카메라-IMU 캘리브레이션 기반 카메라 내부 파라미터의 제약조건 최적화 및 검증

Abstract

In aerial vision-based navigation systems, accurately estimating the misalignment between the camera and the inertial measurement unit (IMU) is critical. However, uncertainty in the camera’s intrinsic parameters can propagate to misalignment estimation and result in geometric errors. This can lead to significant degradation of vision-based navigation performance during high-altitude operations. To address this, we propose a reversal calibration method that leverages extrinsic parameters information to determine the effective range of intrinsic parameters. Based on the results of extrinsic calibration, we define an effective correction region for intrinsic parameters with respect to the physical centers of both sensors. This relationship is then integrated into a constrained optimization framework to refine the intrinsic parameters within physically valid bounds. Experiments were conducted using a camera with a focal length of 35 mm. The results demonstrate that incorporating physical constraints into the optimization process leads to more stable parameter correction, and reduced reprojection errors by up to 60.73%.