SYNTHETIC APERTURE RADAR SIGNAL PROCESSING with MATLAB Algorithms Mehrdad Soumekh ® A Wiley-Interscience Publication JOHN WILEY & SONS, INC. New York / Chichester / Weinheim / Brisbane / Singapore / Toronto‘This book is printed on acid-free paper. €) Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. Published simultaneously in Canada, No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 646-8600. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., I11 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008. To order books or for customer service please, call 1(800)-CALL-WILEY (225-5945), Library of Congress Cataloging-in-Publication Data: Soumekh, Mehrdad. ‘Synthetic aperture radar signal processing with MATLAB algorithms 7 Mehrdad Soumekh p. cm. “A Wiley-Interscience publication.” Includes bibliographical references and index. ISBN 0-471-29706-2 (cloth) L. Synthetic aperture radar. 2. Signal processing—Mathematies 3. MATLAB. I. Title. TK6592.S95S68 1999 621.3848—de21 98-29322 Printed in the United States of America 10To My Beloved Parents Mrs. Malkah and Dr. Eliahou Soumekh and Hertzel, Flora, Lyda, Fereshteh, Benhoor, and ShiaoyanCONTENTS Preface xv Introduction xix 1 Synthetic Aperture Radar / xix The Book / xxi Organization / xxii SAR and ISAR Databases / xxiv Range Imaging 1 Ll 1.2 13 14 LS 16 System Model / 7 Reconstruction via Matched Filtering / 14 Range Solution / 17 Data Acquisition and Signal Processing / 17 Time Domain Sampling | 18 Time Interval of Sampling 1 19 Number of Time Samples | 19 Reconstruction Algorithm / 20 Reconstruction via Pulse Compression for Chirp Signals / 23 Signal Model | 23 Reconstruction | 25 Range Resolution | 28 Time Domain Sampling | 28 Residual Video Phase Error | 31 Upsampling to Recover Alias-Free Echoed Signal | 31 Electronic Counter-Countermeasure (ECCM) via Amplitude Modulation of Chirp Signals | 34 viiviii = CONTENTS 1.7 Frequency-Dependent Target Reflectivity / 39 Reconstruction via Target Signature Matched Filtering | 44 1.8 MATLAB Algorithms / 45 2 Cross-range Imaging 47 2.1. System Model / 54 2.2. Spherical PM Signal within an Infinite Aperture / 60 2.3. Reconstruction via Matched Filtering: Infinite Aperture / 64 Synthetic Aperture (Slow-time) Sampling | 66 Cross-range Resolution | 66 2.4 Spherical PM Signal within a Finite Aperture / 67 Instantaneous Frequency | 67 Slow-time Fourier Transform | 68 Slow-time Angular Doppler Spectrum | 71 2.5 Reconstruction via Matched Filtering: Finite Aperture / 72 2.6 Cross-range Resolution / 75 2.7. Data Acquisition and Signal Processing / 75 Synthetic Aperture Sampling for a Broadside Target Area | 75 Synthetic Aperture Sampling for a Squint Target Area | 78 Reducing PRF via Slow-time Compression | 80 Cross-range Gating via Slow-time Compression | 88 2.8 Reconstruction Algorithm / 89 Baseband Conversion of Target Area | 90 Zero-padding in Synthetic Aperture Domain | 91 Slow-time Doppler Domain Subsampling | 93 Reducing Bandwidth of Reconstructed Image | 94 2.9 Synthetic Aperture-Dependent Target Reflectivity / 95 eae AM-PM Signal Model | 96 Slow-time Fourier Transform of AM-PM Signal / 98 Reconstruction | 114 Representation in Slow-time Angular Doppler Domain | 121 Reconstruction via Target Signature Slow-time Matched Filtering / 127 Type 1: Generalization of Spotlight SAR 1 127 Type 2: Generalization of Stripmap SAR I 130 Type 3: Partial Observability | 135 MATLAB Algorithms / 135CONTENTS = ix 3 SAR Radiation Pattern 137 ae 3.2 3.3 34 3.5 3.6 Transmit Mode Radar Radiation Pattern / 142 Synthetic Aperture (Slow-time) Dependence | 156 Radiation Pattern in Three-Dimensional Spatial Domain / 157 Radar Footprint | 157 Slant-Range | 162 ‘Transmit-Receive Mode Radar Radiation Pattern / 163 ‘Transmit-Receive Mode Radar-Target Radiation Pattern / 165 Polarization / 171 MATLAB Algorithms / 174 4 Generic Synthetic Aperture Radar 176 41 4.2 43 44 45 4.6 47 48 49 4.10 System Model / 186 Fast-time Fourier Transform / 195 Slow-time Fourier Transform / 196 Reconstruction / 196 Digital Reconstruction via Spatial Frequency Interpolation / 198 Baseband Conversion of Target Area | 201 Interpolation from Evenly Spaced Data | 203 Interpolation from Unevenly Spaced Data | 204 Digital Reconstruction via Range Stacking / 206 Digital Reconstruction via Time Domain Correlation and Backprojection / 212 Time Domain Correlation Algorithm | 212 Backprojection Algorithm | 214 Frequency and Synthetic Aperture Dependent Target Reflectivity / 215 Motion Compensation Using Global Positioning System / 217 Spatial Frequency Modeling of Motion Errors | 218 Narrow-Beamwidth Motion Compensation | 219 Wide-Beamwidth Motion Compensation | 223 Three-Dimensional Wide-Beamwidth Motion Compensation | 226 Motion Compensation for Backprojection | 230 Motion Compensation Using In-Scene Targets / 230 Narrow-Beamwidth Motion Compensation | 230x CONTENTS 4.11 4.12 4.13 414 4.15 Wide-Beamwidth Motion Compensation | 231 Three-Dimensional Wide-Beamwidth Motion Compensation | 234 Polar Format Processing / 234 Plane Wave Approximation-Based Reconstruction | 234 Narrow-Beamwidth Approximation | 238 Narrow-Bandwidth and Narrow-Beamwidth Approximation | 239 Wavefront Curvature Compensation | 242 Motion Compensation Using Global Positioning System | 243 Conventional ISAR Modeling and Imaging / 244 ISAR Model | 246 Slow-time Compression or Motion Compensation | 246 Polar Format Processing | 247 Range-Doppler Imaging / 249 Fresnel Approximation-Based Reconstruction | 251 Narrow-Bandwidth and Narrow-Beamwidth Approximation | 255 Three-Dimensional Imaging with Two-Dimensional Azimuth and Elevation Synthetic Apertures / 256 System Model | 257 Reconstruction | 258 Electronic Counter-Countermeasure via Pulse Diversity / 259 Spotlight Synthetic Aperture Radar 262 5.1 5.2 53 54 55 Mechanically Beam-Steered Spotlight SAR / 270 Mechanical Beam Steering | 270 System Model | 274 Reconstruction | 274 Electronically Beam-Steered Spotlight SAR / 277 Electronic Beam Steering { 277 System Model | 280 Reconstruction | 281 Bandwidth of Spotlight SAR Signal / 282 Single Target | 283 Target Area | 286 Resolution and Point Spread Function / 290 Data Acquisition and Signal Processing / 293 Fast-time Domain Sampling and Processing | 2945.6 CONTENTS Xi Slow-time Domain Sampling and Processing / 297 Reducing PRF via Slow-time Compression / 302 Digital Spotlighting | 306 Subaperture Digital Spotlighting 1 317 Reconstruction Algorithms and SAR Image Processing / 319 Digital Reconstruction via Spatial Frequency Interpolation | 319 Reconstruction in Squint Spatial Coordinates | 329 Slow-time Doppler Domain Subsampling | 335 Reducing Bandwidth of Reconstructed Image | 336 Digital Reconstruction via Range Stacking | 349 Digital Reconstruction via Time Domain Correlation and Backprojection | 354 Effect of Slow-time Doppler Filtering | 362 Effect of Motion Errors in Slow-time Doppler Spectrum | 369 5.7. MATLAB Algorithms / 370 6 Stripmap Synthetic Aperture Radar 373 6.1 System Model / 381 6.2 63 64 65 6.6 Radar Radiation Pattern | 382 Stripmap SAR Signal Model | 384 Reconstruction / 387 Bandwidth of Stripmap SAR Signal / 388 Planar Radar Aperture | 389 Curved Radar Aperture | 391 Resolution and Point Spread Function / 391 Data Acquisition and Signal Processing / 397 Fast-time Domain Sampling and Processing | 398 Slow-time Domain Sampling and Processing | 402 Slow-time Compression and Processing | 404 Subaperture Digital Spotlighting | 409 Reducing Side Lobes Doppler Aliasing via Slow-time Upsampling | 415 Reconstruction Algorithms and SAR Image Processing / 418 Digital Reconstruction via Spatial Frequency Interpolation | 418 Slow-time Doppler Domain Subsampling | 425 Reducing Bandwidth of Reconstructed Image { 428 Digital Reconstruction via Range Stacking | 445,xii CONTENTS: 67 68 Digital Reconstruction via Time Domain Correlation ‘and Backprojection | 446 Effect of Beamwidth (Slow-time Doppler) Filtering | 451 Effect of Motion Errors in Slow-time Doppler Spectrum | 453 Subpatch “Mosaic” Digital Reconstruction with Subaperture Data | 456 Moving Target Detection and Imaging / 465 SAR Signal Model for a Moving Target with a Constant Velocity | 467 Three-Dimensional Imaging in Motion-Transformed Spatial Domain and Relative Speed Domain | 412 Moving Target Indicator: SAR Ambiguity Function | 474 MATLAB Algorithms / 477 7 Circular Synthetic Aperture Radar 486 ae pee bea 74 eae 7.6 17 78 System Model / 491 CSAR Signal Model / 493 Fourier Properties of Slant Plane Green's Function | 494 Reconstruction / 497 Slant Plane to Ground Plane Transformation | 498 Ground Plane CSAR Reconstruction | 500 Bandwidth of CSAR Signal / 507 Resolution and Point Spread Function / 509 Full Rotation Aspect Angle Measurement | 509 Partial Rotation Aspect Angle Measurement | 512 Data Acquisition and Signal Processing / 515 Fast-time Domain Sampling and Processing | 515 Slow-time Domain Sampling and Processing | 516 Digital Spotlighting and Clutter Filtering | 517 Reconstruction Algorithms and CSAR Image Processing / 518 Digital Reconstruction via Spatial Frequency Interpolation | 518 Reducing Bandwidth of Reconstructed Image | 521 Digital Reconstruction via Time Domain Correlation and Backprojection | 522 ‘Three-Dimensional Imaging / 525 Target Resolvability from Single-Tone Fringe Patterns | 533 Three-Dimensional Imaging with Two-Dimensional Circular and Elevation Synthetic Apertures / 539CONTENTS xiii System Model | 343 Reconstruction | 545 Digital Reconstruction | 551 8 Monopulse Synthetic Aperture Radar 553 8.1 CH 84 8.5 8.6 Along-Track Moving Target Detector Monopulse SAR / 559 Along-Track Monopulse SAR System Geometry | 561 Monostatic SAR Signal Model / 562 Bistatic SAR Signal Model | 564 Synthesis of Monostatic SAR Signal from Bistatic SAR Signal | 565 Moving Target Indicator | 567 Effect of Variations in Altitude and Nonlinear Motion / 569 Effect of Uncalibrated and Unstable Radars / 570 Amplitude Patterns of Monopulse Radars | 571 Instability of Monopulse Radars | 574 Wide-Beamwidth Monopulse Radars | 574 Signal Subspace Registration of Uncalibrated SAR Images / 575 System Model | 576 Signal Subspace Processing | 578 Estimating Calibration Error Impulse Function | 583 Application in MTD Monopulse SAR | 584 Application in Automatic Target Recognition SAR | 585 Slant Plane Topographic Mapper Monopulse SAR / 587 Slant Plane Monopulse SAR System Geometry | 588 Monostatic and Bistatic SAR Signal Models | 588 Narrow-Bandwidth and Narrow-Beamwidth Approximation: Interferometric SAR (IF-SAR) 1 589 Wide-Bandwidth and Wide-Beamwidth Model | 592 Estimating Slant-range Shift via Signal Subspace Processing | 594 Multistatic Monopulse ISAR / 596 Multistatic ISAR Model | 596 Motion Tracking via Signal Subspace Processing ! 