This PDF file contains the front matter associated with SPIE Proceedings Volume 8051, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

The 2-D Fractional Fourier Transform (FRFT) has been shown to be applicable to the Synthetic Aperture Radar (SAR) imaging problem. Streamlined versions presented here makes the 2-D FRFT comparable with and slightly faster than the Range Doppler (RD) and Extended Chirp Scaling (ECS) methods. The 2-D FRFT is streamlined by eliminating redundancy due to the fact that the same fractional angle is applied to each pulse in the SAR phase history's range dimension while one other fractional angle is applied across each range-gate in the phase history's azimuth dimension. Eliminating the redundancy and approximating the 2-D Fractional Fourier Transform operation in each dimension produces several streamlined 2-D FRFT methods as well as a very fast approximate 2-D FRFT. The computational order of the fast approximate 2-D FRFT is less than that of other corrective SAR imaging techniques. Examples of SAR imaging with these streamlined and approximate FRFTs are given as well as a comparison of the computational speed and impulse response of the full, streamlined and approximate 2-D FRFT, and the RD and ECS methods of SAR imaging.

This paper explores the effect of squint angle on the phase errors introduced by the linear phase assumption in the polar format algorithm for SAR imaging. The maximum scene radius for an allowable phase error is derived as a function of squint angle and other parameters. Simulated phase histories for a variety of squint angles are generated and imaged to demonstrate the bound and the effects encountered when it is exceeded.

The polar format algorithm (PFA) is computationally faster than back projection for producing spotlight mode synthetic aperture radar (SAR). This is very important in applications such as video SAR for persistent surveillance, as images may need to be produced in real time. PFA's speed is largely due to making a planar wavefront assumption and forming the image onto a regular grid of pixels lying in a plane. Unfortunately, both assumptions cause loss of focus in airborne persistent surveillance applications. The planar wavefront assumption causes a loss of focus in the scene for pixels that are far from scene center. The planar grid of image pixels causes loss of the depth of focus for conic flight geometries. In this paper, we present a method to compensate for the loss of depth of focus while warping the image onto a terrain map to produce orthorectified imagery. This technique applies a spatially variant post-filter and resampling to correct the defocus while dewarping the image. This work builds on spatially variant post-filtering techniques previously developed at Sandia National Laboratories in that it incorporates corrections for terrain height and circular flight paths. This approach produces high quality SAR images many times faster than back projection.

It is not currently known if it is possible to accurately form a synthetic aperture radar image from N data points in provable near-linear complexity, where accuracy is defined as the ℓ₂ error between the full O(N²) backprojection image and the approximate image. To bridge this gap, we present a backprojection algorithm with complexity O(log(1/ε)N log N), with ε the tunable pixelwise accuracy. It is based on the butterfly scheme, which works for vastly more general oscillatory integrals than the discrete Fourier transform. Unlike previous methods this algorithm allows the user to directly choose the amount of acceptable image error based on a well-defined metric. Additionally, the algorithm does not invoke the far-field approximation or place restrictions on the antenna flight path, nor does it impose the frequency-independent beampattern approximation required by time-domain backprojection techniques.

This paper considers a time domain ultrasonic tomographic imaging method in a multi-static configuration using the propagation and backpropagation (PBP) method. Under this imaging configuration, ultrasonic excitation signals from the sources probe the object imbedded in the surrounding medium. The scattering signals are recorded by the receivers. Starting from the nonlinear ultrasonic wave propagation equation and using the recorded time domain signals from all the receiver sensors, the object is to be reconstructed. The conventional PBP method is a modified version of the Kaczmarz method that iteratively updates the estimates of the object acoustical potential distribution within the image area. Each source takes turns to excite the acoustical field until all the sources are used. The proposed multi-static image reconstruction method utilizes a significantly reduced number of sources that are simultaneously excited. We consider two imaging scenarios with regard to source positions. In the first scenario, sources are uniformly positioned on the perimeter of the imaging area. In the second scenario, sources are randomly positioned. By numerical experiments we demonstrate that the proposed multi-static tomographic imaging method using the multiple source excitation schemes results in fast reconstruction and achieves high resolution imaging quality.

