Accuracy analysis of UAV aerial photogrammetry based on RTK mode, flight height, and number of GCPs

• 1、Guangxi Key Laboratory of Marine Environmental Change and Disaster in Beibu Gulf, Beibu Gulf University, Qinzhou 535011
• 2、College of Marine Sciences, Beibu Gulf University, Qinzhou 535011

Abstract：How to optimize the UAV (Unmanned Aerial Vehicle) aerial photogrammetry scheme to obtain high-precision mapping results is crucial for industrial applications. Based on the summary of many influencing factors, three representative factors, namely RTK (Real-Time Kinematic) mode, flight altitude, and the number of GCPs (Ground Control Points), were selected to analyze the impact of UAV aerial photogrammetry accuracy. Using the latest Mavic 3 industrial UAV launched by DJI, four flight altitude tests with eight sorties were conducted under two RTK modes. The output of 40 aerial photogrammetric results was realized through the DJI Terra platform using five different GCP layout methods. Finally, based on the measured coordinates of CPs(Check Points), the accuracy differences of various aerial photogrammetric results were analyzed. The results showed that: (1) There is a perfect correlation between the flight altitude and the resolution of the UAV aerial photogrammet- ric results. Based on this, conversion formulas were established between the actual image resolution and the flight altitude under different GCP numbers in both RTK and non-RTK modes, which can be used to estimate the flight altitude or resulting image resolution based on application requirements. (2) In RTK mode, regardless of whether GCPs are added, the aerial photogrammetric image\'s horizontal and vertical accuracy decreases with increasing flight altitude. In the non-RTK mode without GCPs, the aerial photogrammetric image\'s horizontal RMSE (Root Mean Square Error) generally increases with increasing flight altitude. At the same time, the vertical RMSE decreases with increasing flight altitude. After adding an appropriate number of GCPs, the horizontal and vertical accuracy of the aerial photogrammetric image decreases with increasing flight altitude innon-RTK mode. (3) When the flight altitude is the same, whether to add GCPs and their number changes have no significant effect on the improvement of the accuracy of aerial photography under RTK mode but can significantly affect the accuracy of aerial photography under the non-RTK mode, that is, the horizontal RMSE and vertical RMSE of the aerial photogrammetric image both decrease rapidly with increasing GCP number before stabilizing. However, whether RTK is turned on or not, excessive GCPs are not conducive to improving the accuracy of aerial photography results. (4) The mapping accuracy of the UAV under RTK mode without GCPs reaches the same level as that under non-RTK mode with GCPs. Therefore, deploying GCPs when performing aerial photography tasks based on the RTK-UAV is unnecessary under appropriate conditions. On this basis, suggestions for optimizing the UAV aerial photogrammetry scheme were summarized, which can provide a reference for related applications.

Keywords： UAV, accuracy, RTK, flight altitude, ground control points, Mavic 3