This paper considers the problem of target delineation in SAR images. A novel technique involving multiple active contours evolving simultaneously is developed and tested using real radar data. In particular, a technique referred to as multiple hypothesis delineation, in which contours can be in several states simultaneously, is developed and shown to lead to considerable improvement in convergence time and delineation accuracy. The technique is applied to two different SAR image applications, the delineation of multiple targets in close proximity, and the automatic estimation of building dimensions by delineation of shadows in a sequence of SAR images of an urban scene. Finally, the incorporation of the delineation algorithms in a software tool enabling rapid interactive building dimension estimation is briefly described.

The improved ground resolution of state-of-the-art synthetic aperture radar (SAR) sensors suggests utilizing this technique for analysis of urban areas. However, building reconstruction from SAR or InSAR data suffers from consequences of the inherent oblique scene illumination, such as foreshortening, layover, occlusion by radar shadow and multipath signal propagation. Especially in built-up areas, building reconstruction is often hardly possible based on single SAR or InSAR data sets alone. An approach is presented to improve the reconstruction quality combining multiaspect InSAR data. Building object primitives are extracted independently for two directions from the magnitude and phase information of the interferometric data. After projection of these initial primitive objects from slant range into the world coordinate system they are fused. This set of primitive objects is used to generate building hypotheses. SAR illumination effects are discussed using real and simulated data. The simulation results have been compared with real imagery. Deviations between simulations and real data were the base for further investigations. The approach is demonstrated for two InSAR data sets of a building group in an urban environment, which have been taken from orthogonal viewing directions with spatial resolution of about 30 cm.

It has recently become apparent that dismount tracking from non-EO based sources will have a large positive impact on urban operations. EO / camera imaging is subject to line of site and weather conditions which makes it a non-robust source for dismount tracking. Other sensors exist (e.g. radar) to track dismount targets; however, little radar dismount data exists. This paper examines the capability to generate synthetic and measured dismount data sets for radar frequency (RF) processing. For synthetic data, we used the Poser^TM program to generate 500 facet models of human dismount walking. Then we used these facet models with Xpatch to generate synthetic wideband radar data. For measured dismount data, we used a multimode (X-Band and Ku-Band) radar system to collect RF data of volunteer human (dismount) targets.

A set of approximate theoretical equations for the Doppler response of monostatic radar signals due to slowly pivoting objects are derived. The treatment is based on physical models extracted from the mechanical engineering community. Potential applications include analysis of load-based vehicle classification and detection of biological movements such as human joint rotations. Several example calculations are presented based on the resulting theoretical formulas. These examples include Doppler calculations for notional first-order vehicle suspension models and first-order human joint (arm/leg) rotation models. Each set of example calculations includes two sets of notional radar parameters in order to provide insight into potential Doppler pivot detection capabilities as a function of basic radar parameters such as frequency and PRF (pulse repetition frequency).

The electromagnetic scattered field from an electrically large target can often be well modeled as if it is emanating from a discrete set of scattering centers (see Fig. 1). In the scattering center extraction tool we developed previously based on the shooting and bouncing ray technique, no correspondence is maintained amongst the 3D scattering center extracted at adjacent angles. In this paper we present a multi-dimensional clustering algorithm to track the angular and spatial behaviors of 3D scattering centers and group them into features. The extracted features for the Slicy and backhoe targets are presented. We also describe two metrics for measuring the angular persistence and spatial mobility of the 3D scattering centers that make up these features in order to gather insights into target physics and feature stability. We find that features that are most persistent are also the most mobile and discuss implications for optimal SAR imaging.

This paper explores three-dimensional (3D) interferometric synthetic aperture radar (IFSAR) image reconstruction when multiple scattering centers and noise are present in a radar resolution cell. We introduce an IFSAR scattering model that accounts for both multiple scattering centers and noise. The problem of 3D image reconstruction is then posed as a multiple hypothesis detection and estimation problem; resolution cells containing a single scattering center are detected and the 3D location of these cells' pixels are estimated; all other pixels are rejected from the image. Detection and estimation statistics are derived using the multiple scattering center IFSAR model. A 3D image reconstruction algorithm using these statistics is then presented, and its performance is evaluated for a 3D reconstruction of a backhoe from noisy IFSAR data.