We present a technique for aperture weighting for use in video synthetic aperture radar (SAR). In video SAR the aperture required to achieve the desired cross range resolution typically exceeds the frame rate period. As a result, there can be a significant overlap in the collected phase history used to form consecutive images in the video. Video SAR algorithms seek to exploit this overlap to avoid unnecessary duplication of processing. When no aperture weighting or windowing is used one can simply form oversampled SAR images from the non-overlapping sub-apertures using coherent back projection (or other similar techniques). The resulting sub-aperture images may be coherently summed to produce a full resolution image. A simple approach to windowing for sidelobe control is to weight the sub-apertures during summation of the images. Our approach involves producing two or more weighted images for each sub-aperture which can be linearly combined to approximate any desired aperture weighting. In this method we achieve nearly the same sidelobe control as weighting the phase history data and forming a new image for each frame without losing the computation savings of the sub-aperture image combining approach.

In this paper, an inversion scheme for near-field inverse synthetic aperture radar (ISAR) data is derived for both two and three dimensions from a scalar wave equation model. The proposed data inversion scheme motivates the use of a filtered back projection (FBP) imaging algorithm. The paper provides a derivation of the the general imaging filter needed for FBP, which will be shown to reduce to a familiar result for near-field ISAR imaging.

We utilize eight circular passes of measured airborne X-Band radar data to form a novel reflectivity surface estimate. Our previous work demonstrated reflectivity surface estimation from narrow aperture multi-pass data and this work extends those results to wide apertures. Narrow aperture surface estimates are co-registered in the three spatial dimensions and combined non-coherently to form a wide-aperture data product. For the purpose of visually conveying the scene reflectivity content a surface is formed from the wide angle data product. The subject of this work is not on the optimality of the methods nor the global convexity of the cost functions. Instead, these results give us one of the first glimpses at measured wide angle three dimensional SAR image products and provide a qualitative benchmark against which to measure future wide angle three dimensional synthetic aperture radar autofocus and imaging algorithms.

We consider a monostatic synthetic aperture radar system traversing an arbitrary trajectory on a non-flat topography. We present a novel edge detection method applicable directly to SAR received signal. Our method first filters the received data, and then backprojects. The filter is designed to detect the edges of the scene in different directions at each pixel reconstructed. The method is computationally efficient and may be implemented with the computational complexity of the fast-backprojection algorithms. We present numerical experiments to demonstrate the performance of our method.

This paper addresses the question of scattering center detection and estimation performance in synthetic aperture radar. Specifically, we consider sparse 3D radar apertures, in which the radar collects both azimuth and elevation diverse data of a scene, but collects only a sparse subset of the traditional filled aperture. We use a sparse reconstruction algorithm to both detect and estimate scattering center locations and amplitudes in the scene. We quantify both the detection and estimation performance for scattering centers over a high dynamic range of magnitudes. Over this wide range of scattering center signal-to-noise values, detection performance is compared to GLRT detection performance, and estimation performance is compared to the Cramer-Rao lower bound.

For large-scale linear inverse problems, a direct matrix-vector multiplication may not be computationally feasible, rendering many gradient-based iterative algorithms impractical. For applications where data collection can be modeled by Fourier encoding, the resulting Gram matrix possesses a block Toeplitz structure. This special structure can be exploited to replace matrix-vector multiplication with FFTs. In this paper, we identify some of the important applications which can benefit from the block Toeplitz structure of the Gram matrix. Also, for illustration, we have applied this idea to reconstruct 2D simulated images from undersampled non-Cartesian Fourier encoding data using three popular optimization routines, namely, FISTA, SpaRSA, and optimization transfer.

