The spotlight mode SAR (SPSAR) image reconstruction problem may be solved using a backprojection algorithm in the spatial domain while the radar return signals being recorded. The resulting SPSAR algorithm, studied by Desai and Jenkins, permits parallelization of the image reconstruction, and obviates a costly Fourier-domain resampling step. We show that by using a radar signal of a specially designed form with backprojection, the reconstruction becomes extremely simple, eliminating the need for Fourier processing of the return signals, by incorporating a derivative of a Hilbert transform into the emitted signal. We show some simulation results using this new method. We also briefly discuss an algorithm for reconstruction of a function residing on a hemisphere from its radar returns without geometric approximation or distortion.

A synthetic aperture radar (SAR) obtains azimuth resolution by combining data from a number of points along a specified path. Uncompensated antenna motion that deviates significantly from the desired path produces spatially-variant errors in the output image. The algorithm presented in this paper corrects many of these motion-related errors. In this respect, it is similar to time-domain convolution, but it is more computationally efficient. The algorithm uses overlapped subapertures in a three- step image-formation process: course-resolution azimuth processing, fine-resolution range processing, and fine-resolution azimuth processing. Range migration is corrected after the first stage, based on coarse azimuth position. Prior to the final azimuth-compression step, data coordinates are determined to fine resolution in range and coarse resolution in azimuth. This coordinate information is combined with measured motion data to generate a phase correction that removes spatially-variant errors. The algorithm is well-suited for real-time applications, particularly where large flight-path deviations must be tolerated.

Migration processing exactly accounts for the wavefront curvature over the imaged scene. Migration processing is therefore capable of forming high resolution SAR images when the data is acquired over a large synthetic aperture collection angle. Because migration processing requires phase compensation to a line corresponding to the nominal SAR flight path, the phase history is chirped over a very large bandwidth, requiring a very high sample rate to prevent aliasing in the frequency spectrum. The sample rate is determined by the size of the synthetic aperture collection angle. When migration processing is applied to a spotlight-mode SAR, this sampling rate can be much higher than that required for normal spotlight-mode processing. In this latter case, the phase history is motion compensated to scene center and the sample rate is determined by the spot size. Higher sampling rates result in large FFTs and may cause range ambiguity problems. ERIM has pursued the development of a variation on migration processing, which we call the Frequency domain Replication and Downsampling (FReD) algorithm, which enables the acquisition of data at normal spotlight-mode rates and which does not require the computation of FFTs any larger than those required for normal spotlight-mode processing. The FReD algorithm is based on the fact that when a discrete, aliased spectrum is replicated a sufficient number of times, the resultant spectrum will contain the desired signal spectrum. What we call the FReD algorithm is discussed in two articles by Prati, et al. Subsequent processing steps extract the desired portion of the spectrum to form an image. This paper will review migration processing, discuss the FReD algorithm and present expressions for the number of operations required for its implementation. Migration-processed, spotlight-mode SAR imagery derived from airborne collected data and demonstrating the utility of FReD are presented.

This paper outlines the methods adopted by the DRA to obtain focused linear synthetic aperture radar (SAR) images from a long- range, airborne, X-band SAR. We compare the theoretical autofocus performance of four algorithms: fitting the target return, phase estimation, contrast optimisation and multilook registration. Residual phase and azimuthal position errors arising from the focus error are derived and the limitations imposed on the four techniques by the actual scene are discussed. The tracking performances of the algorithms are compared at both high and low frequency and a scheme giving a combination of high-frequency tracking with low distortion introduced. These predictions are then compared with typical aircraft instabilities and the compensated motion frequency response derived. Finally, examples from real SAR data are given illustrating the success of these methods, including the residual distortion and the point response function.

A range-Doppler image is corrupted when each pulse in the corresponding range-time array contains a single time-dependent phase error. An algorithm is developed to estimate these phase errors; the errors are removed, and range-Doppler imaging proceeds in the usual manner. The iterative algorithm works homogeneously across the range-time array (no bright scatterers are necessary for focusing). The estimation technique is model based, but image recovery is successful for images that do not fit the model. The technique promises to be a valuable tool for phase recovery in many cases. These may include image recovery for radar data that: (1) were recorded by a system incapable of maintaining coherence; (2) have been corrupted by uncalibrated sensor or target motion; (3) were collected using frequency- jumped center frequencies; (4) have been corrupted due to the ionosphere; or (5) are bistatic radar data received from a noncooperative transmitter. Illustrations of case (1) are provided, using simulations, postmission corrupted field data, and data that are inherently noncoherent.

