UAV back-track in GPS unreliable environment

Abstract

Unmanned aerial vehicle (UAV) is thriving in recent years. Sensors equipped on a UAV enhance its versatility of aiding human’s work instead of just flying. Function like pathing following and back-tracking is in demand for monitoring and transportation. In this thesis, we propose a pipeline of a monocular back-tracking system based on computer vision, which deftly combines classic pose estimation algorithms with optimization methods to locate UAV itself and provide a navigation target with a 6-dimension pose. This system is proved to have stable performance in a static scenario.