A linear approach to visuo-inertial fusion for homography-based filtering and estimation

A solution to visuo-inertial filtering and estimation based on homography, angular velocity, and specific acceleration measurements is proposed. This corresponds to the typical situation of a mono-camera/IMU sensor facing a (locally) planar environment. By lifting the estimation state space to a higher-dimensionnal space, we show that the problem can be formulated as a linear estimation problem. This allows for the application of classical estimation techniques, e.g., Kalman filtering. Based on this linear formulation, we also determine explicitly the motion conditions that ensure uniform observability of the system, and we propose a simple linear Luenberger-like observer. A validation of the proposed solution based on real data is presented.