Recently, considerable attention has been directed to the use of airborne synthetic-aperture radar (SAR) for detecting both large and small underground objects. The reconstruction of images of buried objects from SAR data, however, often relies on assumptions and algorithms developed in the context of free-space radar propagation. Although in some cases the implicit assumption of free-space propagation can lead to acceptable results, in other important cases it will result in severe degradation to images of underground objects. In this paper we investigate the conditions under which significant improvements to image quality result if proper account is taken of the detailed effects of the soil medium on radar wave propagation. These effects are analyzed both in terms of the refraction of plane waves and via reference to general analytic solutions to Maxwell's equations. Quantitative predictions of the improvement in buried-target image focussing are presented for some realistic data collection scenarios, corresponding to airborne and ground-based SAR or microwave holography.

The 3D information content in data taken from a curved SAR aperture path is significant if sizable areas are produced when the data are projected in three orthogonal directions. Such an undersampled data space does not generate visually acceptable 3D images by straight forward Fourier transformation however. The high sidelobes present in such images may be larger than the main response of other significant scatterers thus veiling their appearance. Also such sidelobes block visual lines of sight to remote scatterers for all but very simple images. A coherent adaptation of the CLEAN algorithm is shown to be effective in eliminating sidelobes in such images allowing the target to be effectively visualized in 3D. Clean images of a target consisting of eight corner reflectors show moderate dependence upon aperture subsets taken from a parent dataset. The algorithm appeared to work well to the limits of the dynamic range of the data set. Initial results indicate the possibility of forming high resolution 3D images of targets from SAR data taken along curvilinear aperture paths in operational situations.

A recent 1D, nonparametric procedure to extrapolate a signal is evaluated for resolution enhancement in inverse SAR imaging. This algorithm, denoted adaptive weighted norm extrapolation (AWNE), is based on iterative use of minimum weighted norm extrapolation to provide a stationary extension of the given data. Computing the discrete Fourier transform (DFT) spectrum of this signal completes the last step of this procedure when used as a spectrum estimator. In this paper the AWNE procedure is used in two directions to obtain range-Doppler images with improved resolution. We evaluate the results and compare them with those obtained with the conventional method based on the DFT. Using rectangularly sampled target data of noise corrupted point scatterers, we show that the AWNE approach produces accurate component locations and amplitudes as well as the expected higher resolution. Our examples show results for arrays of 63 by 63 samples extrapolated to 128 by 128 and 256 by 256. Next, we illustrate the application of AWNE to point scatterer data taken on a polar raster after it has undergone polar reformatting. The procedure is insensitive to the finite accuracy of the reformatting procedure needed to fit the samples into a rectangular raster. Finally, we illustrate the use of AWNE on an array of 64 by 64 data points to form the image of a Boeing 727 aircraft and compare results with those obtained via the conventional DFT processing.

Research at the DRA, Malvern, has resulted in a series of algorithms which are capable of yielding focused, undistorted SAR imagery. Unfortunately these can only be implemented in a fraction of a percent of real-time on a standard work-station. In parallel with the algorithm development, therefore, has been research into a real-time implementation on a parallel computer (the Meiko CS1). This paper outlines the principles behind the software architecture design to achieve the desired speed. Processing functions considered include: initial motion compensation (based on accelerometer data), autofocus with phase correction, final processing and an intensity segmentation stage. Real time processing rates of about 10 MBytes/s are now routinely achieved. We indicate the compromises between processor power, available local memory and communications bandwidth needed to achieve real-time operation. Detailed timings derived from the implementation will be presented together with a discussion of the manner in which this could be varied for different SAR configurations. In parallel with the work on producing real-time high quality imagery has gone a program of research into automated image-understanding techniques. This work is now reaching the stage where reliable algorithms for several basic operations, including segmentation and change detections, exist in a form capable of processing continuous imagery at real time or near real-time rates. Provision has been made for the inclusion of these algorithms as postprocessing stages in the real-time SAR processor.

