Register Now

Also, remember to make your
Hotel Reservation Here

Important Dates

Conference: May 7-11, 2012
Reg. Registration Ends: April 20, 2012
Late Registration: April 21 to May 11


SAR Image Exploitation and Compressive Sensing

Information

Date: May 7, 2011
Time: 1:00pm - 5:00pm
Instructor: Les Novak and Ivana Stojanovic

Tutorial Code: T-05

Abstract

Part 1 of this tutorial will focus on SAR image exploitation and the application of coherent, non-coherent, and multi-polarization change detection algorithms to multi-pass complex SAR imagery. Performance of an optimum coherent change detection (CCD) algorithm (an MLE coherence estimator) will be reviewed and examples of CCD performance using Ku- and Ka-band SAR imagery from Sandia SAR sensors will be shown. New CCD performance results obtained using X-band SAR imagery gathered by the General Dynamics Data Collection System (DCS) for AFRL's persistent surveillance “Gotcha” program will be shown; performance of an optimum non-coherent change detection (NCCD) algorithm (a “generalized” likelihood ratio test) will be reviewed and compared with CCD results. Full-polarization change detection (POLCD) performance using Gotcha SAR data will be shown and compared with dual and single polarization results. Foliage penetration (FOPEN) SAR change detection results will be presented using VFH-band imagery gathered by the Swedish CARABAS-II system during a FOPEN CD-experiment in Sweden. The following change detection cases will be presented, including: images gathered at identical aspects; images gathered at 5-degree aspect difference; images gathered at 90-degree aspect difference; and multi-pass images. PD/PFA curves of CCD, NCCD, and single-pass detection performance will be compared. We will also demonstrate the effects of data compression on SAR change detection; we compare and quantify change detection performance using various compression techniques (BAQ phase-history compression, image-based Wavelet compression, etc.).

Part 2 of this tutorial will focus on applications of Compressive Sensing (CS) to SAR image formation. The use of active array antennas in airborne platforms allows the radar to accommodate multiple modes, and it may be necessary to interrupt the SAR data collection periodically or randomly to perform other radar functions. Such interruptions leave data-gaps in the SAR phase-history and lead to degraded imagery. We apply CS-methods to the reconstruction of complex SAR imagery from interrupted phase-history data; we extrapolate the missing data by applying a basis pursuit denoising (BPDN) algorithm in the reconstruction process. Example images formed from periodically or randomly gapped phase-histories are compared before and after CS-processing -- and to illustrate the capabilities of the BPDN reconstruction algorithm we evaluate CCD and NCCD change detection performance using Gotcha SAR Imagery. Also, using Sandia SAR imagery containing a set of military vehicles situated in a meadow area, we demonstrate the excellent SAR image quality achieved from BPDN processing of the gapped phase-history data. Time permitting, we will discuss wide-angle SAR image formation; when observed over wide aspect angles, point target scattering is highly anisotropic. We will present an algorithm that estimates the entire three dimensional angle-dependent scattering field. The algorithm’s performance will be demonstrated on the BACKHOE data set, showing its robustness to data loss.

Instructor Biography

PhotoDr. Les Novak is a Consulting Scientist at Scientific Systems Company, Inc. (SSCI), which he joined in February 2006. At SSCI he performed research on the effects of data compression on coherent and non-coherent SAR change detection -- and currently is investigating the effects of phase-history data-gaps on SAR image quality.  He was at MIT Lincoln Laboratory from 1977 to 2003 where he held the position Senior Staff in the Sensor Exploitation Group. At Lincoln Lab he worked on the development of target detection, acquisition, and classification algorithms for synthetic aperture radar systems. He also performed studies of polarimetric radar signal processing algorithms and super-resolution signal processing algorithms. He developed the polarimetric change detection algorithm (POLCD) used on the DARPA WATCH-IT Program; the POLCD algorithm was transitioned to the FOPEN ATD SAR System where real-time, in-the-air change detection was demonstrated; the POLCD change detection algorithm was transitioned by DARPA to the NGA SAR Exploitation System. From December 2003 to February 2006 he performed research on SAR change detection and model-based ATR at Alphatech/BAE.  

PhotoIvana Stojanovic is a Senior Research Engineer at Scientific Systems Company, Inc. Ivana received the Dipl. Ing. Degree from the School of Electrical Engineering, University of Belgrade, Serbia in 1999.  She worked at Iospan Wireless, CA from 1999-2002 as a wireless systems engineer contributing to the development of the world's first MIMO-OFDM fixed broadband wireless access system. From 2002-2005 she was with Intel, Santa Clara, CA  as a senior engineer, contributing to the design of the standardized IEEE 802.16 WiMax system. She received the M.S. degree in 2007 and the Ph.D. degree in 2012, both in Electrical and Computer Engineering at Boston University. Her thesis research focused on multi-dimensional signal processing, inverse problems, and compressed sensing with the applications to synthetic aperture radar and medical imaging problems such as optical coherence tomography (OCT) and low-dose X-ray Computed Tomography (CT).