We present a feature extraction algorithm to detect scattering centers in three dimensions using monostatic synthetic aperture radar imagery. We develop attributed scattering center models that describe the radar response of canonical shapes. We employ these models to characterize a complex target geometry as a superposition of simpler, low-dimensional structures. Such a characterization provides a means for target visualization. Fitting an attributed scattering model to sensed radar data is comprised of two problems: detection and estimation. The detection problem is to find canonical targets in clutter. The estimation problem then fits the detected canonical shape model with parameters, such as size and orientation, that correspond to the measured target response. We present an algorithm to detect canonical scattering structures amidst clutter and to estimate the corresponding model parameters. We employ full-polarimetric imagery to accurately classify canonical shapes. Interformetric processing allows us to estimate scattering center locations in three-dimensions. We apply the algorithm to scattering prediction data of a simple scene comprised of canonical scatterers and to scattering predictions of a backhoe.

In recent years, UHF synthetic aperture radar has become a growing area of interest among the radar community. Due to their relatively long wavelengths, UHF systems provide advantages that may not be attainable by microwave and millimeter-wave radar systems. These advantages include excellent target detection statistics in high clutter environments, wide-area surveillance, and long stand-off ranges. UHF systems also have proven synergistic properties with higher frequency radar systems in applications such as topographical mapping. However, the ability to study the characteristics of these lower frequency radar systems in a controlled and systematic environment is difficult. In this work, a physical scale modeling process is utilized to generate three-dimensional UHF imagery that may be used to study scattering phenomenology at these wavelengths. Dimensionally and dielectrically scaled targets and scenes are measured in a 6 - 18 GHz microwave compact range to model the backscatter of the full-size target at UHF wavelengths. The microwave compact radar range and transceiver hardware utilized to model UHF radar signature data are briefly described. A description of the image processor used to generate three-dimensional UHF imagery from wide-band/wide-angle data collections is described as well. Finally, imagery of radar signature data collected from a M1A1 Abrams main battle tank model is examined. The high resolution imagery resulting from the wide-band/wide-angle collection will show that sub-wavelength features of ground targets are resolvable at these wavelengths.

Synthetic Aperture Radar (SAR) is capable of producing high-resolution terrain images from data collected by a relatively small airborne or spaceborne antenna. This data collection is done in cross-range or slow-time along flight trajectory and range or fast-time along direction of electromagnetic wave propagation. The slow-time imaging is what distinguishes SAR from its predecessor imaging radars. The high resolution pulse compression based fast-time imaging in range introduces some visual artifacts into SAR imagery due to range skew and phase information anomaly due to residual video phase (RVP). In this paper, we introduce the concept of SAR 2D aperture synthesis that extends the slow-time imaging concept to range and relies on a single frequency instead of chirp. Moreover, our 2D aperture synthesis implementation does not need computationally expensive Stolt interpolation.

Emerging wide-area synthetic aperture radar (SAR) system concepts call for a single data collection platform to orbit a large (e.g, 20 km) spot at a nominal range of 40 km. The large standoff distance and desire for fine resolution, coupled with a need for persistent real-time sensing, pose a significant challenge in terms of clutter-to-noise ratio (CNR) performance and data processing. Increased CNR and reduced processing load can be achieved by decreasing the range of the SAR system and the size of the area of interest. Employing multiple cooperating SAR systems allows the same overall coverage area to be maintained with a patchwork of SAR footprints. This paper analyzes a high-level system architecture, for multiple SAR systems, that provides uninterrupted coverage over a wide area. System analysis includes eclipsing diagrams, CNR performance and mutual interference issues.

