Self-exploding Insulator Detection for Transmission Lines Based on Improved YOLOv5s Algorithm

For the Unmanned Aerial Vehicle (UAV) inspection of high-voltage transmission lines, the common detection methods have low detection accuracy, slow speed and unsatisfactory effect for insulator self-explosion defects. It is improved on the basis of You Only Look Once version 5s (YOLOv5s ) algorithm. Firstly, the lightweight FasterNeXt network is used in the backbone of the algorithm model to reduce its size and computation; secondly, the Convolutional Block Attention Module (CBAM) channel spatial attention module is added to enhance the ability of important feature extraction and improve network performance. The experimental results show a 1.1% increase in accuracy, an 8.9% reduction in the amount of parameters, a 7.5% reduction in floating-point operations, a 1.2M reduction in model size, and a 7Hz increase in detection speed FPS.