Airborne laser scanning data quality enhancement via rigorous calibration and comprehensive strip adjustment techniques

Airborne laser scanning (ALS) serves as an effective approach for rapid geo-spatial data acquisition. However, multi-sensor integration and unpredictable conditions during data collection introduce systematic errors and random errors, causing inaccurate georeferencing and point cloud discrepancies. To enhance the ALS data quality, we propose rigorous system calibration and robust strip adjustment methods. Aiming at ALS systematic errors, we analyze the structure of an airborne system in detail, construct its rigorous observation model, employ robust nonlinear least squares for solving several system calibration parameters together, and develop an interactive user interface (UI) for incorporating ground truths into the calibration workflow. Aiming at ALS random errors, we utilize temporal information for spatial segmentation of overlapping point clouds in the surveyed area, employ a robust iterative closest point (ICP) registration specialized for airborne point clouds, and design an optimization strategy for both point cloud and trajectory corrections. Detailed experiments are conducted for the system calibration and strip adjustment methods. In quality, the calibration enhances the initial fusion accuracy, and the strip adjustment further improves the discrepancy of multi-strip point clouds. In quantity, we correct the offset from several meters to centimeters after calibration, and point cloud matching achieves around 2-time to 3-time improvement in root mean square error (RMSE). Besides, notable advantages in efficiency and convenience are achieved with paralleling strategy and automated workflow. In summary, our contributions include accuracy improvement, the facilitation of user convenience through an automated workflow, while ensuring processing efficiency via a paralleling strategy. Providing insights into ALS error sourcing and mitigation as well as practical guidance for multi-sensor system data processing, our methods show generalization capability that enables the promotion into broadening applications of geo-spatial information science.