A novel measurement model based on abBRIEF for global localization of a UAV over satellite images

摘要

This paper presents a method for global localization and tracking of an Unmanned Aerial Vehicle (UAV) over satellite images. We propose a new measurement model based on a novel version of BRIEF descriptor and apply it in a Monte Carlo Localization system that estimates the UAV pose in 4 degrees of freedom. The model is used to compare images obtained from the UAV downward looking camera against patches of satellite images such as the ones available on Google (TM) Earth. The proposed method was validated using real flights sequences and has yield good results with different maps of the same region spawning many years and covering large areas. (C) 2018 Elsevier B.V. All rights reserved.