Mirror Based IMU-camera and Internal Camera Calibration

In this paper, a novel method for estimating the transformation between an inertial measurement unit (IMU) and a camera together with the intrinsic parameters of the camera is proposed. The method relies on images of reflected feature points in a planar mirror captured by an uncalibrated camera mounted with an IMU. It does not rely on using a fixed calibration pattern in front of the moving camera and the motion is not limited to be planar in front of the mirror. Instead, known feature points located on the camera body are tracked over the time to estimate the IMU-camera transformation and camera intrinsic parameters. A state-space model of the system is derived and then used as input to the Sigma-Point Kalman filter framework. Simulation results show accurate estimation of both IMU-camera translation and rotation parameters as well as the camera intrinsic parameters.