Evaluating the Effectiveness of the DTM Slope-Based Filter in Stripping Off Above Ground Objects From Digital Aerial Imagery DSM in Rural Landscapes

Abstract

Generation of Digital Terrain Models (DTMs) in wide areas constitutes considerable challenges for base mapping. For DTM generation nonground cells have to be eliminated from Digital Surface Models (DSMs) through filtering processing of the DSMs. Different filtering approaches have been developped for surface smoothing and attenuation of the DSM high frequencies including the DTM slope-based filter. This research aimed at assessment of the efficiency of the DTM slope-based filter in stripping off nonground objects from digital aerial imagery DSMs in rural landscapes. The test site located at the west bank of a river at Vaihingen, Germany has a variety of rural landscape features where a DSM of 0.09-meter resolution extracted from digital aerial imagery stereo pairs has been employed in the study. Bare earth models and removed object layers created from filtering the DSM as well as closed gap bare earth models have been qualitatively and quantitively analyzed along with accuracy estimation of the closed gap models. The analysis showed nodata gaps at the positions of the removed cells that increase in sizes with the increases in the filter radius. Also, the standard deviation of the bare model decreased by 12% with filter radius of 37 grid cells. However, with filter radius of 25 cells, the standard deviation of the elevation residuals increased by 1.53% while with filter radius of 29 cells that standard deviation increased by 13.62%. This indicates rapid deterioration in the accuracy the bare earth models created with filter radius greater than 25 grid cells.