Key points are not available for this paper at this time.
The increasing reliance on LiDAR-Camera systems to enhance robustness and accuracy in robotic systems presents new challenges, particularly in extrinsic calibration. Traditional techniques typically involve offline calibration using targets such as checkerboards and assume that the extrinsic parameters between sensors remain constant during subsequent usage. However, this assumption fails to account for the drift in extrinsic parameters caused by external perturbations, such as vibration and deformation, over long-term operation. In this paper, we introduce CalibOnline, a novel method for the online detection and correction of miscalibration in multi-sensor setups. Firstly, we propose a unified representation of lidar and camera data in the form of depth maps to reduce calibration uncertainties stemming from data modality discrepancies. Subsequently, we explore a robust characteristic of this modality, the edge of depth discontinuities, to facilitate efficient matching between depth maps. The impact of extrinsic variations on edge-matching constraints is then analyzed, and a miscalibration detection module is accordingly designed to monitor extrinsic parameters. Finally, the correction of extrinsic parameters is formulated as a problem of on-manifold optimization, which enhances the convergence of the estimated extrinsic parameters. Experimental results across various datasets and scenarios demonstrate the high robustness and accuracy of the proposed method. Code is open-sourced at https://github.com/cchester25/CalibOnline.git.
Feng et al. (Fri,) studied this question.