Monocular Visual-Inertial SLAM with Camera-IMU Extrinsic Automatic Calibration and Online Estimation

An approach of automatic calibration and online estimation for camera-IMU extrinsic parameters in monocular visual-inertial SLAM (Simultaneous Localization and Mapping) is proposed in this paper. Firstly, the camera-IMU extrinsic rotation is estimated with the hand-eye calibration as well as the gyroscope bias. Secondly, the scale factor, gravity and camera-IMU extrinsic translation are approximated without considering the accelerometer bias. All these parameters are refined with the gravitational magnitude and accelerometer bias taken into account at last. Furthermore, the camera-IMU extrinsic parameters are put into state vectors for online estimation. Experiment result with the EuRoC dataset shows that the algorithm automatically calibrates and estimates the camera-IMU extrinsic parameter with the extrinsic orientation and translation's error within 0.5 degrees and 0.02 m separately, which contributes to the rapid use and accuracy of the VI-SLAM system.