598 MATLAB Algorithms / 603 Bibliography 605 Index 611PREFACE SSS Synthetic aperture radar (SAR) is one of the most advanced engineering inven- tions of the twentieth century. SAR is an airborne or satellite-borne radar system that provides high-resolution maps of remote targets on a terrain, a planet, and so forth. The E-3 AWACS (Airborne Warning and Control System) airplanes that are used extensively in surveillance missions in the Persian Gulf region to detect and track maritime and airborne targets, the E-8C Joint STARS (Surveillance Target At- tack Radar System) airplanes that were used to detect and locate ground targets in the Gulf War, and the NASA space shuttles are equipped with this radar system. SAR systems are a highly developed combination of precision hardware and elec- tronic design for data acquisition, and advanced theoretical principles of mathemat- ies and physics to convert the acquired data to high-resolution images, The origin of the SAR theoretical principles can be traced back to Gabor’s theory of wave- front reconstruction (gab); this theory is also the foundation of many other coherent imaging systems in diverse fields such as geophysical exploration and diagnostic medicine [894]. The utility of the wavefront reconstruction theory in SAR was recognized during the inception of this imaging system in the 1950s and 1960s. However, the lack of fast computing machines and advanced digital information (signal) Processing al- gorithms at that time prevented the development of wavefront reconstruction-based SAR imaging methods. The early SAR systems were based on optical (analog) pro- cessing of the measured echoed signal using the Fresnel approximation for image formation [cut; wil]; this SAR processor, in the analog form or its digital version which was introduced for the spaceborne SAR in the late 1970s [cur], is also known as range-Doppler imaging [br69]. In the 1970s another SAR imaging method, known as polar format Processing, was introduced for the high-resolution spotlight SAR systems [aus; wal]; this method was based on what was known as the plane wave approximation. An extensive the- oretical and practical knowledge base has been shaped over the past 40 years by the approximations inherent in these methods; some of these concepts are either incor- rect or not applicable when viewed in the framework of modern high-resolution SAR systems. The first digital signal processing methods for SAR image formation via the wave- front reconstruction theory were introduced in the late 1980s. SAR wavefront recon- struction theory not only provides a tool for SAR image formation but also reveals xvxvi PREFACE functional properties of the SAR signal that contradict or were not predicted in the approximation-based theoretical foundation of either range-Doppler imaging or po- lar format processing. ‘A prominent example involves the slow-time Doppler bandwidth of the spotlight SAR system. Polar format processing theory states that this bandwidth is propor- tional to the beamwidth angle of the physical radar aperture; this also supports the classical SAR theory principles for the stripmap SAR signal. This slow-time Doppler bandwidth dictates the minimum required pulse repetition frequency (PRF) of the radar. However, SAR wavefront reconstruction theory shows that the slow-time Doppler bandwidth of the spotlight SAR signal is proportional to the beamwidth angle of the physical radar aperture plus the beamwidth look angle of the synthetic aperture (Chapter 5). Unfortunately, most of the users of polar format processing who tried to understand and implement the SAR wavefront reconstruction missed this subtle but very important issue. The processing of aliased slow-time Doppler data in X band spotlight SAR sys- tems has been a key reason for the failure to exploit coherent (complex) target sig- natures for automatic target recognition (ATR). Filtering out the higher slow-time Doppler data is a commonly used scheme to deal with the Doppler aliasing problem; this turns out to be a premature and simplistic approach that removes a significant amount of useful information. ‘The wavefront reconstruction theory reveals unique functional properties of and redundancies in the SAR signal that cannot be found in any other information pro- cessing modality. These redundancies could be exploited to recover uncorrupted data from aliased data. ‘Another classic example of how wavefront reconstruction theory has changed the way SAR signals are viewed involves filtering of the acquired SAR data. Conven- tional frequency domain filtering, for example, applying a Hamming window, is known to improve the quality of data in speech processing. Such filters have also been used in range-Doppler imaging or polar format processing to suppress clutter and/or improve the shape of the point spread function. However, these filters mainly reduce the errors that are due to the approximations in range-Doppler imaging or polar format processing; these errors turn out to be more prominent in the higher slow-time Doppler frequencies where the filter attenuates the SAR data. Based on the SAR wavefront reconstruction theory, applying these filters on SAR data results in loss of useful high Doppler frequency information and, consequently, in loss of resolution. ‘The main message of the above examples is that one