Online Inertial-Aided Monocular Camera Self-Calibration

In this paper, we present a novel manner to perform an online camera calibration, without any prior knowledge of the environment, based in the readings of an IMU and in the track of a single 3D point projected into the camera image frames while the camera moves around. Our method is intended to be used alongside with a visual odometry or visual-inertial navigation system, since this type of algorithm strongly depends on the precision of the intrinsic camera parameters but they normally rely in offline calibration methods and take their results as constants. But it is known that there are several factors that can change the intrinsic camera parameters over time. Our approach can be used to correct these values as they are changed since it is an online method. An additional contribution of the method proposed in this work, is that it estimates, alongside with the camera intrinsic parameters, the motion performed by the camera while the algorithm is running. Our approach utilizes an EKF to perform its estimations. We demonstrate the efficiency of the method in a real world experiment, where we could perform the camera calibration and estimate the camera motion with our approach.