We consider synthetic aperture radar system using ultra-narrowband continuous waveforms, which we refer to as Doppler Synthetic Aperture Radar (DSAR). We present a novel image formation method for bi-static DSAR. Our method first correlates the received signal with a scaled or frequency-shifted version of the transmitted signal over a finite time window, and then uses microlocal analysis to reconstruct the scene by a filtered-backprojection of the correlated signals. Our approach can be used under non-ideal imaging scenarios such as arbitrary flight trajectories and non-flat topography. Furthermore, it is an analytic reconstruction technique which can be made computationally efficient. We present numerical experiments to demonstrate the performance of the proposed method.

We propose a single-receive, multiple-transmit channel imaging radar system that limits received data rate while also providing spatial processing for improved detection of moving targets. A multi-input, single-output (MISO) system uses orthogonal waveforms to separate spatial channels at the single receiver. The use of orthogonal waveforms necessitates several modications to both synthetic aperture radar imaging and adaptive space-time beamforming. An orthogonal frequency-division transmit waveform scheme is proposed, and we derive the attendant extensions to the standard backprojection and space-time beamforming algorithms.. We demonstrate imaging and moving target detection results using data from an airborne X-band system. We conclude with a discussion of the clutter covariance matrix of the resulting space-time beamformer and a suggested waveform scheduling scheme to minimize the rank of the observed clutter subspace.

The paper presents results from a bistatic SAR experiment conducted using two airborne SAR systems operating in the high VHF- and low UHF-band. The Swedish SAR system LORA operated together with the French SAR system SETHI and collected data in different bistatic geometries in the frequency band 222-460 MHz and using HH-polarization. The two SAR systems were synchronized using the 1PPS GPS-signal. Data were collected during four flight missions over the main test site with forested terrain and buildings as well as controlled target deployments. A fifth mission was included over a second test site with an extensive data base of forest parameters but without target deployments. The bistatic radar data have been processed to SAR images and first analysis completed. Results show significant suppression of strong forest clutter and that the effect increases with bistatic elevation angle. The clutter reduction is observed in areas with dominating double-bounce scattering. Data analysis shows that forest clutter can be suppressed by 10 dB for a bistatic elevation angle of 10°.

Synthetic Aperture Radar (SAR) often suffers from interference signal from various radio sources. In general, notch filters or band elimination filters have been utilized to eliminate such interference signal; however, if the bandwidth of the interference signal is relatively wide, the gap in the spectrum caused by the band elimination filter could significantly degrade the original image. We propose an algorithm to suppress relatively wide bandwidth interference while maintaining the image quality. In the algorithm, spatially variant interference suppression filter is generated based on the signal of interference free band, then the filter is applied to the interference contaminated image. Unlike the band elimination filter, the spatially variant interference suppression filter preserves the signal component within the interference contaminated band; therefore, the image distortion caused by the spectrum gap can be largely eliminated. The algorithm has been tested with a simulated interference contaminated image generated from the real 10cm resolution airborne Ku band SAR image and a TerraSAR-X image. It has been shown that while conventional band elimination filter would degrade the image quality, the image quality of the interference suppressed image obtained by the proposed algorithm is satisfactory.

We consider nonparametric adaptive spectral analysis of complex-valued data sequences with missing samples occurring in arbitrary patterns. We first present two high-resolution missing-data spectral estimation algorithms: the Iterative Adaptive Approach (IAA) and the Sparse Learning via Iterative Minimization (SLIM) method. Both algorithms can significantly improve the spectral estimation performance, including enhanced resolution and reduced sidelobe levels. Moreover, we consider fast implementations of these algorithms using the Conjugate Gradient (CG) technique and the Gohberg-Semencul-type (GS) formula. Our proposed implementations fully exploit the structure of the steering matrices and maximize the usage of the Fast Fourier Transform (FFT), resulting in much lower computational complexities as well as much reduced memory requirements. The effectiveness of the adaptive spectral estimation algorithms is demonstrated via several 2-D interrupted synthetic aperture radar (SAR) imaging examples.

