A Target-based Calibration Method for LiDAR-Visual-Thermal Multi-Sensor System.
Songyu Yuan,Tian Xie,Shiqiang Zhu,Yeheng Chen,Yuehua Li,Tao Zheng,Jason Gu
DOI: https://doi.org/10.1109/ROBIO55434.2022.10011752
2022-01-01
Abstract:As the basis of multi-sensor fusion, accurate extrinsic calibration among multi-sensors is vital for hetero-geneous information fusion. However, most existing methods only focus on the calibration between two specific heteroge-neous sensors, such as camera-LiDAR, LiDAR-thermal, and etc. which may cause inevitable accumulation when applied to a multi-sensor system. To address this problem, a novel calibration target and a high-accuracy method are proposed to simultaneously calibrate the intrinsic and extrinsic of a 3D LiDAR, a thermal camera and a visible camera without any user intervention. The proposed method only relies on one calibration board, to calibrate all the intrinsic and extrinsic parameters in one-step, without any strict pose requirement or long-time optimization. Furthermore, 2D-3D corresponding features can be extracted with higher precision by considering sensor model, comparing to the common feature extraction methods. Experiments with Velodyne VLP-16 LiDAR, ZED camera and D843NT thermal camera demonstrate that the proposed method can complete calibration among three of them automatically. Finally, this paper also presents a target-based method of uniformly expressing the calibration errors among multi-sensor system, with a competitive performance against most state-of-the-art methods.