A highly regular and computationally efficient SAR image formation processor architecture has been developed for wide-area ground surveillance systems. The architecture is based on the repetitive use of a very simple small FFT kernel (instead of a large 2D FFT) providing a variable resolution capability. The first stage of the planar subarray processor (PSAP) forms coarse images and constructs polar formatted, pseudo spotlight SARs. The second stage evaluates the requisite 2D transforms in a multiple- stage block FFT, providing variable resolution images. The fold- FFT is paramount to the computational efficiency of the approach.

Advanced algorithms for synthetic aperture radar (SAR) imaging have in the past required computing capabilities only available from high performance special purpose hardware. Such architectures have tended to have short life cycles with respect to development expense. Current generation Massively Parallel Processors (MPP) are offering high performance capabilities necessary for such applications with both a scalable architecture and a longer projected life cycle. In this paper we explore issues associated with implementation of a SAR imaging algorithm on a mesh configured MPP architecture.

An automated system for the SAR/ISAR imaging of rigid bodies which are undergoing arbitrarily complicated unknown motions is being developed. This system determines, from only the radar data, all observable parameters of motion, on a pulse by pulse basis. The approach makes it possible to: (1) exploit any type of relative motion: translational, rotational, two dimensional, three dimensional, deterministic, or stochastic; no prior parametric assumptions on the functional form of the motion are required; (2) require only the radar data; no ancillary motion measurement system on either the radar platform or on the target is required; (3) automatically provide all the motion information needed to form correctly scaled images, without cross range scale ambiguities; (4) make full use of all the radar data; no signals returning from a target are discarded; and (5) require a known computation time, which is not signal dependent, as all iterative processes used have known, guaranteed convergence rates.

Ultra-wideband radar (UWB) has been shown to be among the most powerful techniques available for underground and obscured object detection. The value of such systems is that they combine the penetration enhancement associated with VHF/UHF (and lower) frequencies with the resolution of wide absolute bandwidth. Such systems necessarily make use of much of the frequency spectrum already in heavy use by other services, such as television and mobile communications. Although this spectral overlap provides occasion for adverse consequences in both directions, to date the principal consequence has been often-severe impact on UWB radar measurements. Even in remote locations, the average interference power often exceeds receiver noise by many dB, becoming the limiting factor on system sensitivity. Nor are UWB radar designers free to overcome this interference by increasing radar power, since regulatory sanction for UWB operation will depend on maintaining sufficiently low spectral power densities to assure that other, prior, services are not appreciably degraded. Given the importance of radio frequency interference (RFI) on practical ultrawide band ground penetrating radar systems, it is important to consider how and to what extent the effects of RFI noise may be reduced. The overall problem of RFI and its impacts will be described and several signal processing approaches to removal of RFI will be discussed. These include spectral estimation and coherent subtraction algorithms and various filter approaches, which have been developed and applied by the signal processing community in other contexts. These methods will be applied to several different real-world experimental data sets, and quantitative measures of the effectiveness of each of these algorithms in removing RFI noise will be presented. Although computationally-intensive, most of the techniques to be described achieve substantial increases in S/RFI without requiring concomitant increases in radar average power.

This paper describes foliage/ground-penetration radar experiments and interferometric synthetic aperture radar experiments sponsored by the Advanced Research Projects Agency in 1992-3. In the foliage/ground penetration experiments, airborne synthetic aperture radar data have been collected at a variety of sites in three bands, 20-90 MHz, 100-500 MHz, and 1200-1300 MHz. Foliage penetration data were collected on Panamanian rain forest and on Northern Maine woods, and ground penetration data were collected on buried objects in a desert site in Arizona. Interferometric SAR data have been collected on various terrain to extract terrain height, and on military vehicles such as tanks and trucks to assess the utility of height information for target detection and recognition.