Multichannel Autofocus (MCA) assumes that there exits a region of low return in the focused image and solves for the correction filter that minimizes the energy in the presumed low-return region. Provided that the lowreturn region is precisely known, the algorithm yields a superior restoration compared to other autofocus methods. Fourier-domain MCA (FMCA) is a generalization of this algorithm that works for practical ranges of look angles. However, both MCA and FMCA assume a planar wavefront, which makes them inapplicable to near-field imaging scenarios where there is a significant amount of wavefront curvature. We propose an autofocus algorithm that builds upon MCA, with a modification that takes into account wavefront curvature. In this setting, the demodulated data can no longer be interpreted as 2-D Fourier samples of the underlying image. Therefore, we make use of the linear relationship between the correction filter and the reconstructed image via Convolution Backprojection (CBP) along curves. Under the far-field assumption, our algorithm is equivalent to FMCA with a Jacobian-weighted 2-D periodic sinc-kernel interpolator when the presumed low-return regions are the same. However, our algorithm has the distinct advantage of being able to select the presumed low-return region within a continuous set of coordinates. We present simulation results showing that our algorithm outperforms other algorithms for the case with wavefront curvature.

A method for obtaining windowing functions for focused range-doppler imaging of rotation objects is given. Focused range-doppler imaging of rotation objects involves the evaluation of a non-standard transform rather than the discrete Fourier transform. As such, familiar windowing functions such as Hamming are inappropriate for sidelobe control. Previous methods of evaluation involve resampling and interpolation to emulate the discrete Fourier transform. In this paper, a correction factor applied to any standard window function is derived and results shown.

General purpose graphical processing units or GPGPUs have emerged in recent years as the power horse behind many large scale computing efforts. For example, the recent unveiling of the world's fastest supercomputer has achieved this feat by utilizing low cost and high performance GPGPUs. Additionally, in the past year the synthetic aperture radar (SAR) community has started to utilize GPGPUs as well. The utilization of GPGPUs to date has been limited mainly to SAR image formation and in this capacity tremendous performance improvements over the same CPU based algorithms have been demonstrated. However, image formation is only one of many necessary steps towards SAR image exploitation. Image registration, filtering, interpolation and interferometric flattening are equally important steps in obtaining many of the desired output products such as coherence change detection (CCD) products and terrain adjusted interferograms. We will demonstrate that by transitioning the entire SAR image exploitation processing chain from image formation through product generation onto a GPGPU, it is possible to achieve more than an order of magnitude in performance improvements. In this paper we will review results presented at last year's SPIE conference regarding SAR image formation and present new results obtained for coherent exploitation of SAR data including CCD and interferometric SAR processing. In addition to presenting these results, we will discuss challenges associated with migration of CPU-based exploitation algorithms to the GPGPU environment, as well as to discuss possible future improvements using these powerful new devices and associated software tools.

This paper describes several alternative techniques for detecting and localizing slowly-moving targets in cultural clutter using synthetic aperture radar (SAR) data. Here, single-pass data is jointly processed from two or more receive channels which are spatially offset in the along-track direction. We concentrate on two clutter cancelation methods known as the displaced phase center antenna (DPCA) technique and along-track SAR interferometry (AT-InSAR). Unlike the commonly-used space-time adaptive processing (STAP) techniques, both DPCA and AT-InSAR tend to perform well in the presence of non-homogeneous urban or mountainous clutter. We show, mathematically, the striking similarities between DPCA and AT-InSAR. Furthermore, we demonstrate using experimental SAR data that these two techniques yield complementary information, which can be combined into a "hybrid" technique that incorporates the advantages of each for significantly better performance. Results are generated using the Gotcha challenge data, acquired using a three-channel X-band spotlight SAR system.

We consider moving target detection and velocity estimation for multi-channel synthetic aperture radar (SAR) based ground moving target indication (GMTI). Via forming velocity versus cross-range images, we show that small moving targets can be detected even in the presence of strong stationary ground clutter. Furthermore, the velocities of the moving targets can be estimated, and the misplaced moving targets can be placed back to their original locations based on the estimated velocities. An iterative adaptive approach (IAA), which is robust and user parameter free, is used to form velocity versus cross-range images for each range bin of interest. Moreover, we discuss calibration techniques to combat near-field coupling problems encountered in practical systems. Furthermore, we present a sparse signal recovery approach for stationary clutter cancellation. We conclude by demonstrating the effectiveness of our approaches by using the Air Force Research Laboratory (AFRL) publicly-released Gotcha airborne SAR based GMTI data set.

