Exploiting sparsity for persistent scatterer detection to aid X-band airborne SAR tomography
This thesis evaluates the potential for using line of sight returns and return signals from underneath a forest canopy using X-band, airborne synthetic aperture radar (SAR) tomography. Approximately 30% of the Earth’s land surface is covered by vegetation, therefore global digital elevation models (DEMs) contain a signal from the forest canopy and not the ground. By uncovering new techniques to find the ground signals, using data collected from airborne platforms as verification, such procedures could be applied to currently operational and future X-band, spaceborne systems with the aim of resolving much of the vegetation bias on an international scale. Data from three sources is presented; data collected from Selex ES’s SAR systems, the GOTCHA dataset and simulated data. Before carrying out tomography it is shown that SAR interferometry (InSAR) can successfully be applied to X-band, helicopter data. A scatterer defined as a candidate persistent scatterer (CPS) is introduced, where the pixels are stable and coherent over a matter of days. An algorithm for selecting CPSs is developed by exploiting sparsity and a novel choice of hard thresholding operator. Using simulated forestry and SAR information the effects of changing input parameters on the outcome of the tomographic profile is analysed. What is found in this study is that model simulations demonstrate that ground points can be detected if the platform motion is relatively stable and that temporal decorrelation over the forest volume is kept to a minimal. An understory can confuse the tomographic profile since less line of sight observations can be made. By combining line of sight observations alongside new tomography techniques on high resolution SAR data this thesis shows it is possible to detect ground scatterers, even at X-band.