The effectiveness of applying incoherent change detection to multipass SAR images and targets in foliage is affected by the operating radar frequency band. Incoherent change detection is achieved by taking the weighted difference of the magnitude of two well registered passes of SAR imagery. Items which change between two passes, such as a target present in the first pass and not present in the second pass, will appear in the weighted difference image. With well-registered wideband SAR imagery, images can be divided into frequency bands and evaluated using incoherent change detection. An Environmental Research Institute of Michigan (ERIM) Rail SAR experiment provides such a data collection. The Rail SAR is characterized by polarimetric, wideband (400 MHz - 1.3 GHz), multipass (with and without targets), well-registered SAR images. The ERIM Rail SAR data is divided into a number of frequency bands which simulate the high- band Stanford Research Institute, International (SRI) ultra- wideband radar (UWBR) (350-550 MHz), the Loral miniature SAR (MSAR) (500-800 MHz), and the Naval Air Warfare Center (NAWC) P-3 upgraded UWBR (200-900 MHz) sensor. This paper shows how these sensors work on targets in foliage using incoherent change detection and provides an experimental measurement of upper-bound performance.

A new technique, developed at Lincoln Laboratory, utilizes algorithms developed for two-pass change detection to exploit the differences in aspect angle dependency between target scatterers and clutter scatterers. This technique, referred to as split- aperture detection, involves forming several aspect-angle-diverse SAR images from a single flight pass over a given area by separating the synthetic aperture into sub-apertures during image formation. Change detection algorithms are then applied to these aspect-angle-diverse looks, in the hopes that clutter returns will be nearly isotropic over small variations of aspect angle and will look similar in each image, while man-made objects will provide anisotropic returns over the same angular variation and, therefore, will appear brighter in one image. In this case, the change detection algorithms will significantly suppress the background clutter (thereby significantly reducing the number of false alarms) while enhancing target detectability.

During 1993, a series of experiments were performed under the Advanced Research Projects Agency (ARPA) sponsorship using the SRI ultra-wide band UHF synthetic aperture radar (SAR). These experiments were performed over a variety of clutter backgrounds to assess the foliage penetration capability of the technology and to investigate target detection in clutter. Experiments were conducted observing tropical rain forest backgrounds in Panama, several different desert backgrounds in the Yuma vicinity, and the mid-latitude temperate forest of Maine. SAR images were formed from the raw data using Differential GPS to aid in the focusing. The three locations represent different levels of foliage cover, ranging from the sparsely vegetated desert sites to the triple canopied rain forest. The characteristics of each site are discussed first through a presentation of photography and SAR imagery. The clutter characteristics are studied through a comparison of the cumulative distributions, which are plotted using a variety of conventions. For each case, at least one reference target is included in the test scene. The signal of that target as processed by a common algorithm will be compared to the processed clutter distribution.

An adaptive image formation algorithm is proposed to exploit the aspect-angle dependence of man-made scatterers in foliage penetrating (FOPEN) synthetic aperture radar (SAR). Man-made scatterers often exhibit a strong dependence on the aspect angle between the orientation of the scatterer and the line of sight of the radar. More specifically, the return from a man-made target is greater when the target is oriented broadside with respect to the radar. Conventional SAR image formation processing assumes that backscatter is independent of the aspect angle; by relaxing this assumption, it is possible to reformulate the image formation process to improve the separability of man-made scatterers vs. natural clutter. We propose an image formation process that adapts the length and position of the aperture used during the cross-range compression stage. The algorithm identifies the locations that are likely to correspond to aspect- dependent scatterers. In the vicinity of such scatterers, the algorithm chooses the aperture to match the expected return from a man-made scatterer. Elsewhere, the algorithm uses the full aperture. The resulting imagery enhances man-made targets relative to the background clutter and facilitates improved detection performance.

Most practical automatic target detection/recognition algorithms (ATR) need to incorporate feature mapping and extraction. The performance of the detection algorithms is strongly affected by the effectiveness of feature mapping. For this paper, the effectiveness is evaluated not only by the number of features needed to represent a signal for a given mean-square magnitude error of approximation, but also by the separability of those features from the clutter background. To investigate the effectiveness of a class of linear feature mapping and selection techniques, a comparative study is conducted in this paper using the data collected by MIT Lincoln Laboratory on the SRI Ultra Wideband (UWB) SAR sensor. This data collection was sponsored by ARPA. The feature mappings considered in this paper include conventional image transforms and modern multiresolution decompositions. The traditional image transforms, such as the discrete Fourier transform and the discrete cosine transform provide frequency domain information in the data, while the modern multiresolution decomposition, such as the discrete wavelet transform, are well-known for their capability to provide both spatial and frequency information. In the transform domain, the significant features, i.e., the features with the greatest energy, are extracted by a zonal-filtering with selected masks. The separability of the extracted target significant features from the natural clutter features are compared in terms of the estimated effective GSNR offered by the different feature mappings using the imaged vehicles and the clutter data obtained from the SRI radar. Finally the experimental receiver operating characteristic curves are also presented for the selected feature mappings.

