Inertial-aided state and slope estimation using a monocular camera

This paper aims at estimating the metric state of a quadrotor using a monocular camera and an Inertial Measurement Unit (IMU) based on a homography model. Angular velocities are obtained from the on-board IMU and subtracted from the homography matrix. The reshaped vector from the homography matrix is directly fused with acceleration measurements using an Extended Kalman Filter. The visual inertial fusion framework is capable of estimating distance to the plane, speed, acceleration biases and surface normal. The feasibility of our method is proven using simulation and on real sensory data recorded from our quadrotor platform. It is shown that our approach is superior over traditional method that decomposes the homography matrix for the estimation of surface normal.