SIXTH ANNUAL
ASAP '98 WORKSHOP

Adaptive SAR Imaging
and Its Impact on ATD/R
and
Image Exploitation

 

 

Stuart R. DeGraaf
Northrop Grumman Corporation
Electronic Sensors and Systems Division
P.O. Box 746, MS 1155
Baltimore, MD 21203
email: srd@cestua.md.essd.northgrum.com

Abstract Conventional spotlight SAR imaging exploits the Fourier transform pair relationship between signal history and scene reflectivity, and therefore employs a weighted 2-D FFT to form a SAR image. Since SAR collection apertures are of finite size in wavenumber space, the spatial resolution afforded by Fourier imaging is inherently limited. Finite resolution leads to speckle or scintillation, which is caused by random constructive and destructive interference of independent unresolved scatterers. Modern spectral estimation (super resolution) techniques offer improved image resolution and contrast, and more importantly, reduced clutter speckle and target scintillation. Ongoing research at Northrop Grumman has shown that two methods offer enormous benefits for SAR imaging and automatic target detection and recognition. The minimum variance method (MVM) and Pisarenko's method reduce target scintillation and clutter speckle through the signal history domain forward-backward subaperture averaging for covariance matrix estimation. MVM (and Pisarenko) make homogeneous clutter distributions appear more log-normal. As a result, MVM reduces the false alarm rate of a standard two-parameter CFAR screener by more than an order-of-magnitude and MVM imagery takes on a more "optical" appearance. Classical image processing tools, such as Laplacian sharpening filters, work better on MVM and Pisarenko imagery than on Fourier. Pisarenko's method reduces misclassification of military vehicles by roughly a factor of three, and may reduce the number of target templates required for target recognition. MVM improves image compressibility by an order-of-magnitude. MVM improves Bayesian SAR scene segmentation performance. Modern spectral estimation methods promise to improve dramatically all facets of SAR imaging and exploitation.

The presentation will review the rationale for MVM and Pisarenko imaging methods, show examples of adaptive SAR imagery, and illustrate our ATR results.



Disclaimer

Direct comments and questions to: webmaster@ll.mit.edu

MIT Lincoln Laboratory. All rights reserved.