This paper develops a hierarchical Bayes model for multiple-pass, multiple antenna synthetic aperture radar (SAR) systems with the goal of adaptive change detection. The model is based on decomposing the observed data into a low-rank component and a sparse component, similar to Robust Principal Component Analysis, previously developed by Ding, He, and Carin1 for E/O systems. The developed model also accounts for SAR phenomenology, including antenna and spatial dependencies, speckle and specular noise, and stationary clutter. Monte Carlo methods are used to estimate the posterior distribution of the variables in the model. The performance of the proposed method is analyzed using synthetic images, and it is shown that the performance is robust to a large space of operating characteristics without extensive tuning of hyperparameters. Finally, the method is applied to measured SAR data, providing competitive results compared to standard methods with the additional benefits of uncertainty characterization through a posterior distribution, explicit estimates of both foreground and background components, and flexibility in including other sources of information.

This paper investigates the performance of single-channel SAR-GMTI systems in the focusing and detection of translating ground targets moving in the presence of a clutter background. Specifically, focusing and detection performance is investigated by applying the Moving Grid Processing (MGP) focusing technique to a scene containing an accelerating target moving in the presence of both uniform and correlated K-distributed clutter backgrounds. The increase in detection sensitivity resulting from the focusing operation is found to result from two separable effects, target focusing and clutter defocusing. While the detection sensitivity gain due to target focusing is common for both clutter types, the gain due to clutter defocusing is found to be significantly greater for textured clutter than for uniform clutter, by approximately 5 to 6 dB in the simulated scenario under consideration. This paper concludes with a discussion of the phenomenological causes for this difference and implications of this finding for single channel SAR-GMTI systems operating in heterogeneous clutter environments.

This paper develops the theory for waveform-diverse moving-target synthetic-aperture radar. We assume that the targets are moving linearly, but we allow an arbitrary flight path and (almost) arbitrary waveforms. We consider the monostatic case, in which a single antenna phase center is used for both transmitting and receiving. This work addresses the use of waveforms whose duration is sufficiently long that the targets and/or platform move appreciably while the data is being collected.

We present a novel passive image formation method for moving targets using distributed apertures capable of exploiting information about multiple-scattering in the environment. We assume that the environment is illuminated by non-cooperative transmitters of opportunity with unknown location and unknown transmitted waveforms. We develop a passive measurement model that relates the scattered field from moving targets at a given receiver to the scattered field at other receivers. We formulate the passive imaging problem as a generalized likelihood ratio test for a hypothetical target located at an unknown position, moving with an unknown velocity. We design a linear discriminant functional by maximizing the Signal-to-Noise Ratio (SNR) of the test-statistic, and use the resulting position- and velocity-resolved test-statistic to form the image. Our imaging method can determine the two- or three-dimensional velocity vector as well as the two- or three-dimensional position vector of a moving target without the knowledge of transmitter locations and transmitted waveforms. We present numerical experiments to demonstrate the performance of our passive imaging method operating in multiplescattering environments. The results show that the point spread function of the reconstructed images improves when the information about multiple scattering is exploited.

Measurement times for synthetic aperture radar (SAR) image collection can take from the order of seconds to minutes and consequently the technique is subject to imaging artefacts due to target motion. For example, imaged moving targets can be displaced and unfocussed and similarly for vibrating targets. Current understanding of this phenomenon is somewhat esoteric however this paper puts forward and demonstrates a visual explanation via the physics of modulated scatterer SAR images in the Fourier domain. This novel approach has led to an imagery analyst aid which associates a distinctive signature to modulated scatterer artefacts in SAR imagery and to an associated filter.

