Low-cost Extrinsic Calibration Between LIDAR and Thermal Camera for Indoor Mapping

Abstract. Building on the concept of LIDAR–RGB extrinsic calibration using simple box targets (Pusztai and Hajder, 2017), we propose a low-cost and effective method for extrinsic calibration between a thermal camera and a LIDAR. Our approach utilizes a single, inverted half-box target featuring a heated asymmetric circles grid on one of its planes, designed for high visibility in thermal imagery. Thermal camera poses are estimated using OpenCV’s calibration toolbox, while LIDAR poses are computed by extracting the three planar surfaces of the calibration target. The extrinsic transformation between the thermal camera and the LIDAR is then determined using a cycle consistency constraint. We evaluate the method in terms of calibration accuracy and precision. Additionally, the setup can be used to assess the time synchronization between the thermal and LIDAR data streams by visualizing the projection errors on dynamic scene elements. Additional results can be found in the video at youtu.be/0ODHZe4rVec. The source code is available at github.com/SaxionMechatronics/CycleCameraLiDARCalibration.