A universal method for extrinsic calibration of LiDAR and camera

LiDAR, an essential sensor in computer vision, is highly regarded due to its robust resistance to interference and its ability to provide accurate range measurements, even with sparse data acquisition. Conversely, cameras provide dense image data yet are vulnerable to external factors such as lighting conditions. Due to their strong compatibilities and collaborative capabilities, the current research focus has pivoted towards the combined utilization of both sensors in tasks such as robotic mapping, autonomous driving, and 3D reconstruction. Typically, a common multi-sensor setup involves the combination of LiDAR and pinhole cameras. In this study, we investigate the possible applications of wide field-of-view multi-sensor configurations. We employ the RS-Bpearl spherical LiDAR (Bpearl) along with sets of pinhole and fisheye cameras to build a multi-sensor configuration with a wide field of view. Due to the distinctive properties of the Bpearl spherical LiDAR beam, we propose a universal extrinsic calibration algorithm for LiDAR and cameras. The code is open-source, and the repository is accessible at: https://github.com/kakghiroshi/LiDAR_cam_universal_calibration.