Comparative Accuracy Assessment of Nadir and Hybrid RTK-UAV Photogrammetry for Surface Detail Extraction

Abstract

This study evaluates the metric reliability of nadir and hybrid RTK-UAV photogrammetric models for surface detail extraction on a two-story building without using Ground Control Points (GCPs). Two photogrammetric models were generated: a standard model based only on nadir images and a hybrid model combining nadir, oblique, and manual façade-oriented images. Both models were assessed using terrestrial reference data obtained from GNSS and reflector less total station measurements. The evaluation was performed at three levels: point-based, length-based, and surface-based comparisons. The point-based results indicate that the nadir model allowed 21 of the 35 reference points to be reconstructed, whereas the hybrid model allowed the 3D positions of all. The hybrid model also substantially improved vertical and overall 3D accuracy, reducing the ΔZ MAE from 0.355 m to 0.029 m and the 3D RMSE from 0.454 m to 0.046 m. In the length-based analysis, the nadir model was included where both endpoints of a selected linear element could be clearly identified. The nadir model allowed only 4 of the 14 selected length pairs to be measured, whereas the hybrid model allowed all 14 length pairs to be evaluated. On the other hand, the hybrid model had close agreement with the reference lengths, with a MAE of 0.012 m and RMSE of 0.016 m. Also, there is no statistically significant difference observed between reference and model lengths. In the surface-based comparison, the nadir model could represent only 4 of 7 reference surfaces, a limited number of reference surfaces, whereas the hybrid model enabled all selected surfaces to be evaluated, including upper-floor and under-eave regions. Although the nadir model produced lower average errors on some reconstructed surfaces, its limited surface coverage reduced its practical applicability. Overall, the findings demonstrate that hybrid RTK-UAV image acquisition provides a more robust and reliable solution than nadir-only acquisition for building surface documentation, particularly in façade areas with restricted visibility.