We present a low-complexity method for compression of raw Synthetic Aperture Radar (SAR) data. Raw SAR data is typically acquired using a satellite or airborne platform without sufficient computational capabilities to process the data and generate a SAR image on-board. Hence, the raw data needs to be compressed and transmitted to the ground station, where SAR image formation can be carried out. To perform low-complexity compression, our method uses 1-dimensional transforms, followed by quantization and entropy coding. In contrast to previous approaches, which send uncompressed or Huffman-coded bits, we achieve more efficient entropy coding using an arithmetic coder that responds to a continuously updated probability distribution. We present experimental results on compression of raw Ku-SAR data. In those we evaluate the effect of the length of the transform on compression performance and demonstrate the advantages of the proposed framework over a state-of-the-art low complexity scheme called Block Adaptive Quantization (BAQ).

Being able to recognize one object from another is vital research to our society because it can save lives, improve national security, and improve existing technology such as object avoidance, tracking, etc. In this research we are trying to classify Synthetic Aperture Radar (SAR) images of vehicles from one another no matter if the vehicle is rotated or occluded. The dataset that is being used for this research is the Commercial Vehicle (CV) Data Domes obtained fromWright Patterson Air Force Base (WPAFB). To accomplish this task we used Local Feature Extraction (LFE) to extract the features and then K-nearest neighbor (KNN) was used to classify the vehicles. Overall this method performed well in that the algorithm was able to correctly identify the vehicles 97.6% to 100% accuracy. Currently the algorithm can not handle translation, so the next step of this research is to be able to use the glint information to register the vehicles to a desired location and then perform our algorithm which we believe that registering the image would have a significant improvement to the current results.

Phase fluctuation is one of the inherent characteristics for complex radar targets.The primary objective of this work is to compare the phase fluctuation characteristics of ships on sea surface with that of the sea clutter based on complex high resolution range profiles (HRRPs). The statistics of the HRRP phase gradient are studied using alpha-stable distribution. Numerical simulation results show that the HRRP phase gradient of a ship on sea surface behave significantly different from that of the sea clutter, suggesting that the statistics of HRRP phase gradient provide useful information for ship discrimination from sea clutter.

A method of Coherent Change Detection (CCD) performance prediction is described based on (1) a simple analysis of the frequency support overlap between the two collects which encapsulates imaging geometry effects and (2) a method of relating environmental effects to average scatterer disturbance which can be assessed empirically. The strength of this approach is that, once the average disturbance for a particular environmental effect has been established from one system, it can be extrapolated to all other systems since mean disturbance is system-independent. Validation and application of this approach using simulated examples is presented.

We present an image quality metric and prediction model for SAR imagery that addresses automated information extraction and exploitation by imagery analysts. This effort drarws on our team's direct experience with the development of the Radar National Imagery Interpretability Ratings Scale (Radar NIIRS), the General Image Quality Equations (GIQE) for other modalities, and extensive expertise in ATR characterization and performance modeling. In this study, we produced two separate GIQEs: one to predict Radar NIIRS and one to predict Automated Target Detection (ATD) performance. The Radar NIIRS GIQE is most significantly influenced by resolution, depression angle, and depression angle squared. The inclusion of several image metrics was shown to improve performance. Our development of an ATD GIQE showed that resolution and clutter characteristics (e.g., clear, forested, urban) are the dominant explanatory variables. As was the case with NIIRS GIQE, inclusion of image metrics again increased performance, but the improvement was significantly more pronounced. Analysis also showed that a strong relationship exists between ATD and Radar NIIRS, as indicated by a correlation coefficient of 0.69; however, this correlation is not strong enough that we would recommend a single GIQE be used for both ATD and NIIRS prediction.

