William W. Irving

Multiscale segmentation and anomaly enhancement of SAR imagery

We present efficient multiscale approaches to the segmentation of natural clutter, specifically grass and forest, and to the enhancement of anomalies in synthetic aperture radar (SAR) imagery. The methods we propose exploit the coherent nature of SAR sensors. In particular, they take advantage of the characteristic statistical differences in imagery of different terrain types, as a function of scale, due to radar speckle. We employ a class of multiscale stochastic processes that provide a powerful framework for describing random processes and fields that evolve in scale. We build models representative of each category of terrain of interest (i.e., grass and forest) and employ them in directing decisions on pixel classification, segmentation, and anomalous behaviour. The scale-autoregressive nature of our models allows extremely efficient calculation of likelihoods for different terrain classifications over windows of SAR imagery. We subsequently use these likelihoods as the basis for both image pixel classification and grass-forest boundary estimation. In addition, anomaly enhancement is possible with minimal additional computation. Specifically, the residuals produced by our models in predicting SAR imagery from coarser scale images are theoretically uncorrelated. As a result, potentially anomalous pixels and regions are enhanced and pinpointed by noting regions whose residuals display a high level of correlation throughout scale. We evaluate the performance of our techniques through testing on 0.3-m resolution SAR data gathered with Lincoln Laboratorys millimeter-wave SAR.

A multiresolution approach to discrimination in SAR imagery

We develop and test a new algorithm for discriminating man-made objects from natural clutter in synthetic-aperture radar (SAR) imagery. This algorithm exploits characteristic variations in speckle pattern as image resolution is varied from course to fine. We model these variations as an autoregression in scale, and then use the autoregressive model to define a multiresolution log-likelihood ratio discriminant. We incorporate this discriminant into the existing Lincoln Laboratory SAR system for automatic target recognition (ATR), and test the augmented system by applying it to millimeter-wave SAR imagery having 0.3 m resolution and representing 56 square kilometers of terrain. At a probability of detection of 0.95, the addition of the multiresolution discriminant reduces the number of natural-clutter false alarms by a factor of six.

Multiresolution approach to discriminating targets from clutter in SAR imagery

We develop and extensively test a new algorithm for discriminating man-made objects from natural clutter in synthetic-aperture radar (SAR) imagery. The novel feature of our approach is its exploitation of the characteristically distinct variations in speckle pattern for imagery of man-made objects and of natural clutter, as image resolution is varied from coarse to fine. We treat these characteristics using stochastic framework, specifically tailored for multiresolution random processes and fields. Within the framework, we build a pair of multiscale models: one for SAR imagery of natural clutter and another for imagery of man-made objects. We then use these models to define a multiresolution discriminant as the likelihood ratio for distinguishing between the two image types, given a multiresolution sequence of images of a region of interest (ROI). We incorporate this likelihood ratio into an existing, established discriminator that was developed at Lincoln Laboratory (LL) as part of a complete system for automatic target recognition (ATR). To classify a given ROI, we merge the information provided by our likelihood ratio with the measured values of a small number of size and brightness features. We have applied the resulting new discriminator to an extensive data set of 0.3-meter resolution, HH polarization imagery gathered with the LL millimeter-wave SAR. The detection results are extremely good. In particluar, the new discriminator achieves a significant improvement in receiver operating characteristics, compared to an optimized version of the standard discriminator that is traditionally used in the LL ATR system. This result conclusively demonstrates that multiresolution methods have an effective and important role to play in SAR ATR algorithms.

