Modelling of GNSS Positioning Errors in a GNSS/INS/LiDAR-integrated Georeferencing

Abstract

Kinematic laser scanning is an efficient and highly accurate method for the acquisition of 3D topographic data. A primary task in kinematic laser scanning is the transformation of the laser scanner measurements from a local scanner coordinate system to a global georeferenced coordinate system. This requires knowledge of the scanners’ trajectory (position and orientation over time). The trajectory is typically computed via Kalman filtering of Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS) measurement data. However, this trajectory often exhibits significant errors, which become apparent when point clouds acquired at different times overlap spatially. In these overlapping areas, corresponding points may be exploited in a subsequent strip adjustment to improve the trajectory and the system calibration and therefore the final point cloud. However, ignoring the raw GNSS and INS measurements and their statistical properties can lead to overoptimistic estimates and deformation of the laser point cloud. In this contribution, a method is presented which integrates the GNSS position, raw inertial measurements and laser scanner correspondences into one adjustment and explicitly models the time-correlated stochastic errors of the GNSS positioning solution. This method for direct georeferencing considerably reduces the discrepancies between overlapping point clouds, as demonstrated on an airborne laser scanning dataset in comparison with an existing state-of-the-art strip adjustment implementation.