cannot view the SAR signal and its properties via the range-Doppler or polar format processing theory and then use the SAR wavefront reconstruction theory for image formation. Unfortunately, most of the SAR hardware systems that have been developed since the 1970s result in aliased SAR data when their acquired data are viewed via the SAR wavefront reconstruction theory. Does this mean that the wealth of the data collected by these SAR systems are contaminated and that these SAR systems must be rebuilt? As we will discover, the answer to this question is no.PREFACE = xvii This book provides a foundation in signal processing and includes digital algo- rithms for the SAR wavefront reconstruction theory. It both establishes the con- straints for acquiring the SAR data from the system design point of view and provides digital signal and image processing algorithms for proper and alias-free implemen- tation of the SAR wavefront reconstruction. As will become clear in this book, the unique properties of the SAR signal cannot be found in any other information pro- cessing system, such as speech processing, or the conventional beamforming meth- ods for array processing. For instance, the unconventional SAR wavefront reconstruction theory shows that under certain conditions, an aliased SAR database can be used to recover its corre- sponding alias-free SAR data (Chapters 2 and 5). This scheme is used to recover the alias-free SAR data of a SAR system that was built based on the polar format processing theory. The SAR wavefront reconstruction theory has also brought new tools for address- ing classical SAR processing issues such as motion compensation and radar cal- ibration (Chapter 4). The motion and calibration errors in SAR can be shown to amplitude modulate the SAR signal which itself is a phase-modulated signal (Chap- ters 2 and 3). The combined AM-PM signal then dictates the point spread function of the SAR system. This concept has been used by Papoulis to analyze the Fourier properties of optical AM-PM signals [pap]; in the 1960s O’Neal also predicted such properties for the SAR signals. ‘A major issue associated with using SAR for automatic target recognition (ATR) is the variations of the radar radiation pattern over time or the calibration problem. Owing to the calibration errors, the point spread function of a target in a SAR image varies from one experiment to another. The AM-PM modeling of uncalibrated SAR signals enables the user to address the calibration problem via the signal subspace least squares methods which are used for blind deconvolution [s98a; wid] (Chap- ter 8). A similar calibration problem is also encountered in moving target detection (MTD) using along-track monopulse SARs. In recent years attempts to use dis- placed phase center arrays (DPCAs) to emulate along-track monopulse MTD-SARs over a relatively long slow-time have been unsuccessful due to the calibration er- rors. Signal subspace processing of uncalibrated along-track monopulse MTD-SARs is also a digital signal processing approach used to encounter calibration errors (Chapter 8). In addition to the traditional linear motion stripmap and spotlight SAR systems (Chapters 5 and 6), newer SAR modalities are also being investigated. One of these new SAR systems involves a circular flight path over a desired (spotlighted) target area, The objective in the circular SAR (CSAR) systems is to acquire the radar signa- ture of the targets in the spotlighted scene over all possible aspect angles (Chapter 7). Such a database is useful for target detection and classification in ATR-SAR prob- Jems. CSAR also provides some form of three-dimensional imaging capability that the linear SAR systems do not possess. Meanwhile, a CSAR signal has functional properties and, as a result, digital signal processing problems that are different from those of the linear SAR signal.xviii PREFACE ‘The theoretical principles in this book are accompanied by results from four sets of realistic SAR, ISAR, and CSAR (turntable) databases. MATLAB* algorithms are also provided for simulation, digital signal processing, and reconstruction of various one-dimensional radar and multidimensional SAR imaging modalities. These algo- rithms are available at the ftp site ftp://ftp.mathworks.com/pub/books/soumekh. The key to the success of a high-resolution SAR system is 90 percent due to good radar hardware, and 10 percent due to intelligent digital signal processing, though some radar scientists might argue that the ratio is likely to be 99 to 1. It can be argued that the development of each of these two aspects of the SAR system based on mature and concrete foundations will enhance the success of the other. Acknowledgments Iam grateful for my collaboration with Rolan Bloomfield, S. I. Chou, Robert Dinger, Robert DiPietro, Ronald Fante, Lawrence Hoff, Liz Jones, Scott Jones, William Kendall, Eric Nelson, David Nobles, Richard Perry, Michael Pollock, Edward Rupp, Roger Rysdyk, Alan Stocker, Steve Tran, Jay Trischman, Michael Wicks, Christo- pher Yerkes, and William Zwolinski. I am thankful to S. I. Chou, Ira Ekhaus, Joseph Fitzgerald, Robert Nelson, Thierry Rastello, and Christopher Yerkes for their com- ments and suggestions in the course of this project. Tam also grateful to the members of the radar groups who collected the ISAR, SAR, and CSAR (turntable) databases