Využitie UAV technológie pre klasifikáciu a mapovanie krajiny vo fluviálnej geomorfológii

Abstract

The aim of this paper is to present the possibilities of UAVs (Unmanned Aerial Vehicles) as photogrammetry payload carriers for data acquisition and fluvial landform identification and mapping. The manual and automatic classification of the Belá River riparian zone for landscape object identification and the analyses of the point cloud density after vegetation filtration was performed. The HEXAKOPTER XL including the Sony NEX 6 camera with 16 – 50 mm lens for landscape monitoring features was used. Data was processed in Agisoft PhotoScan software. The RMSE (root mean square error) of aligned images was 60.121 mm (x coordinate), 43.7584 mm (y coordinate) and 29.46 mm (z coordinate). The resulting point cloud was semiautomatic classified in the software Terrasolid – Terrascan (Microstation), in the following six classes: high vegetation (over 5 m), medium vegetation (from 1.5 m to 5 m), small vegetation (from 0.2 m to 1.5 m), topographic surface and water surface. Orthophotomosaic was classified in ArcGIS software by supervised Maximum Likelihood Classification (MLC). Here training site signatures identified the five land cover categories (water area, bar surface, vegetation, Large Woody Debris – LWD and bare surface). The classification of photogrammetric derived point clouds increases the accuracy elevation model, but on the other hand, does not capture the real terrain and topography under the vegetation.