A new SAR target detection technique is developed which we call the joint spatial-polarimetric whitening filter (SPWF) detector. This SPWF detector is an extension of Novak's polarimetric whitening filter (PWF) detector. The derivation of the new SPWF detector presented here is motivated by a new derivation of the PWF developed by Wang and Lee. The SPWF detector is shown to perform better than the PWF detector and single channel detectors using real millimeter-wave fully polarimetric 1 ft x 1 ft spatial resolution SAR data.

This paper contains the first step in an overall approach to low- frequency SAR target detection and classification through direct exploitation of signature physics. The SAR imaging algorithm has been modified to take some of the available signature information into account. More specifically, a simple detection filter is constructed to approximately match the dominant dihedral response from a side of a man-made vehicle, which is not dominant in tree responses. The matching takes place along the image formation integration path in raw dwell data that produces a single image pixel. This integration path measures response versus viewing angle. Since target orientation is unknown, a filter bank is used. Although the dihedral RCS is also matched in frequency, the primary filter gain comes from matching the limited azimuthal response of the dihedral. Further gain may be provided by polarization diversity, but this is unknown since the radar under study measures only HH polarization. In a detection performance comparison with conventional image formation and processing techniques, the current matched filter image formation technique appears to be the only one offering significant improvement. Since the technique is fundamentally limited by the ability to view a vehicle side, it is recommended to increase the radar azimuthal beamwidth to ninety degrees or more. Our overall concept to detection and classification uses a multistep, decision-directed, signature-primitive-based approach that follows the current imaging technique with two layers of complex spatial matched filtering, one for final detection and one for classification, as described at the end.

The potential for automatic target recognition (ATR) processing of foliage-penetrating (FOPEN) SAR imagery requires very high bandwidth occupancies for sufficient range resolution to be achieved for the ATR task. The Army Research Laboratory (ARL) ultra-wideband (UWB) FOPEN SAR--with greater than 95% bandwidth occupancy--provides a suitable testbed for evaluation of UWB ATR techniques. Target signatures in these data are characterized by two temporally distinct responses: an early-time (driven) response, and a late-time (resonant) response. We propose an ATR technique for UWB data based on recognition of targets by their resonant signatures. Current resonance-extraction techniques, such as the singularity-expansion method, hinge on contemporary adaptations of Prony's algorithm; this method, however, generally performs poorly in the presence of noise and is computationally intensive. We propose a form of resonance analysis by application of linear-transform methods, using both classical Fourier techniques and contemporary multiresolution approaches. Target- declaration confidences are established by the correlation of two sets of spectral coefficients--one set from the transformed image data, and the other from a synthetic target template. This permits a fast, efficient scheme for recognition of target resonance effects in UWB imagery. UWB images from the ARL UWB FOPEN SAR instrumentation system were analyzed with canonical targets (dipoles) of differing dimensions and orientation. Results are presented and summarized for each of the targets and transform methods employed in the analysis.

This paper presents performance results for a new, fully polarimetric SAR ATR algorithm suite. The new algorithm suite is a modified version of the current Lincoln Laboratory baseline ATR algorithm suite. The detection (plus clustering) portion of the ATR system remains unchanged; however, the new ATR system uses a larger, optimized feature set for discrimination, and a multi- layered scheme for target classification. The new ATR system was found to provide improved performance over the baseline system when tested on a large data set of tactical targets (both bare and netted) and a large set of ground clutter (which contained numerous man-made objects).

