A vision-based robust hovering control system for UAV

Abstract

This paper introduces an algorithm for real-time line detection and tracking utilizing the Graphic Processing Units (GPUs) for UAV's vision-based hovering control system. We concentrate that there are many of lines where UAV can fly, and extract meaningful line to grasp of vehicle's attitude. We implemented image processing techniques on GPUs for real-time performance because detection and tracking of lines need huge computational resources. Experiments show affordable frame throughput that our approach is feasible in real flight. © Springer-Verlag Berlin Heidelberg 2014.