Digital Elevation Model Production Using Point Cloud Acquired by Unmanned Aerial Vehicles

Abstract

Currently, point cloud data acquired by using unmanned aerial vehicles (UAVs) are mostly used for the production of digital surface models (DSMs). This paper shows the possibility of digital elevation model (DEM) production with point cloud data acquired using UAVs. In this study, which was conducted in Korea, we used 314 images acquired with a DJI Inspire-2 UAV. To extract ground data, we performed point cloud auto-classification with six types of software, Pix4DMapper, GlobalMapper, Inpho, Trimble Business Center (TBC), Metashape, and Terrascan, and the obtained results were compared. Pix4DMapper was used for point cloud extraction, and all six types of software used the default options for point cloud autoclassification. A point cloud acquired using LiDAR classifies vegetation using an echo, but a point cloud extracted from an image has no echo, making it impossible to classify vegetation using an echo. In this study, DEMs were produced using ground data classified automatically with the six types of software, and they were compared with a DEM produced by the Korean National Geographic Information Institute (NGII). As a result, the DEM error rate was 38-47% depending on the type of software used. In a mountainous area, the dense forest made it impossible to extract ground data, resulting in a very high error rate of 82-92%.