An In-Depth UAS-PPK Georeferencing Analysis at Different Flight Heights for an Urban Area

Acquiring UAS images in urban environments presents a significant importance for various applications but presents challenges, particularly with flight missions and the reliability of the Global Navigation Satellite System (GNSS) signal. Georeferencing the UAS images in real-world locations is typically done through GNSS-assisted Real-Time Kinematic (RTK)/ Post-Processing Kinematic (PPK) or indirect georeferencing methods. Due to the time-intensive nature for ground-based measurement and materialization of Ground Control Points (GCPs), the GNSS-assisted RTK/PPK method is often preferred. Nevertheless, the accuracy provided by onboard devices is highly contingent on various factors. In densely built-up areas, RTK positioning is prone to signal interruptions, signal losses, radio link outages, and multipath effects, the PPK georeferencing offers a more reliable and precise solution. The aim of this research is to perform an in-depth analysis of the PPK georeferencing results when using three different Continuously Operating Reference Station (CORS) stations and one local base station, with focus only on nadir flights. Two flights were conducted over an 8.6 ha urban area at 60 m and 100 m heights with a DJI Phantom 4 Pro v2 equipped with a TeoKIT GNSS PPK module. The accuracy of the PPK georeferencing process was assessed using 33 Check Points (ChPs) across different scenarios with varying numbers of GCPs and PPK trajectory computations derived from distinct base stations. Comparing the results at 100 m and 60 m flight heights, the best accuracy, with an RMSE of 3.7 cm for all CORS stations, was achieved at the 100 m flight height, reflecting an accuracy improvement of approximately 43% for the RS2 local base station and 75% for the CORS stations with respect to 60 m flight height. The effectiveness of our proposed workflow is demonstrated by the results obtained for the RS2 local base station at 60 m flight height, achieving the highest accuracy (9.2 cm) compared to the CORS stations in the 0 GCP scenario, with a 30%-40% improvement due to good satellite visibility and low levels of multipath and SNR interference. However, the RMS values for the 60 m flight were four times worse than the 100 m flight, primarily due to multipath effects and signal interruptions.