Geometric accuracy evaluation of geospatial data using low-cost sensors on small UAVs

Abstract

The recent development and proliferation of Unmanned Aircraft Systems (UASs) has made it possible to examine environmental processes and changes occurring at spatial and temporal scales that would be difficult or impossible to detect using conventional remote sensing platforms. However, new methodologies need to be codified in order to be compared with traditional photogrammetric products. This can be done by testing geometrical accuracies reached by the models when external orientations have changed. In this paper two dense point clouds, derived from the same spatial database, were compared to evaluate the discrepancies resulting from two different relative orientations: the first one based on the GPS position of each UAV frame and the second one based on GCPs, measured through GNSS positioning. The two dense point clouds presented an average offset of 3.4 cm and a standard deviation of 5cm, proving that relative accuracy is only influenced by the matching intensity. To assign three different absolute orientations, the georeferencing procedure of the same orthomosaic was then verified based on GCPs coming from three different open geo-data sources. By evaluating the position discrepancies of some Independent Check Points (ICPs), the three open geo-data sources provided three estimates of different root-mean-square error (RMSE) positional accuracy, of which the absolute geometrical precision was the related function.