High-Resolution Map Generation Algorithm for UAV Large-Scale Scene Matching Navigation

Abstract

Aiming at the guarantee issue of the reference map for the visual scene matching navigation system, this paper proposes a scene matching reference map generation and optimization method based on Inertial/Vision/GNSS fusion, which enables real-time generation and optimization of large-scale reference maps. During the real-time navigation of an unmanned aerial vehicle (UAV), the algorithm first uses the loose integration of Inertial/GNSS navigation results to complete the geographic information correction of the real-time map. Further, a global reference coordinate system mapping algorithm based on geographic information is designed. Then, the XFeat algorithm is applied to solve and splice the matching relationships of keyframes. Finally, the geographic coordinate annotation is completed. Flight tests demonstrate the effectiveness of the proposed algorithm, which can realize the real-time generation of large-scale reference maps. Meanwhile, the comparison test at ground integrated navigation observation points verifies that the geographic information error of this algorithm is better than 0.05 meters. Experimental results demonstrate that the proposed method in this paper exhibits advantages in map construction efficiency, positioning precision, and coverage scale.