Kolmogorov-Arnold network and efficient multi-scale attention graph neural network based multi-object detection for power transmission lines

The power system is a core infrastructure for national economy and energy security, and the health status of transmission lines directly affects its reliability. When traditional image processing methods are used for transmission line detection, their generalization ability is insufficient in complex scenarios. Although existing deep learning models have made progress, they still face challenges such as multi-scale differences in transmission line images and interference from complex backgrounds. To this end, this paper proposes a Kolmogorov-Arnold Network and Efficient Multi-Scale Attention Graph Neural Network (KE-GNN) model based on graph convolutional networks, which converts images into graph structures and processes them via graph convolutional networks. Additionally, it enhances the multi-scale feature transformation capability by combining Kolmogorov-Arnold Network (KAN) and improves the detection robustness in complex environments by using Efficient Multi-Scale Attention (EMA). Experiments show that KE-GNN has practicality in the task of transmission line target detection, providing a new solution for the intelligence of power inspection.