A novel autonomous navigation approach for UAV power line inspection

Abstract

This paper presents a novel autonomous navigation approach for high voltage transmission line inspection. For the achievement of robustly continuous navigation along transmission line, the inspection Unmanned Aerial Vehicle (UAV) system equipped with High Definition (HD) Pan Tilt Zoom (PTZ) and high performance embedded processors is constructed. In order to cope with complex background, a Detecting-Tracking Visual Strategy (DTVS) which consists of tower detection based on Faster Region-based Convolutional Neural Network (Faster R-CNN) and tower tracking by Kernelized Correlation Filters (KCF) is developed. Further, a projective model retrieving three-dimensional course from two-dimensional images along with a corresponding inspection path are proposed, which makes inspection more secure and robust. In particular, we implement a general ROS interface to facilitate the development of similar tasks which simultaneously integrate object detection and tracking. Finally, experiments are conducted and the results verify the effectiveness of the proposed navigation approach.