An Iterative-Optimization-Based Calibration Framework for VIO With Limited Prior Conditions

Most of the visual inertial positioning accuracy is constrained by the calibration accuracy of the sensor, however, there are few navigation systems that take the time offset and extrinsic parameters into account at the same time. In this paper, an iterative-optimization-based calibration framework for VIO (visual-inertial odometry) with limited prior conditions is proposed. In particular, a three-stage calibration method is presented, which includes the time offset estimation stage, the extrinsic and intrinsic parameters calibration stage, and global temporal and IMU parameter optimization stage. In the first stage, the time offset calibration model is proposed, which estimates the time delay by iteratively optimizing the orientation residual equation. In the second stage, based on the assumption of constant angle velocity and linear velocity between two continuous keyframes, an iterative-based coarse-to-fine calibration strategy is presented to calibrate the initial extrinsic and IMU intrinsic parameters. In the third stage, a global optimization-based algorithm is introduced to obtain a more accurate time offset value and IMU intrinsic parameters. What’s more worth mentioning is that the proposed method does not require any prior parameters, which greatly improves the universality of the navigation system. To evaluate the effectiveness and accuracy of the proposed algorithm, the method is tested on both simulation and public dataset, and also compared with classical systems. Abundant experimental results illustrate that the proposed method can effectively estimate the time offset, the IMU intrinsic and extrinsic parameters.