New Methodology for Precise UAV Surveys with a Single Ground Control Point

Abstract

The advances in aerial mapping using unmanned aerial vehicles (UAVs) are increasingly allowing for the surveying of areas that are difficult or unsafe to access. Considering that the cameras boarded on UAVs are non-photogrammetric sensors, positional accuracy and precision are paramount to ensuring that the final product is geometrically consistent and that its coordinates are precise enough within an acceptable range (which depends on the purpose of each application). With this in mind, this work evaluates the positional accuracy of aerial surveys with UAVs supported by the post- processed kinematic (PPK) technique for the generation of cartographic products in compliance with the Brazilian cartographic accuracy standard PEC-PCD. Furthermore, this study proposes a methodology for inserting virtual control points using coordinates adjusted and inserted into the main point of selected images, using the package of solutions for GNSS RTKLIB processing, to forgo Ground Control Points (GCPs) without loss of quality of the cartographic product. This research used a UAV from DJI (model Mavic 2 Pro), equipped with a dual frequency GNSS receiver, where positional records were synchronized with the camera’s shooting times. To validate the technique, 20 checkpoints with known coordinates were selected, and these points were used in statistical tests to assess the accuracy and precision of the survey. The orthomosaic and the digital elevation model were used as the reference for positional coordinates to be compared against the PEC-PCD standards. The results show that the discrepancy is, on average, 0.126 m for planimetric coordinates and 0.066 m for altimetry. No outliers were found, suggesting a Gaussian sampling distribution. The application of Student’s t- test indicated the existence of a small bias, suggesting the displacement of coordinates along the three axes. The chi-square test showed results below the tabulated limits, attesting to the high precision of the survey. Finally, the root mean squared error (RMSE) was lower than the standard error limit for PEC-PCD class A planialtimetric cartographic products on a 1:1,000 scale. The viability of this technique is thus confirmed.