Photogrammetry meets Computer Vision: Simplified LiDAR to Camera Calibration

Calibration of camera extrinsics and intrinsics can be done in a laboratory using a 3D test field (preferred photogrammetric method) or more automated by use of planar calibration targets (preferred computer vision method). Within this paper, we focus on a new concept for the calibration of the relative orientation of cameras vs. low-cost (but dense) LiDAR sensors using virtual LiDAR intensity images. This allows using the same workflows already developed for cameras and can deliver highly accurate results for registration of LiDAR point clouds to images. In this paper, we show results from the calibration of a combined RGB/TIR camera system complemented by a Livox Mid-360 LiDAR sensor. As a proof of concept, the calibration is performed by photogrammetric means in a specialized laboratory and then compared to the results of a computer vision workflow based on planar targets. Both approaches yield comparable results within the determined uncertainty range, simultaneously avoiding complicated measurement or computation in 3D point clouds. Overall, relatively simple planar targets combined with the generation of virtual LiDAR intensity images are adequate and sufficient for LiDAR to camera calibrations.