We consider the problem of jointly forming images and characterizing anisotropy from wide-angle synthetic aperture radar (SAR) measurements. Conventional SAR image formation techniques assume isotropic scattering, which is not valid with wide-angle apertures. We present a method based on a sparse representation of aspect-dependent scattering with an overcomplete basis composed of basis vectors with varying levels of angular persistence. Solved as an inverse problem, the result is a complex-valued, aspect-dependent response for each spatial location in a scene. Our non-parametric approach does not suffer from reduced cross-range resolution inherent in subaperture methods and considers all point scatterers in a scene jointly. The choice of the overcomplete basis set incorporates prior knowledge of aspect-dependent scattering, but the method is flexible enough to admit solutions that may not match a family of parametric functions. We enforce sparsity through regularization based on the ℓ_k-norm, k < 1. This formulation leads to an optimization problem that is solved through a robust quasi-Newton method. We also develop a graph-structured interpretation of the overcomplete basis leading towards approximate algorithms using guided depth-first search with appropriate stopping conditions and search heuristics. We present experimental results on synthetic scenes and the backhoe public release dataset.

Recent technology developments in digital radio, low-cost inertial navigation systems and unmanned air vehicle design are converging to enable and make practical several new radar sensing modes such as simultaneous SAR/GMTI from persistent staring-mode radar, 3D SAR from a single-pass, single phase center radar and wide-angle radar tracking of dismounts. One of the challenges for algorithm developers is a lack of high-quality target and clutter signature data from the new radar modes. AFRL's Sensor Directorate and SET Corporation are developing a compact, low-cost wide-angle radar test bed capable of simulating a variety of radar modes, including 3D SAR, SAR/GMTI from staring-mode radar and ultra-fine resolution range-Doppler. We provide an overview of the wide-angle radar test bed architecture, its modular design and our implementation approach. We then describe several non-conventional wide-angle radar sensor modes and outline a corresponding series of test bed data collection experiments that could be used to support the development of new tracking and recognition algorithms.

Construction of the new 350GHz compact range has been completed and it is able to collect fully polarimetric scaled X-band radar data with 6-inch full-scale range resolution. In order to investigate the reproduction of X-band data using scale models, fully polarimetric high-resolution radar signature data has been collected on several targets which include a high-fidelity in-house built 1/16^th scale T72 Main Battle Tank (MBT) and a commercially available 1/35^th scale model T72 modified to match its features. A correlation study of ISAR images has been performed between the X-band data sets collected on these models, a full-scale T72, a 1/35^th scale model heavy equipment transporter, and several different 1/16^th scaled targets of similar size. The ISAR images formed from the data were compared using several techniques which include a two-dimensional cross-correlation of the images against one another, and the comparison of the images pixel-by-pixel to measure the percentage differences. It will be shown that the T72 data sets compare well across the three different radar platforms. It has also been found that there are persistent sharp features in the two-dimensional cross-correlation maps that are located where the real target is matched even when other parameters have changed by a significant amount. These features continue to occur when the target has been imbedded in a complex two-target scene with the heavy equipment transporter.

The convolution backprojection algorithm is an accurate synthetic aperture radar imaging technique, but it has seen limited use in the radar community due to its high computational costs. Therefore, significant research has been conducted for a fast backprojection algorithm, which surrenders some image quality for increased computational efficiency. This paper describes an implementation of both a standard convolution backprojection algorithm and a fast backprojection algorithm optimized for use on a Linux cluster and a field-programmable gate array (FPGA) based processing system. The performance of the different implementations is compared using synthetic ideal point targets and the SPIE XPatch Backhoe dataset.

