ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume V-1-2020
https://doi.org/10.5194/isprs-annals-V-1-2020-245-2020
https://doi.org/10.5194/isprs-annals-V-1-2020-245-2020
03 Aug 2020
 | 03 Aug 2020

AERIAL-TRIANGULATION AIDED BORESIGHT CALIBRATION FOR A LOW-COST UAV-LIDAR SYSTEM

J. Li, B. Yang, C. Chen, W. Wu, and L. Zhang

Keywords: Low-cost, UAV-LiDAR System (ULS), Boresight Calibration, Dynamic Networks

Abstract. The Laser-IMU boresight calibration is the precondition for an Unmanned Aerial Vehicle (UAV)-Light Detection and Ranging (LiDAR) system (ULS). The existing methods achieve good performance for calibrating ULSs with high-precision positioning and orientation systems (POS) (e.g., APX-15), in which, the systematic errors of the high-precision POS can be ignored, only the boresight parameters are estimated. However, these methods have difficulties in calibrating the low-cost ULSs with low-precision POS. To overcome the impact of the systematic errors of the low-precision POS on boresight calibration, an aerial-triangulation aided boresight calibration is proposed in this paper. It simultaneously estimates the laser-IMU boresight angles and system states (e.g. trajectory) by setting the point clouds derived from aerial-triangulation (AT point clouds) as the reference. Firstly, the planar voxels from the AT point clouds are extracted, due to the fact that they are more reliable in AT point clouds. Secondly, raw laser observations are matched with the extracted planar voxels to establish laser matching observations. Thirdly, a Dynamic Network (DN) is built using the GNSS observations, inertial observations, and laser matching observations to simultaneously optimize the initial laser-IMU boresight angles and the system states. All the sensor observations involved in the ULS are modeled with proper error models, which are essential for analyzing and refining the data quality of the low-cost ULS. The proposed method was tested to calibrate a low-cost ULS, KylinCloud-II, in a calibration field. It showed that the average distance between the laser point clouds and the referenced AT point clouds was decreased from 2.560m (RMSE = 3.88m) to 0.08m (RMSE = 0.99m).