Transmission Line Nut-Pin Defect Detection Based on Improved YOLOv8

Despite the advancements in deep learning techniques, the detection of transmission line nut-pin defects remains a significant challenge. This paper introduces a YOLOv8-based algorithm, YOLOv8-DND (You Only Look Once v8-Dowel Nut Detection), to address the issue of insufficient feature information due to the small target size of nut-pin defects. By incorporating the ECBAM attention mechanism, the CVT cross- layer feature map interaction module, and the DND detection head into YOLOv8, the algorithm achieves precise detection of transmission line nut-pin defects. The experimental results demonstrate that YOLOv8-DND is effective in detecting nut-pin defects among the YOLO series algorithms.

I. Introduction

In the maintenance and management of transmission lines, the detection of nut-pin defects is crucial to ensuring the safety of operations. The application of deep learning algorithms for automated, high-precision detection of these defects has emerged as a research focus within the power industry. This paper presents a novel method for detecting transmission line nut-pin defects, leveraging the enhanced YOLOv8 algorithm to increase accuracy and bolster the power system's operational safety.