Smart Fault Detection and Monitoring of Power Line by Drones

Abstract

In this paper, we introduce a novel automatic power line inspection system based on automatic vision. This system uses UAV inspection as the main inspection method, optical images as the main data source, and deep learning as the backbone of data analysis. To facilitate the implementation of the system, we solve three major challenges of deep learning in vision-based power line inspection: (i) lack of training data; (ii) class imbalance; (iii) detection of small parts and faults. First, we create four medium-sized datasets for training component detection and classification models. Next, we apply a series of effective data enhancement techniques to balance the unbalanced classes. Finally, we propose a multi-stage component detection and classification method based on a single-shot multi-box detector and a deep residual network to detect small components and faults. The results show that the proposed system can quickly and accurately detect common failures of power line components, including the lack of a top cover, cracks on the rod and cross arm, woodpecker damage to the rod, and rot on the cross arm. Field tests show that our system has broad prospects in the Smart monitoring and inspection of power line components and the valuable addition of smart grids.