cited in this book; these groups are identified in the introductory chapter. I also wish to extend my gratitude to those who provided the data, Many thanks to Betty Brown, Rosalyn Farkas, and Donald DeLand for helping me in editing and preparing this book. The work reported in this book was supported by the following government agencies and private industries: National Science Foundation, Grant MIP-9004996, P. Ramamoorthy, Program Director; Bell Aerospace (Textron), William Zwolin- ski, Technical Coordinator; Space Computer Corporation, Alan Stocker, Technical Coordinator, Summer Faculty Fellowships at Naval Research and Development (SPAWAR Systems Center), Robert Dinger (1992-95) and Lawrence Hoff (1995), Technical Coordinators; Summer Faculty Fellowship at Rome Laboratory, Air Force Office of Scientific Research, Michael Wicks, Technical Coordinator; Naval Re- search and Development, Grants N66001-95-M-1383 and N66001-7052-7595, Michael Pollock, Technical Coordinator, MITRE Corporation, Richard Perry, Technical Coordinator; Office of Naval Research, Grants N00014-96-1-0586 and NO00014-97-1-0966, William Miceli, Program Officer; and Air Force Office of Sci- entific Research, Grant 99-NM-080, John Sjogren, Program Officer. Their generous support of this work is greatly appreciated. MEHRDAD SOUMEKH Buffalo, NY October 1998 *MATLAB® is a registered trademark of the MathWorks, Inc. For MATLAB product in- formation, please contact: The MathWorks, Inc., 3 Apple Hill Drive, Natick, MA 01760-2098 USA.INTRODUCTION SYNTHETIC APERTURE RADAR Vision is perhaps the most critical component of the human sensory system. The lens in the human eye collects the optical waves that are being reflected from the objects in its surrounding medium, which are then interpreted as visual information in the brain. Human-made vision systems imitate the function of the eye-the sensor—and that of the brain—the processor. Human-made vision systems are primarily built to im- prove on our ability to resolve targets, for example, binoculars, radars, and sonars, or to capture the image of a scene, for example, cameras. In either case the ability to view or capture a scene improves with a larger lens aperture (in a binocular or camera), a larger radar antenna aperture, or larger acoustic transducer aperture; the key to better vision is a larger aperture. Unfortunately, it is extremely difficult to either build or maintain a physically large aperture vision system. (Recall the Hubbel telescope.) In the 1950s an invention in radar by Wiley revolutionized the way human-made vision systems are constructed. This invention was called synthetic aperture radar, or SAR. The principal idea behind SAR is to synthesize the effect of a large-aperture physical radar, whose construction is infeasible. What is the significance of a larger aperture? This becomes evident in the following example. The lateral or cross-range resolution of a D = 1-meter diameter radar antenna with wavelength 4 = 1 meter at the range R = 1000 meters is Ri Lateral resolution = pe 1000 meters, which is very poor. Yet, based on the SAR theory and signal processing, if we move this small 1-meter aperture radar along an imaginary aperture with length Deg = 1000 meters (Dem stands for effective radar diameter or aperture), then the lateral resolution with wavelength 2 = 1 meter at the range R = 1000 meters becomes Ri Lateral resolution = —— = | meter; Dest this is a tremendous improvement over the 1000-meter laterai resolution of the small 1-meter physical radar antenna, The imaginary aperture of Derr = 1000 meters is xixxX INTRODUCTION called a synthetic aperture. In recent years the principle of synthetic aperture has also been utilized in other human-made vision systems, such as sonar. Is synthetic aperture imaging a novel human invention? Not quite. While this has not been clearly demonstrated due to the complexity of biological structures, it is believed that the acoustic or ultrasonic sensory systems in dolphins and bats perform some form of synthetic aperture processing. A dolphin or a bat uses lateral motion to create a synthetic aperture. The lateral motion-based synthetic aperture formation and processing is referred to as Doppler processing in the SAR literature. Since synthetic aperture imaging was originally introduced for radar systems, there has been a common misconception that synthetic aperture radar and sonar are exclusively for military purposes. However, synthetic aperture imaging has commer- cial as well as military applications. Some of the commercial and military applica tions of synthetic aperture imaging are: ‘Topographic imaging of the surface of planets * Topographic imaging of the ocean floors © Assessing the condition of crops * Airborne reconnaissance * Underground resources (e.g., oil) exploration © Surveillance and air traffic control * Automatic aircraft landing * Mine detection * Concealed target detection in foliage © Interior imaging of buildings in a rescue or hostage crisis © Detecting buried historical objects * Missile detection and tracking © Ground moving target detection and tracking * Medical ultrasonic imaging of obstructed biological structures © Nondestructive testing of obstructed mechanical structures ‘Some of the above applications, such as topographic imaging, are heavily inves- tigated and even commercialized, Some, such as medical ultrasonic imaging, are behind in development due to misunderstandings and the unconventional nature of synthetic aperture imaging. For instance, one of the areas of interest in diagnostic medicine echo imaging is to obtain quantitative as well as qualitative information re- garding the condition of the cardiovascular system using ultrasonic sources. Due to the low signal to clutter power ratio in these problems, some form of synthetic aper- ture phased array data collection strategy is required [s97] (see also the introduction to Chapter 2).INTRODUCTION — xxi A medical ultrasonic imaging group has investigated and developed a circular synthetic aperture phased array imaging system for this problem.