Abstract:
This thesis describes the orthorectification process for commercial satellite Worldview2 Multi-Spectral Imagery (MSI) with a Rational Polynomial Coefficient (RPC) sensor model by using a Digital Elevation Model (DEM) raster of terrain heights from Light Distance and Ranging (LiDAR) data. The orthorectification process is presented that uses a LiDAR DEM raster with Ground Sample Distance (GSD) that matches the nearly 0.5 meters nominal GSD of the panchromatic overall spectral brightness, and about 2 meters nominal GSD of each spectral band, for the pixel footprint of the Worldview2 unrectified MSI. It presents a process to encourage adding extra DEM layers such as slope and aspect into the MSI feature extraction process. This study tests a hypothesis to see if the orthorectification process produces adequate registration of the DEM and orthoimage for combined spectral and terrain reasoning, by measuring the offsets between the input DEM and the output orthoimage.
The accuracy assessment is directed within an urbanized landscape that contains numerous elevation discontinuities, where the vertical sidewalls of buildings cause cliffs within the terrain surface. Rendering anomalies of feature ghosts from the elevation discontinuities are described along with a suggested solution to the problem. A method also is presented that measures positional inaccuracy between conjugate features within the image of the terrain surface heights and within the initial orthoimage, and then it entails removal of the measured systematic error (shift) within a reproduced orthoimage. This study suggests or provides solutions to identified problems within conventional orthorectification processing regardless of the sensor model. This affirms the alternative hypothesis that the conventional orthorectification process can be adjusted to produce sufficient registration of the DEM and orthoimage for combined spectral and terrain reasoning, if the suggestions from this thesis are implemented.
