Surface Modelling Based on Unmanned Aerial Vehicle Photogrammetry and Its Accuracy Assessment

Our study provides insight into modelling and assessment of a digital surface model of experimental plot. An unmanned aerial vehicle equipped with low-cost RGB camera was used. The camera system made it possible to acquire very high resolution imagery to generate ortho-mosaics and digital surface models through automatic 3D photo-reconstruction methods by Agisoft Photoscan solution. The unmanned aerial vehicle was operated manually by a pilot over the suggested study sites with concentrating to the problem of relief. The flight plan included acrossand along-track overlaps (i.e. about 30 % and 60 %) using parallel flight lines. There were acquired 251 perpendicular images per 11.5 ha of the suggested plot, taken from an altitude of 50 m above the ground. Ortho-mosaic resolution in set height was 0.02 m·pixel. Registration to the coordinate system was done by ground control points (GCP) measured by the real-time kinematic global positioning system. In this purpose, 21GCP were surveyed. Over 24 million georeferenced points were used in the digital surface generation process. The resulting digital surface model was compared with various numbers of GCP used for orthophoto generation. Accuracy of the digital surface model with various numbers of GCP used was evaluated with the help of the Root Mean Square Error (RMSE) method. The RMSE value was 0.29m with the number of 21accuracy checkpoints measured by GPS in real-time kinematic mode per 11.5 ha.