This paper presents a description of a Bayesian neural network (BNN) automatic target recognition (ATR) algorithm specifically developed to efficiently exploit the correlated relationships between the joint dimensional distribution properties of the multiple features. A multidimensional clustering process is used to define a Parzen estimate which approximates the Bayesian a posteriori decision function, but without limiting constraints regarding multimodal or correlated density function forms in multidimensional feature space. The mathematical derivation is placed in the form of a multilayer neural network, including Kernel function variance adaptation to accommodate small training data sets, and Kernel function tail saturation for robust handling of multidimensional feature data not represented in the training data. The BNN is shown to be a hybrid Bayesian/neural net classifier. It has a robust Bayesian statistical classifier as an initial state, and a feedback learning structure which enables learning corrections localized only to the feature space in the vicinity of the error. In addition, a method is described for integrating scene context information, spatial constraints, and a priori probability information into the BNN classifier. Example uses of the BNN algorithm are discussed for recent programs which obtained multiple feature ATR improvements from using the BNN with multiple frequencies, polarizations, aspect angles, and 3D interferometric SAR features.

The need to understand the statistical and spectral characteristics of synthetic aperture radar (SAR) data prompted the analysis of ADTS SAR data. The analysis included statistical and spectral analyses, with an emphasis on understanding the impact of the SAR processing on the scene properties. Several models were developed, including a SAR Processing model and a scene generating model which explicitly models the scene speckle. As a result, a documented and tested FORTRAN program is available for spectral analysis. The statistical and spectral characteristics form the basis of ongoing SAR modeling.

Ultrawideband (UWB) SAR images often contain clutter and other types of spatially correlated noise. To reduce these effects, we have developed a complex, multichannel, autoregressive image- modeling and inverse-filtering algorithm. Our algorithm extends a previously reported technique to include complex image data and a solution to the directionality problem of 2D AR models. A multichannel image pixel, represented by a vector, is modeled as an autoregressive process with a nonsymmetric half-plane (NSHP) support region. The parameters of this NSHP AR model, represented as matrices, are estimated by minimizing a squared residual. To account for possible nonstationarity in the image statistics, the estimation is performed over a series of small subwindows, which lends a degree of adaptabtility to the algorithm. Once estimates of the complex AR parameter matrices are obtained, they are used in an inverse filter. To account for the directionality inherent in the NSHP filter, we have adopted a bi-directional filtering approach, with the results of the two filters combined such that targets are enhanced and clutter is minimized. This algorithm has been applied to various UWB SAR images and shown to be an effective whitening filter. The resulting signal-to-clutter enhancement depends on a variety of factors, including clutter statistics, target size, and AR model order. A version of the whitening algorithm has been incorporated into our standard set of tools for processing UWB SAR images.

This paper demonstrates the importance of texture in SAR image interpretation. Initilly, a simulated annealing technique for despeckling SAR images is outlined. This yield an optimum estimate of the underlying image cross-section. On attempting to apply the method to distinguishing virgin forest from clearings in the Amazon rain forest, we find that the information is not carried in the intensity itself but in the image texture. We discuss the choice of texture estimator that conveys the maximum amount of information and apply the simulated annealing technique to this optimum texture image. We demonstrate that the two categories (forest/clearing) are now almost completely separated with a low misclassification rate. Finally, we examine the sensitivity of the method to the SAR resolution and show that the texture would generally not be visible in an ERS1 image at the conventional incidence angle whereas it could just be with the maximum available incidence angle.

This paper discusses the use of modern 2D spectral estimation algorithms for SAR imaging. We provide a synopsis of the algorithms available, and discuss their relative merits for SAR imaging. Three related algorithms, minimum variance method, reduced-rank minimum variance method, and adaptive sidelobe reduction, offer particular promise for SAR imaging. We review these methods in detail, and develop interferometric variants for use with displaced aperture interferometric SAR systems. Examples illustrate that MVM and ASR both offer significant advantages over Fourier methods for estimating both scattering intensity and interferometric height using data collected of the area around the University of Michigan stadium by ERIM's interferometric X- band DCS SAR.

During an ARPA/Army topographic mapping study, it was shown that interferometric synthetic aperture radar elevation data (IFSARE) can distinguish trees and artificial structures such as buildings and towers from open areas on the basis of height. This result suggests the use of thresholded heights or height-based edge detection as a potential tool for image segmentation in an ATR system. With somewhat more height resolution and accuracy than was available during this study, one may also contemplate height- based change detection as a means of finding mobile targets. This paper describes the radar and navigation systems used for the study. It then summarizes the main algorithm steps used for motion compensation, geopositioning and interferometric phase unwrapping, starting from range compressed data. Sample imagery and related results are shown. Data from preliminary tests of an improved navigation system, intended ultimately to allow automated coregistration of overlapping images, are summarized.

