Accuracy assessment of post-processing kinematic georeferencing based on uncrewed aerial vehicle-based structures from motion multi-view stereo photogrammetry

Abstract

We verified a method of acquiring digital surface models of steep, forested, small watershed topography, where real-time kinematic processing is difficult because of the presence of interfering objects. Our approach involved using an uncrewed aerial vehicle with a built-in global navigation satellite system, which reduces time and labour costs. We tested the applicability of structure-from-motion multi-view stereo processing, and post-processing motion corrections of positional coordinate data were also tested. Nine verification points were established in a small 0.5 km² watershed, with check dams established in the headwaters of the forested area. The position accuracy and overall working time of verification points extracted by the method were compared with the position accuracy and work time obtained by a field survey using a conventional total station. The results show that the vertical error between the total station and each verification point at an altitude of 150 m ranged from 0.006 to 0.181 m. The working time of the survey was 13% of that of the total station survey. The proposed workflow shows that safe and non-destructive topographic surveying, including fluvial geomorphological mapping, is possible with a vertical error of ±0.103 m in small watersheds. This method will be useful for rapid topographic surveying in inaccessible areas during disasters, namely, debris flow monitoring at check dam sites and efficient topographic mapping of steep valleys in forested areas where positioning ground control points is laborious. It should also be of widespread interest to geographers working with spatial challenges related to land management.