A hierarchical approach for refining point cloud quality of a low cost UAV LiDAR system in the urban environment

Insufficient accuracies of the low end position and orientation system (POS) used in low cost UAV LiDAR systems (ULSs) cause the direct georeferencing data to lead to poor point cloud quality. Trajectory correction and scan-to-map matching are two commonly used strategies for point cloud quality refinement. The existing trajectory correction strategies work with the assumption that POS errors can be modeled as a time-variant function, which cannot be applied to the low end POS. The existing scan-to-map matching methods have difficulty refining the ULS point clouds due to the large gaps between scan lines. This paper proposes HR-ULS, hierarchical refinement for low cost ULS point cloud quality in the urban environment, to solve these challenges. HR-ULS separated the raw laser scanning point clouds into a set of scan-blocks and refined the point cloud quality with a hierarchical strategy, resulting in local and global optimization, respectively. First, the internal scan-block matching (ISBM) estimated multiscale distributions for each laser frame and calculated relative motions iteratively to achieve local map consistency in each scan-block. Second, the multiview scan-block matching (MSBM) took inertial, Global Navigation Satellite System (GNSS), and laser measurements into a unified adjustment framework to correct the trajectory, achieving global map consistency between scan-blocks. Comprehensive experiments evaluated the proposed HR-ULS with the point clouds captured by a low-cost ULS in three typical urban areas. They showed that the average plane fitting RMSE of the ULS point clouds was improved from 0.34 m to 0.09 m, and the average checkpoint offset was improved from 1.86 m to 0.21 m, achieving an identical level of accuracy with that of direct georeferencing using a high end POS, APX-15-UAV.