The convolution/back-projection (CBP) algorithm has recently once again been touted as the "gold standard" for spotlight-mode SAR image formation, as it is proclaimed to achieve better image quality than the well-known and often employed polar formatting algorithm (PFA). In addition, it has been suggested that PFA is less flexible than CBP in that PFA can only compute the SAR image on one grid and PFA cannot add or subtract pulses from the imaging process. The argument for CBP acknowledges the computational burden of CBP compared to PFA, but asserts that the increased image accuracy and flexibility of the formation process is warranted, at least in some imaging scenarios. Because CBP can now be sped up by the proper algorithm design, it becomes, according to this line of analysis, the clear algorithm of choice for SAR image formation. In this paper we reject the above conclusion by showing that PFA and CBP achieve the same image quality, and that PFA has complete flexibility, including choice of imaging plane, size of illuminated beam area to be imaged, resolution of the image, and others. We demonstrate these claims via formation of both simulated and real SAR imagery using both algorithms.

While the chief cause of defocus in airborne spotlight-mode imagery is uncompensated errors in the measurement of the aircraft position as it traverses the synthetic aperture, another physical phenomenon can cause blurring in the formed SAR image as well. This is the injection of phase errors into the collected SAR phase history data by random fluctuations in the index of refraction as the microwave pulses propagate through an atmosphere that contains irregularities in the tropospheric water vapor distribution. In this paper, we show that in SAR imagery collected under certain conditions, these phase errors can be detected and corrected using a robust autofocus algorithm such as Phase Gradient Autofocus (PGA). The phase errors are confirmed as having been propagation-induced by demonstrating that they exhibit a power-law spectrum described by Tatarski, based on the turbulence model of Kolmogorov.

This paper discusses a multistage approach to entropy minimization for SAR image autofocus. The new algorithm is compared to existing approaches, including point based autofocus, sub-aperture based autofocus, and hybrid methods. Monte Carlo statistical results are presented for simulated clutter scenes and point target scenes. The new minimum entropy autofocus provides improved speed and accuracy in correcting azimuth phase errors in both scenarios.

This paper describes a challenge problem whose scope is detection of stationary vehicles in foliage using VHF-band SAR data. The data for this challenge problem consists of images collected by the Swedish CARABAS-II system which produces SAR images at the low VHF-band (20-90 MHz). At these frequencies the electromagnetic energy from the radar penetrates the foliage of the forest, providing a return from a target concealed in a forest. Thus, VHF-band SAR technology transforms the foliage penetration problem into a traditional detection problem where the goal is to reduce the false alarm rate (FAR). Reducing the FAR requires suppressing the clutter in a VHF-band SAR image which is dominated by larger tree trunks, buildings and other man-made objects. The purpose of releasing the CARABAS-II data set is to provide the community with VHF-band SAR data that supports development of new algorithms for robust target detection with a low false alarm rate. The set of images supports single-pass, two-pass and multi-pass target detection.

In this paper a target detection technique based on a rotational invariant wavelet-based scheme is presented. The technique is evaluated on SAR imaging and compared with a previous fractal-based technique, namely the extended fractal (EF) model. Both techniques attempt to exploit the textural characteristics of SAR imagery. Recently a wavelet/fractal feature set, similar to the proposed one, was compared with a feature set similar to EF for a general texture classification problem. The wavelet technique yielded lower classification error than EF, which motivated the comparison between the two techniques presented in this paper. Experimental results show that the proposed technique has the potential for providing lower false alarm rates compared to EF.

This research compares alternative performance metrics to those more commonly used in target detection system performance evaluation. The alternative performance metrics examined here include the Fisher ratio, a modification of the Dice similarity coefficient, and the Youden index. These metrics are compared to metrics that have been previously introduced for such target detection system performance evaluation: the receiver operating characteristic (ROC) curve (and the related summary area under the ROC curve (AUC) value), and the confidence error generation (CEG) curve (and the related summary root square deviation (RSD) value). The ROC curve is a discrimination metric that measures the ability of a target detection system to distinguish between target and non-target. The CEG curve quantifies detection system knowledge of its own performance. An approach is presented that combines such metrics; this combination may be dynamically adjusted and updated based on current and future evaluation requirements for particular target detection systems.

