Development of a stream DTM generation methodology using UAV-based SfM and LiDAR point cloud

Filtering ground points and generating accurate Digital Terrain Models (DTMs) are crucial for hydrological and ecological modeling in riverine environments. This paper proposes a methodology for DTM generation in stream areas by integrating point clouds from a UAV-based structure from motion (SfM) and light detection and ranging (LiDAR) while applying ground filtering algorithms. The study area, the Bokha Stream in Icheon City, South Korea, was surveyed using Zenmuse L1 (LiDAR) and Phantom 4 multispectral (SfM). Water and non-water areas were classified using the normalized difference water index (NDWI), and three ground filters, cloth simulation filter (CSF), progressive triangular irregular network (PTIN), and simple morphological filter (SMRF), were applied to remove vegetation and error points. The methodology was validated using 11 cross-sections, with accuracy assessed via mean absolute error (MAE) and root mean square error (RMSE). Among the filters, SMRF demonstrated superior performance, achieving an MAE of 0.160 m and an RMSE of 0.214 m. It effectively removed vegetation and preserved riparian terrain but tended to underestimate elevations in water areas. The results highlight the potential of the proposed method to enhance the accuracy of DTMs, offering benefits for applications in mesoscale hydrological and ecological modeling in small streams.