O³ LiDAR–Camera Calibration: One-Shot, One-Target and Overcoming LiDAR Limitations

This paper proposes a novel method for extrinsic calibration between LiDAR and camera sensors given a single observation of one calibration target board. Particularly, in order to improve the accuracy and robustness we focus on the characteristics of the LiDAR sensor data which are often ignored in the previous work. First, LiDAR points are the intersections between rays and the object surface. Hence, noise in the measurement of the time of flight produces errors in the distance along the ray. Therefore, we need to project LiDAR points to the target plane along the ray directions, not along the surface normal direction. Second, LiDAR scans are not continuous but discrete for each bearing angle. That is why sometimes it is called a range image. Thus, edge points extracted based on the range discontinuity along each channel are in fact not on the real edges but within the surface boundaries. Experimental results with simulated and real data show that we overcome the limitations of the previous work and obtain a better accuracy given a single target observation and higher robustness with noisy LiDAR data, thanks to incorporating the LiDAR characteristics in the LiDAR-camera calibration.