* However, this is achieved by dropping one element (or subaperture) and turning on another in the suc- cessive transmit/receive modes of the array; this would make the acquired data sus- ceptible to various system noise sources (thermal, quantization, nonlinear scattering, etc.). The reconstruction algorithm for this system, which does not exploit wavefront reconstruction theory, is also based on a time-consuming array beamformer which makes the approach susceptible to motion artifacts. Meanwhile, the radar community, where synthetic aperture imaging got its first start, has been more willing than the medical imaging community to accept and adapt modern synthetic aperture imaging systems and their wavefront reconstruction theory-based signal and image processing. This might be due to the fact that the radar engineers are more capable of comprehending a mathematically oriented concept that requires extensive digital signal processing for its implementation. This book is intended to provide a basic understanding of the wavefront re- construction signal theory principles behind SAR imaging, and their digital im- plementation. This is done by not only presenting a theoretical foundation for the multidimensional signals of a SAR system but also providing a physical meaning for these signals. The analysis and wording in the book are for synthetic aperture radar. Yet the general treatment of the synthetic aperture imaging problem, which is applicable in near-range as well as far-range and wideband sources as well as narrowband sources at any carrier frequency, makes the analysis suitable for the syn- thetic aperture-based imaging systems of sonar, diagnostic medicine, geophysical exploration, nondestructive testing, and other applications of this type. Itis anticipated that this approach would remove the myth that SAR signal theory and digital signal processing is a raw and difficult mathematical concept. Synthetic aperture-based imaging systems have the potential to dominate most human-made vision systems, With a better understanding of SAR signal ‘processing, the hidden power and applications of synthetic aperture imaging would be more readily captured and pursued, and put to a much wider use than they have been up to the present time. THE BOOK With the advent of powerful digital signal processing algorithms, both multidimen- sional signal analysis and reconstruction in imaging systems may now be formulated via more concrete theoretical principles and foundations. These theoretical founda- tions, which were deemed to be impractical for implementation in realistic imaging systems, not only can be used to develop accurate and computationally manageable reconstruction algorithms, but they also provide a new understanding of the informa- tion content of the signals that are utilized for image formation. *M. O'Donnell and L. Thomas, “Efficient synthetic aperture imaging from a circular aper- ture with possible application to catheter-based imaging,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, 39: 366-380, 1992. See also papers by the first author in the May 1997 and January 1998 issues of the same journal.xxii INTRODUCTION This is particularly true in SAR imaging. High-resolution SAR offers one of the prime examples of the role of digital signal processing and signal theory in advanc- ing this field. Digital image formation in SAR is heavily dependent on the discrete Fourier analysis of the SAR signal and the target function. This discrete-time and discrete-space analysis imposes restrictions on the parameters used to acquire SAR data, such as the radar pulse repetition frequency (PRR), the fast-time A/D conver- sion rate, and the size of the imaging area for its alias-free implementation. The successful implementation of the advanced SAR multidimensional digital signal processing algorithms requires a thorough understanding of the SAR signal theory and its sampling constraints. The digital signal processing issues associated with the new high-resolution SAR systems have brought new complexities and also new misunderstandings for those familiar with the traditional SAR systems. The purpose of this book is to present a multidimensional signal processing frame- work for high-resolution SAR data acquisition, image formation, and analysis. In addition to the development of the theoretical principles, special emphasis is given to constraints on system parameters, digital signal processing, and digital image re- construction problems of SAR; MATLAB algorithms and numerical examples are provided to assist the reader in this endeavor. The book is intended as a research and teaching monograph