TOPOGRAPHY-PRESERVING, NON-LINEAR INPAINTING FOR AUTONOMOUS BARE EARTH DIGITAL ELEVATION MODEL (DEM) RECONSTRUCTION

Harris describes a novel way to autonomously inpaint missing data into high resolution single reflective surfaces utilizing a variant of the Navier-Stokes equations. One product of this process is a high-resolution bare earth Digital Elevation Model (DEM) with the same resolution as the input data. Inpainting allows generation of high resolution bare earth DEMs in both high and low frequency terrain environments for urban 3-D modeling. Having this bare earth DEM accounts for a dramatic increase in accuracy in all other steps of the urban 3-D modeling process. The LiteSite™ toolkit has the capability to automatically extract buildings and vegetation from an urban scene. The resulting DEM from this step of the process acts as input to the inpainting process. The expected building and vegetation base heights can then be inpainted into the area of extraction where data is now missing. The inpainting process maintains building and vegetation base height consistency in the inpainted regions and performs high accuracy interpolation and edge propagation on DEMs. Inpainting is highly effective for data sets where the occurrence of missing data is common and undesired. Examples of such data include Shuttle Radar Topography Mission (SRTM), LIDAR, IFSAR, correlated DEMs from imagery, or any other single reflectance collection data set. This technology preserves height contours. A more accurate bare earth product allows for better automated building vector extraction and therefore reduces manual building vector editing. Automated texturing of 3D model products with aerial or satellite imagery is accomplished using Harris’ RealSite TM Toolkit.