This paper considers target detection with SAR and coherent subtraction. We shall show with experimental data that the coherent subtraction technique may be used to suppress outliers and obtain approximate Gaussian distributions for clutter and noise. We shall also derive generalized likelihood ratio (GLR) detection algorithms that may be used with SAR images that are obtained with coherent subtraction. We shall analytically compare the performance of (a) a single pixel detector, (b) a detector using complete knowledge of the target signature information and known orientation information, (c) a detector using incomplete knowledge of the target signature information and known orientation information, (d) a detector using unknown target signature information and known orientation information, and (e) a detector using unknown target signature information and unknown orientation information.

The signature of a target imaged by a millimeter-wave SAR is highly variable. Various viewing angles will cause different scattering centers to be illuminated, the returns from which can vary greatly with minor changes in viewing angle, and the coherence of the radiation induces speckle noise. Using fully polarimetric turntable (inverse SAR) data, we have undertaken some basic investigations of the persistence of scatterers as a function of azimuth for a number of depression angles from 15 degrees to 32 degrees. Although many scatterers persist for only a few degrees of azimuth, enough persist for 10 to 20 degrees to make model-based recognition feasible. Based on these results, we have developed an experimental system for target recognition. The system uses the functional template approach for detection, pose estimation, and initial hypothesis ranking. The best-matching template defines an area where so-called bright-points are extracted, resulting in a binary feature map that shows the location of strong scatterers. Back-end recognition consists of matching these feature maps to target appearance models that capture the location of scatterers that produce strong returns and are sufficiently persistent with changes in viewing angle. The performance of the hypothesis generation via functional templates is briefly reviewed, both for ISAR data and for SAR data. Recognition results obtained with the new back-end recognition system are also presented for the case of ISAR data.

The subject of interferometric synthetic aperture radar (IFSAR) for high-accuracy terrain elevation mapping continues to gain importance in the arena of radar signal processing. Applications to problems in precision terrain-aided guidance and automatic target recognition, as well as a variety of civil applications, are being studied by a number of researchers. Not unlike many other areas of SAR processing, the subject of IFSAR can, at first glance, appear to be somewhat mysterious. In this paper we show how the mathematics of IFSAR for terrain elevation mapping using a pair of spotlight mode SAR collections can be derived in a very straightforward manner. Here, we employ an approach that relies entirely on Fourier transforms, and utilizes no reference to range equations or Doppler concepts. The result is a simplified explanation of the fundamentals of interferometry, including an easily-seen link between image domain phase difference and terrain elevation height. The derivation builds upon previous work by the authors in which a framework for spotlight mode SAR image formation based on an analogy to 3D computerized axial tomography (CAT) was developed. After outlining the major steps in the mathematics, we show how a computer simulator which utilizes 3D Fourier transforms can be constructed that demonstrates all of the major aspects of IFSAR from spotlight mode collections.

High resolution in synthetic aperture radar (SAR) could be achieved by the convolution of the received signal with the reference function. The right selection of the reference function is necessary to obtain high accuracy in this processing. The reference function of the azimuth processing is characterized by two parameters: the central Doppler frequency and the speed of the azimuth signal frequency modulation. The principles of the proposed algorithms are discussed in this report. These algorithms are intended for more precise estimation of the reference function parameters; they utilize received realization of the reflected signal.

Adaptive high-definition imaging (HDI) is a data-adaptive approach to SAR image reconstruction based on superresolution techniques originally developed for passive sensor arrays. The problem at hand is the detection and recognition of ground-based targets in a clutter-dominated environment via UHF and L-Band foliage-penetrating SAR. Unfortunately, the resolution achieved in conventionally generated images is limited due to longer wavelengths and smaller bandwidths, as compared to high- resolution X- and Ka-Band SAR. A comparison of imaging techniques is presented, including conventional imaging, a 2D technique based on the MLM (Capon) algorithm, and a 2D version of the MUSIC algorithm. Results are presented for Wideband Rail SAR measurements of reflectors both in and out of foliage, demonstrating resolution improvement and clutter rejection. Also, results of processing data from an airborne wideband UHF SAR further demonstrate significant rejection of clutter which promises significant improvements in false-alarm performance.