for radar signal and image processors, and for individuals interested in applications of multidimensional signal processing. It is written from the point of view of a signal processor. The style is tailored for the members of the signal processing community who are interested in learning about the signal processing theory and algorithms for SAR. The book is also of interest to radar engineers who are eager to understand the principles of emerging advanced multidimensional signal processing algorithms for SAR. ‘As background for this book, a thorough understanding of Fourier analysis and phase modulation is needed. This requirement is certain to be met by signal proces- sors and radar engineers, though someone with a background in wave theory would also be able to follow the development in the text. The analytical development is pre- sented in the framework of signal processing and basic communication theory, and ‘on some occasions, a brief physical interpretation of a topic is also provided. ORGANIZATION ‘Synthetic aperture radar is an echo-mode array imaging system. Similar to the other array imaging systems, SAR provides a multidimensional database (i.e, the acquired data) that can be manipulated via signal processing means to obtain a multidimen- sional image that carries information on the target area under study (i.e., the target re- flectivity function). The signal (information) subspace of the acquired data is formed by varying the radar frequency and radiation pattern in the target area. The simplest way to look at two-dimensional SAR imaging is via the principle that range imaging, the x domain information, is constructed via the variations of the radar signal frequency (bandwidth), and that the cross-range imaging, the y domain information, is formed through the variations of the radar radiation pattern (syntheticINTRODUCTION —Xxili aperture or radar aspect angle). However, the mapping of the two-dimensional signal subspace of the acquired SAR data into the two-dimensional target function is more complicated than two separable one-dimensional mappings. In fact the theory of high-resolution SAR image formation is based on a nonseparable two-dimensional inverse problem [394]. Our discussion in this book begins with the basic principles behind one-dimen- sional range imaging via utilization of a radar bandwidth (Chapter 1) and cross-range imaging via utilization of a synthetic aperture (Chapter 2). These basic principles, and their associated data acquisition and digital signal processing issues, will as- sist us in developing more concrete high-resolution SAR imaging algorithms in the later chapters. In particular, spectral (Fourier) decomposition of propagating waves and the wavefront reconstruction theory, which are the theoretical foundation for for- mulating SAR signal theory and imaging in this book, are presented in Chapter 2. Chapter | also includes a discussion on the design and processing of radar signals for electronic counter countermeasure (ECCM), for example, phase-modulated (phase- perturbed) chirp signals. The discussion in Chapter 3 brings out the special role of the radar radiation pat- tern and its Fourier properties using the wavefront reconstruction theory. These are critical for the reader in comprehending the origin of the SAR signal and the multidi- mensional signal subspace that it occupies. Once this is established, the formulation of the SAR digital signal processing and imaging algorithms is based on an efficient manipulation of this signal subspace. ‘The material in Chapter 4 provides the reader's first exposure to the principle be- hind SAR imaging. In this chapter a SAR scenario is considered, called generic SAR, that lacks many practical components of a realistic SAR system. Yet the generic SAR provides a simple way to view the multidimensional SAR signals and their Fourier properties. The problem of interpolation from unevenly spaced data for digital re- construction of the SAR image is discussed in this chapter. ‘An alternative approach, which is referred to as range stacking, is outlined that does not require any interpolation. The chapter also contains a discussion on two other digital reconstruction methods for SAR which are known as-the time domain correlation (TDC) and backprojection imaging. Modeling and imaging airborne targets using inverse SAR (ISAR) principles are formulated. Motion compensation using global positioning system (GPS) data, and in-scene targets are examined, Application of radar signaling methods of Chapter 1 and pulse diversity for ECCM, and their associated signal processing in SAR, are out- lined. The chapter concludes with a discussion on three-dimensional imaging using a two-dimensional planar synthetic aperture in the azimuth and elevation domains, Chapters 5 and 6 discuss system modeling and imaging for broadside and squint spotlight SAR, and stripmap SAR within the framework of the spectral properties of the SAR radiation pattern that is developed in Chapter 3. This provides the reader with an understanding of the information contents of the two SAR systems, and the limitations of their corresponding acquired data signal subspaces. Chapter 7 is concerned with a SAR geometry in which a circular aperture is formed by the motion of the radar-carrying aircraft. This system is a spotlight SARxxiv INTRODUCTION system which has