The target classification algorithm community is making a special effort to explicitly treat operating conditions (OCs) in classifier assessments and performance modeling. This is necessary because humans do not intuitively appreciate what makes classification difficult for computers-it just seems so easy to us. In analyzing OCs, some OCs are more direct or primitive while others are more abstract or integrating. These more abstract or "Derived OCs" provide an intermediate step between direct OCs and classifier performance. Similar to the target, sensor, environment partition of OCs, the AFRL COMPASE Center introduces the "Mossing 3" partition of derived OCs into "Clarity," "Uniqueness," and "Conformity." Clarity is primarily concerned with the relevant information content available in the sensor data. Uniqueness is about the inherent separability between the types of objects to be classified (i.e., the library) and between all those types and objects not known to the classifier. Conformity is about the relationship between the OCs of the test instances and the OCs represented in the library types or training data. Furthermore, by analyzing derived OCs from multiple perspectives, informative subpartitions of the Mossing 3 are created. Clarity measures are well developed, particularly as image quality metrics. The other partitions are less well developed, but relevant work exists and is brought into context. While derived OCs and the Mossing 3 partition are not a complete solution to performance modeling, they help bring in powerful existing technologies and should enrich and facilitate dialogue on classifier performance theory and modeling.

In this paper, we introduce a novel joint sparse representation based automatic target recognition (ATR) method using multiple views, which can not only handle multi-view ATR without knowing the pose but also has the advantage of exploiting the correlations among the multiple views for a single joint recognition decision. We cast the problem as a multi-variate regression model and recover the sparse representations for the multiple views simultaneously. The recognition is accomplished via classifying the target to the class which gives the minimum total reconstruction error accumulated across all the views. Extensive experiments have been carried out on Moving and Stationary Target Acquisition and Recognition (MSTAR) public database to evaluate the proposed method compared with several state-of-the-art methods such as linear Support Vector Machine (SVM), kernel SVM as well as a sparse representation based classifier. Experimental results demonstrate that the effectiveness as well as robustness of the proposed joint sparse representation ATR method.

Conventional target recognition approaches for SAR include template matching and feature-based classification. However, unlike visual imagery, Synthetic Aperture Radar (SAR) presents a unique challenge in that many attributes, such as scattering centers, are extremely pose dependent and wink in and out with even minor viewing geometry changes. This work implements a highly efficient biologically-inspired 3D template-based approach, the Map Seeking Circuit (MSC) algorithm, for target recognition in SAR. Instead of exhaustively searching a high dimensional state space, the MSC algorithm efficiently searches a superposition hypersurface to estimate target location and 3D pose. Results are shown from applying the algorithm to real SAR datasets.

Morphological operators are commonly used in image processing. We study their suitability for use in synthetic aperture radar (SAR) image enhancement and target classification. Morphological operations are nonlinear operators defined by set theory. The dilation and erosion operations grow or shrinkimage features that match to a predefined structuring element. The opening and closing operations are combinations of successive dilation and erosion. These morphological operations can visually emphasize scattering of interest in an image. We investigate whether these operations can also improve target classification performance. The operators are nonlinear and image dependent; thus we cannot predict performance without empirical testing. We test and evaluate the morphological operators using simulated and measured SAR data. Results show the dilation operator is most promising for increasing match score and separation between classes in the decision space.

The automatic target recognition (ATR) performance of SAR with subsampled raw data is investigated in this paper. Two schemes are investigated. In scheme A, SAR images are reconstructed from subsampled data by applying compressed sensing (CS) techniques and then targets are classified using either the mean-squared error (MSE) classifier or the point-feature-based classifier. Both classifiers recognize a target by using the magnitude information of dominant scatterers in the image. They fit nicely with the CS framework considering that CS approaches can efficiently recover the bright pixels in SAR images. In scheme B, the smashed-filter classifier is employed without image formation. Instead it makes the classification decision by directly comparing the observed subsampled data with data simulated from reference images. The impact of various subsampling patterns on ATR is investigated since CS theory suggests that some patterns lead to better performance than others. Simulation results show that compared with images formed by the conventional SAR imaging algorithm, CS reconstructed images always lead to much higher recognition rates for both the classifiers in scheme A. The MSE classifier works better than the point-feature-based classifier because the former takes into account both the magnitudes and locations of bright pixels while the latter uses the locations only. The smashed-filter classifier is computationally efficient and can accurately recognize a target even with strong subsampling if appropriate reference images are available. Its application in practice is difficult because it is sensitive to the phases of complex-valued SAR images, which vary too much for different observation angles.