Simulation study of homography-based vision-aided inertial navigation for aerial vehicles

This research proposes a novel scheme for vision-aided inertial navigation based on planar homography. By exploiting the multiplicative group property of homography matrices, extended Kalman filter-based vision aiding is formulated with a measurement model of the group difference between image-estimated and navigation homographies. A simple compensation for depth distortion of the homography matrix is also proposed, and shown to be effective in simulation.