Assessing Impacts of Grass on Vertical Accuracy of Digital Surface Models Derived from Unmanned Aircraft Systems Imagery

Abstract

Digital surface models (DSM), which are 3D representations of the Earth's surface including all natural and built features, can be created many ways, including directly through lidar acquisitions or indirectly through photogrammetric manipulation of aerial imagery. DSMs created using aerial imagery collected via unmanned aircraft systems are prone to having high elevation error in places with vegetation, because the photographs capture only the top-most surface and do not penetrate to the underlying topography. While many studies have acknowledged this error, few have quantified its magnitude and extent especially for areas involving wild grasses or areas of drastic elevation change. While a distinct correlation between grass height and DSM vertical accuracy was not found, this study assesses possible reasons for model inaccuracy and possible future improvement.