An estimation-theoretic technique for motion-compensated synthetic-aperture array imaging

• Main Author: Logan, Cedric Leonard
• Other Authors: Alan S. Willsky.
• Format: Others
• Language: English
• Published: Massachusetts Institute of Technology 2005
• Subjects: Electrical Engineering and Computer Science.
• Online Access: http://hdl.handle.net/1721.1/9122

Thesis (Sc.D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2000. === Vita. === Includes bibliographical references (p. 351-354). === Synthetic-Aperture Radar (SAR) is an imaging technique that achieves high azimuth resolution by using coherent processing to exploit the relative motion between an airborne or spaceborne radar antenna and the imaged target field (effectively synthesizing the effect of a larger aperture array). From an estimation-theoretic perspective, this thesis addresses the following limitations of conventional imaging techniques for the spotlight-mode version of SAR: sidelobe imaging artifacts and loss of resolution for stationary SAR scenes containing high-amplitude scatterers, and blurring and object-displacement artifacts in the presence of moving targets. First, this thesis presents a generalized estimation-theoretic SAR imaging framework which exploits the idea of L1-norm regularization. Some results are included which demonstrate the utility of this approach for reducing sidelobes and improving resolution for stationary SAR images. A parameterized L-norm-based moving-target imaging technique is also presented. For the case of a single moving target, this technique is able to compensate for the blurring due to temporally-constant velocity rigid-body motion (even if the target scatterers are closely-spaced). However, the motion-induced object-displacement compensation performance of this technique is significantly affected by velocity estimation errors. This thesis also presents an estimation-theoretic moving-target SAR imaging framework which uses a multi-dimensional matched-filter for computing a set of scatterer-velocity estimates which are used as initial conditions for an L1-norm-based estimation algorithm which assumes that the target scatterers have temporally-constant spatially-independent velocities. Therefore, this framework is able to image a moving target and nearby high-amplitude stationary clutter simultaneously. This framework also shows potential for imaging targets with non-rigid body motion. However, the motion-induced object-displacement compensation performance of this approach is significantly affected by cross-scatterer interference effects. === by Cedric Leonard Logan. === Sc.D.