the capability to provide the imaging scene signature at all aspect angles. The system also provides some form of three-dimensional imaging capabil- ity. A discussion on three-dimensional SAR imaging using a combination of circular and elevation synthetic apertures is also provided. The final chapter of the book, Chapter 8, provides a discussion on a specialized SAR system that uses two or more receiving radars, with varying spatial coordi- nates, to record echoed data due to transmission from a single radar. This system, which is called monopulse SAR, has utility in detecting moving targets in an imaging scene (along-track monopulse SAR), and topographic terrain mapping (slant plane monopulse SAR which is also called interferometric SAR, or IF-SAR). Signal sub- space processing methods for registering uncalibrated SAR images and motion track- ing, and its applications in monopulse SAR systems, are also examined. ‘The signal processing concepts for spotlight SAR in Chapter 5 are accompanied by the real data from a spotlight inverse SAR (ISAR) system; the ISAR data are from an airborne commercial aircraft. The signal processing principles and proper- ties for stripmap SAR in Chapter 6 are accompanied with the real data from an ex- perimental foliage penetrating (FOPEN) ultra wideband UHF stripmap SAR system; this database is also studied in the framework of a spotlight SAR system in Chap- ter 5. The analytical study of circular SAR in Chapter 7 is accompanied by the real turntable data of a T-72 tank. Chapter 8 also uses the real ISAR data to demonstrate registration and tracking applications of the signal subspace processing methods Most chapters of the book (except for Chapters 4 and 7) conclude with a dis- cussion on MATLAB algorithms for the signal and image processing principles that are developed in that chapter; for these programs, output figures are identified by P. ‘Throughout each chapter these program figures are referenced in addition to the fig- ures and examples of that chapter. SAR AND ISAR DATABASES As we mentioned earlier, a set of realistic SAR and ISAR databases are used in this book to assist the reader in visualizing and comprehending the SAR signal theory principles. In this section we identify the origin of these databases, and the parame- ters that were used to acquire them. X Band ISAR Data of a Commercial Aircraft This ISAR database was collected by the Radar Group at the SPAWAR Systems Center (formerly Naval Command, Control and Ocean Surveillance Center) at San Diego. The radar signal was a chirp burst. The following are the parameters of this ISAR database: * Fast-time (A/D converter) sample spacing: A, = 2 ns * Pulse repetition frequency of the radar; PRF = 200 HzINTRODUCTION xxv ° Fast-time frequency bandwidth of the radar: (@min/27, Omax/27] * [9, 9.5] GHz ° Speed of the aircraft (estimated): v, = 92.22 m/s * Slant range and cross-range: (X,, Ye) = (5.23, 2.08) km * Number of fast-time samples: 256 * Number of slow-time samples: 256 * Start point of fast-time sampling: T, = 37.27 us We will see in our discussion on spotlight SAR in Chapter 5 that the above pa- rameters translate into the following: * Cartier frequency of the radar: «a¢/21 = 9.25 GHz * Sample spacing in the synthetic aperture domain: A, = v,/PRF = 0.23 m * Synthetic aperture: [—L, L] = [—59, 59] m * Slow-time data acquisition period: 2L/v, = 1.28 s * Squint angle: 6, = arctan(¥,/X¢) = 21.66° An interesting feature of this ISAR database is that the aircraft possessed nonlin- ear motion components (pitch, yaw, roll, acceleration, etc.) that make the processing and interpretation of the data more complicated. We will show the effects of these nonlinear motion components in the reconstructed ISAR image in Chapter 5. This database will also be examined in the framework of ISAR tracking in Chapter 8. The other interesting feature of this ISAR database is its nonzero squint angle 6, which has an important impact on the point spread function of the ISAR imaging system. UHF Band Stripmap SAR Data This SAR database is part of a set of stripmap SAR databases collected by the radar group at SRI International to study foliage penetrating (FOPEN) characteristics and potential of the UHF band SAR for reconnaissance applications. An impulse-type radar signal was used to radiate the target area. The imaging area was composed of a near-range foliage region and clear land be- hind it. Various stationary trucks and corner reflectors were positioned in the foliage and clear land regions. As we will see, the SAR data also contain contributions from a surrounding farm and vehicles that were in motion in the area (though these were not the intended targets). Also there exists a foliage area at the far-range that shows up in the reconstructed SAR images. We will present results for two sets (i.e., two different runs of the radar-carrying aircraft) of these SAR databases. We will use one of the stripmap SAR databases (aircraft runs) within a relatively small synthetic aperture in Chapter 5 (spotlight SAR) to show its similarity to the spotlight SAR data, In Chapter 6 (stripmap SAR), we will examine the SAR data for the same run of the radar-carrying aircraft within a larger synthetic aperture interval. We will also present results for the SAR data which were collected during the other run of the radar-carrying aircraft.