In this paper we propose a Wiener filter-based change detection algorithm for the detection of mines in Synthetic Aperture Radar (SAR) imagery. By computing second order statistics, the Wiener filter-based method has demonstrated improved performance over Euclidean distance. It is more robust to the presence of highly correlated speckle noise, misregistration errors, and nonlinear variations in the two SAR scenes. These variations may result from differences in the data acquisition systems and varying conditions during the different data collect times. A method very similar to the Mahalanobis distance was also implemented to detect mines in SAR images and has shown similar performance to the Wiener filter-based method. We present results in the form of receiver operating characteristics (ROC) curves, comparing simple Euclidean difference change detection, Mahalanobis difference-based change detection, and the proposed Wiener filter-based change detection in both global and local implementations.

The enhancement and improvement of classifiers for SAR signatures are a permanent challenge. The focus of this paper is the development of an integrated decision-and-reject method suitable for a kernel-machine-based target classification framework for SAR scenarios. The proposed processing chain consists of a screening process identifying ROIs with target cues, a pre-processing, and a high-performance classifier. A feasible screening method has to provide a maximum of detections namely object hypotheses while the false alarm rate is of lower interest. Therefore the quality of the following classification step significantly depends on the capability of reducing the false alarms. In complex scenarios standard approaches may classify clutter objects incorrectly as targets. To overcome this problem a novel classification scheme was developed. Class discriminating information is computed in a pre-classification step by a family of two-class kernel machines. Thus, a feature vector for an additional classification stage is provided. A comparative assessment was done using a SAR data set provided by QinetiQ. First results are given in terms of ROC curves.

We consider the joint inverse problems of sensor data registration and automatic target recognition. Single-platform, multi-sensor registration is posed as a model-based, data fusion problem using Bayesian and maximum likelihood frameworks. The sensor model parameters typically consist of platform pose parameters, sensor pointing angles, and internal calibration factors, and these are used to define a transformation that maps raw data recorded in the sensor frame to a ground-referenced, world coordinate system. The fusion estimation problem is one joint inversion since the sensor model parameters common to multiple sensors are simultaneously estimated (along with sensor-specific model parameters). For the ATR problem we pose the joint optimization problem over these sensor model parameters (constrained by the global scene) and target model parameters (e.g., for selected target chips). In addition, we pose a cooperative inversion approach that captures uncertainty from the system model estimation process for use in a refined ATR inversion. The latter consists of a search over target model parameters and a constrained system model parameter space with realization samples consistent with the estimated system model covariance. Estimation robustness is achieved through use of a hybrid global/local search method (to avoid final convergence to local minima), robust kernels that down-weight data residual outliers (generated from test and reference image feature correspondences), and the use of multi-sensor data to increase the number and diversity of data constraints. In summary, we have developed a model-based fusion approach which draws on well-developed methods in photogrammetry, computer vision and automatic target recognition for enhanced registration and recognition performance.

There is a strong and growing need for automatic target recognition (ATR) technologies. Those technologies have made great strides; however, there is a general sense that they are not having the full impact desired. This paper develops a value-based framework for considering how ATR technology can be made more relevant and then introduces and expands on two elements within that framework: 'enhancements' and 'accommodations'. Value is used here as the degree to which a technology's benefits exceed the technology's costs. Value may be improved by increasing benefits or decreasing costs; but it may be as important that the uncertainty about benefits and costs be reduced. Enhancements and accommodations are distinguished here from the 'core ATR'. While it is generally appreciated that improved core ATR performance could improve value, enhancements and accommodations might be overlooked by those focused on ATRs. Enhancements are ways of making the overall system, inclusive of a core ATR, more capable. Accommodations are ways of making the problem easier for the core ATR. An example enhancement is technology to fuse the output of the core ATR with other sources. An example accommodation is for the user to agree to limit the target set to large, and therefore more easily recognized, objects. This paper encourages the consideration of this framework and outlines a number of candidates for enhancements and accommodations for synthetic aperture radar (SAR) ATR, including humans-in-the-loop, change detection, fusion, modeling confusers, group detection, adaptive algorithms, class make-up, and scene-based decisions.

