Accuracy Validation of UAV-based Photogrammetry on Small Rivers using Two-Course Flight Design

Abstract

We conducted a UAV photogrammetry experiment using the "along-river two-course flight" method on a small river. The results showed that even when the RTK-GNSS positioning solutions for the camera locations were utilized in the bundle adjustment, vertical errors exceeding 0.1 m occurred at ground verification points, and the camera parameter estimation was insufficient. This revealed that the weakness of the along-river two-course flight method is not only the previously pointed-out issue of being prone to georeferencing rotation around the course axis.

A potential cause for the insufficient accuracy in camera parameter estimation is the instability of the estimation process, which stems from the possible indeterminacy of internal parameters when the flight courses are straight. Therefore, as countermeasures, we investigated a method to avoid the effects of indeterminacy by adding a small number of supplementary images, and a method to improve the stability of camera parameter estimation by adding non-vegetation tie points. We demonstrated that these methods have the potential to significantly reduce verification point errors.