Sparse reconstruction in monostatic and multistatic SAR
Embargo Date
Indefinite
OA Version
Citation
Abstract
The focus of this thesis is the development of robust signal processing paradigms for synthetic aperture radar (SAR) imaging problems with non-conventional SAR configurations. Conventional SAR systems with collocated transmit and receive antennas are designed for imaging of stationary ground scenes over narrow aspect observation apertures and conventional methods for SAR imaging require regular, dense sampling of measurement space. Conventional SAR is a mature technology and early challenges in producing high-quality SAR imagery are now replaced by challenges of integration of SAR into larger multi-mode radar systems capable of complex pervasive area sensing missions. Multi-mode operation and collection of better target attributes to aid pervasive sensing impose non-conventional configurations on SAR. Non-dense, reduced data sampling of the measurement space due to gapped aperture sensing and multistatic geometries, large aperture extents, as well as target motion violate the assumptions used in conventional SAR algorithms.
In this thesis we address new challenges which arise from such non-conventional SAR configurations. First, we utilize insights from the compressed sensing literature in the design of sensor geometries for imaging of point-like scattering with reduced data requirements and we explore the trade-offs of different monostatic and multistatic geometries. We propose to use the coherence parameter associated with a sensing geometry as a reconstruction quality predictor. We demonstrate that a sparse reconstruction method outperforms the conventional method even for non-optimized sampling patterns in terms of coherent-change detection capabilities arising in persistent area surveillance. Additionally, we propose a multi-component reconstruction algorithm that addresses spatially non-localized scattering by its sparse representation in a learned, redundant dictionary. This algorithm reduces background speckle noise, while preserving high-resolution strong scattering.
Another challenge is imaging of moving targets. We propose a method that extracts target motion attributes in addition to creating focused images using multistatic sensing. When sensing over large apertures, the isotropic scattering assumption of the scene is violated and conventional algorithms perform poorly and lead to target defocusing. We develop an algorithm that removes such distortions and jointly estimates targets: reflectivity responses over wide aspect angles.
Finally, we adapt and apply methods that utilize sparse representations m a learned, redundant dictionaries to other applications such as fast angiographic imaging in optical coherence tomography (OCT) and low-dose imaging in X-ray computed tomography (CT).
Description
Thesis (Ph.D.)--Boston University
PLEASE NOTE: Boston University Libraries did not receive an Authorization To Manage form for this thesis or dissertation. It is therefore not openly accessible, though it may be available by request. If you are the author or principal advisor of this work and would like to request open access for it, please contact us at open-help@bu.edu. Thank you.
PLEASE NOTE: Boston University Libraries did not receive an Authorization To Manage form for this thesis or dissertation. It is therefore not openly accessible, though it may be available by request. If you are the author or principal advisor of this work and would like to request open access for it, please contact us at open-help@bu.edu. Thank you.