Abstract:

LiDAR and camera sensor fusion is important for autonomous robots to perceive the world. To fuse the measurements from these two sensors, extrinsic calibration is required beforehand. A common approach to register the camera and LiDAR coordinate systems is to align the 3D edges detected by the LiDAR with the 2D edges captured by the camera, and state-of-the-art algorithms can achieve pixel-level precision. However, one practical problem is that inaccurate camera intrinsic parameters would influence the accuracy of extrinsic calibration. Therefore, we propose to conduct camera intrinsic calibration and LiDAR-camera extrinsic calibration simultaneously in target-less environments.