Vision-based map referenced UAV navigatiobn using terrain classification

Abstract

Satellite navigation, which is adopted by most UAVs today, is reliable and efficient, but it is pointed out that the satellite service may be limited by the service provider, or the signal reception may be obstructed. Therefore, there is increasing interest and research on the auxiliary navigation system that can be used for absolute positioning even when satellite navigation is disabled. This paper propose a new navigation system that estimates the position of an aircraft by comparing the map database with the real-time aerial image obtained by an low-cost image sensor. The operation principle of the proposed navigation method is described by introducing the semantic segmentation of aerial image, map databases and aerial image matching using ICP algorithm, and implementation of integrated navigation using Kalman filters.