Without motion compensation, Synthetic Aperture Radar (SAR) images of the ground are generally blurred. In 1997, MITRE reported the development technique called the Keystone Process for removing the range migration caused by the radial velocity component of each pixel's movement within the scene, whether moving or stationary with respect to the ground. When applied to multiple phase center phased array radar data, this first pass process allows for automated detection of moving targets via phase thresholding. Once detected in phase space, the moving targets can be individually and automatically focused using the procedures previously reported. Automated positioning of the detected target within the formed image is then accomplished (georegistration). We can easily detect and accurately georegister bright (large radar cross-sections) moving targets using a phase threshold technique reported herein. However, we have found that, for smaller targets, the phase differences between the cells containing the moving target can be greatly distorted by the presence of strong ground clutter. Only after the ground clutter is cancelled will the phase difference be sufficiently dominated by the target response to allow accurate geopositioning. Herein we describe one technique whereby the clutter may be cancelled by using multiple phase centers.

This paper examines the theory, application, and results of using single-channel synthetic aperture radar (SAR) data with Moving Reference Processing (MRP) to focus and geolocate moving targets. Moving targets within a standard SAR imaging scene are defocused, displaced, or completely missing in the final image. Building on previous research at AFRL, the SAR-MRP method focuses and geolocates moving targets by reprocessing the SAR data to focus the movers rather than the stationary clutter. SAR change detection is used so that target detection and focusing is performed more robustly. In the cases where moving target returns possess the same range versus slow-time histories, a geolocation ambiguity results. This ambiguity can be resolved in a number of ways. This paper concludes by applying the SAR-MRP method to high-frequency radar measurements from persistent continuous-dwell SAR observations of a moving target.

We propose a novel approach to focus and geolocate moving targets in synthetic aperture radar imagery. The initial step is to detect the position of the target using an automatic target detection algorithm. The next step is to estimate the target cross-range velocity using sequential sub-apertures; this is done by forming low resolution images and estimating position as a function of sub-aperture, thus yielding an estimate of the cross-range velocity. This cross-range estimate is then used to bound the search range for a bank of focusing filters. Determining the proper velocity that yields the best focused target defines an equation for the target velocity, however both components of the targets velocity can not be determined from a single equation. Therefore, a second image with a slightly different heading is needed to yield a second focusing velocity, and then having a system of two equations and two unknowns a solution can be obtained. Once the target velocity is known the proper position can be determined from the range velocity. Synthetic data will be used with a point source target and both background clutter and noise added. The results support the development of staring radar applications with much larger synthetic aperture integration times in comparison to existing SAR modes. The basic idea of this approach is to trade-off the development of expensive phased-array technology for GMTI applications with the potential development of advanced processing methods that show potential for processing data over very large aperture integration intervals, to obtain similar GMTI geolocation results that would be compatible with current radar technology.

SAR-MTI is a generalization of SAR processing and can work with only a single-phase center. SAR-MTI requires formation of a stack of SAR images assuming different sensor ground speeds. Each image will capture a different set of target velocities, and the complete set of images will focus all target speeds less than a desired maximum speed regardless of direction and target location. SAR-MTI has no minimum detectable velocity. SAR-GMTI will have higher resolution than standard GMTI because the SAR based processing allows longer coherent processing intervals (CPI). A realistic dense moving target scenario was simulated using SAR-GMTI detection and standard GMTI detection was simulated for this scenario and the frequency of unresolved targets was compared. SAR-GMTI suffered many fewer unresolvable target pairs because of its smaller resolution cell and because it inherently detects in a 3 dimensional space vs. a two dimensional space. Furthermore SAR-GMTI does not have a clutter notch, which eliminated about 15% of the moving targets for standard GMTI.