Evaluation of the influence of the number of GCPS on the measurement quality of a photogrammetric block captured with an RTK UAV in geographical environments with high topographic roughness

This study investigates the impact of using Ground Control Points (GCPs) on the measurement quality of photogrammetric blocks in high topographic roughness areas, captured with an RTK-enabled UAV. The research evaluates eight scenarios with varying numbers of GCPs to determine their effect on horizontal and vertical Root Mean Square Error (RMSE), Mean Error (ME), and Standard Deviation Error (SDE). The study area was characterized by high topographic roughness, with a UAV equipped with RTK GNSS used to capture images, and 117 control points (CP) established for accuracy assessment. The UAV was flown in a double grid pattern with nadir and oblique images, ensuring 85% overlap. The GCPs were distributed in different configurations across the area, ranging from zero to six GCPs per scenario. When no GCPs were used, the horizontal ME was 0.0693 m (X-axis) and 0.0686 m (Y-axis), while the vertical ME was significantly higher at -0.1781 m. The SDE followed a similar trend, with the vertical SDE being the highest. The RMSE values were 0.0707 m (X-axis), 0.0701 m (Y-axis), and 0.1900 m (Z-axis), indicating substantial bias and variability, particularly in the vertical axis. The inclusion of two GCPs had minimal impact on ME, but starting from three GCPs, ME decreased significantly, especially in the Z-axis. The SDE remained consistent across different GCP scenarios, and the RMSE showed a marked reduction with the addition of three GCPs, with no significant improvement observed beyond this number. The optimal number of GCPs was identified as four to five, providing the best accuracy results without an unnecessary increase in GCPs. The study confirmed that using RTK GNSS onboard UAVs enhances elevation accuracy, especially when coupled with a minimum number of GCPs. Significant errors were noted in areas with abrupt elevation changes, and the central placement of a single GCP did not significantly reduce RMSE. Linear regression analysis demonstrated a strong positive correlation between UAV RTK-derived elevations and terrestrial GNSS RTK measurements, with a high coefficient of determination (R² > 0.99). The spatial configuration of GCPs was crucial, with edge placements proving more effective than central ones. Histograms of vertical differences between control points and UAV RTK measurements indicated a systematic bias reduced with the addition of GCPs. The research also examined the relationship between control point error and distance to the nearest GCP, finding no significant correlation. This study underscores the importance of GCPs in improving the measurement quality of photogrammetric blocks in areas with high topographic roughness. Without GCPs, UAV RTK systems were insufficient for achieving decimeter-level accuracy. Including three to five GCPs significantly enhanced accuracy, with no substantial benefits beyond five GCPs. This finding suggests that an optimal GCP configuration balances the number and placement to minimize errors efficiently. The research provides practical recommendations for UAV photogrammetry, emphasizing the need for adequate GCPs in complex terrains. Future studies should consider increasing the number of control points for better RMSE analysis and spatial error verification, and flight planning based on surface characteristics is advised to improve data accuracy in topographically complex areas. The findings and recommendations presented in this study offer valuable insights for researchers and practitioners using UAV photogrammetry, contributing to more precise and reliable data for various applications.