captured by unmanned aerial vehicles (UAVs) and determine the optimal resolution for nail image recognition through image segmentation. We then enhance the dataset through data augmentation. Next, we innovate the network structure on the foundation of the conventional YOLOv5 algorithm. Finally, a comparative experiment is conducted with representative algorithms. The experimental results demonstrate that the improved algorithm achieves higher detection accuracy compared to traditional nail defect detection algorithms.