Fine Registration of Low-Resolution Infrared Camera and Low-Beam LiDAR

Extrinsic calibration forms the foundation for data fusion with multi-sensors. However, challenges arise for low resolution infrared cameras and low-beams LiDAR extrinsic calibration, which are resulted by technology and cost considerations. Traditional calibration methods using chessboard patterns are no longer suitable for these two kinds of sensors. Therefore, this paper proposes a novel extrinsic calibration method based on special feature points from infrared images and LiDAR point clouds. The framework involves two main steps: an initial extrinsic guess using Perspective-n-Points and refinement of extrinsic parameters through similarity optimization. This method impressively shortens the calibration time compared to manual calibration and improves accuracy and robustness. In order to verify the efficacy of this approach, practical experiments were carried out using authentic datasets acquired from a physical unmanned vehicle.