Optimal polarimetric processing for enhanced target detection

The results of a study of several polarimetric target detection algorithms are presented. The study concerns the Lincoln Laboratory millimeter-wave SAR sensor, a fully polarimetric, 35 GHz synthetic-aperture radar. Fully polarimetric measurements (HH, HV, VV) are processed into intensity imagery using adaptive and nonadaptive polarimetric whitening filters (PWFs), and the amount of speckle reduction is quantified. Then a two-parameter CFAR (constant false alarm rate) detector is run over the imagery to detect the targets. Nonadaptive PWF processed imagery is shown to provide better detection performance than either adaptive PWF processed imagery or single-polarimetric-channel HH imagery. In addition, nonadaptive PWF processed imagery is shown to be visually clearer than adaptive PWF processed imagery.<<ETX>>

A canonical correlations approach to multiscale stochastic realization

We develop a realization theory for a class of multiscale stochastic processes having white-noise driven, scale-recursive dynamics that are indexed by the nodes of a tree. Given the correlation structure of a 1-D or 2-D random process, our methods provide a systematic way to realize the given correlation as the finest scale of a multiscale process. Motivated by Akaikes use of canonical correlation analysis to develop both exact and reduced-order models for time-series, we too harness this tool to develop multiscale models. We apply our realization scheme to build reduced-order multiscale models for two applications, namely linear least-squares estimation and generation of random-field sample paths. For the numerical examples considered, least-squares estimates are obtained having nearly optimal mean-square errors, even with multiscale models of low order. Although both field estimates and field sample paths exhibit a visually distracting blockiness, this blockiness is not an important issue in many applications. For such applications, our approach to multiscale stochastic realization holds promise as a valuable, general tool.

Bounding performance of peak-based target detectors

We develop a theoretical bound on the receiver operating characteristics (ROC) associated with a target detector that bases its decision on the spatial distribution of extracted peak locations in synthetic-aperture (SAR) imagery. There are three basic steps to our analysis. In the first step, we formulate statistical models for both target and clutter regions of interest that have passed through a pre-screening stage. From these models, we then infer a corresponding detection procedure, which takes the form of a generalized likelihood ration test (GLRT). Finally, in the third step, we again use our statistical models to determine the ROC performance of the GLRT. This third step is where we believe our primary technical innovation lies. In the presence of uncertainty regarding target type and pose, it is generally difficult to obtain analytical performance expressions. We circumvent this problem by treating statistically the database of target exemplars. This approach has tow benefits. First, it suppresses the details of the database of exemplars, alleviating the need to specify the spatial configuration of points for every database entry. second, it leads to analytical performance expressions, which can be calculated with considerably less effort than is possible with Monte Carlo simulation.

Multiresolution model development for overlapping trees via canonical correlation analysis

A class of multiscale stochastic models has been introduced in which Gaussian random processes are described by scale-recursive dynamics that are indexed by the nodes of a tree. One of the primary reasons the framework is useful is that it leads to an extremely fast, statistically optimal algorithm for least-squares estimation in the context of 2-D images. We refine this approach to estimation by eliminating the visually distracting blockiness that has been observed in the previous work. We eliminate the blockiness by discarding the standard assumption that distinct nodes at a given level of our tree correspond to disjoint portions of the image domain; as a consequence of this simple idea, a given image pixel may now correspond to several tree nodes. We develop tools for systematically building overlapping-tree multiscale representations of prespecified statistics, and we develop a corresponding estimation algorithm for this processes. In this way, we achieve nearly optimal estimation results, we generate corresponding error covariance information, and we eliminate blockiness without sacrificing the resolution of fine-scale detail.

A new model for high-resolution polarimetic SAR clutter data

A new product clutter model aimed at fitting the cumulative distribution function of high resolution SAR data is proposed. The model uses only the sample mean and variance of a region, and the number of uncorrelated images of the regions to estimate the cumulative distribution function of the power return. It is shown that the model is capable of accurately modeling grass and tree clutter for both single- and multipolarization data.

High resolution radar target identification

The application of neural networks to the synthetic aperture radar (SAR) automatic target recognition (ATR) problem is discussed. In particular, initial studies investigating the use of an ART-2 self organizing neural-network in a 2-D SAR ATR system are summarized. The performance of the new neural net pattern-matching algorithm is compared with that of the baseline correlation pattern matching algorithm developed previously. This comparison includes evaluating the ability of the pattern matcher to reject nontargets (clutter discretes) and to classify the remaining detections into tank/APC/Howitzer categories.<<ETX>>

Adaptive Processing of Polarimetric SAR Imagery

This paper presents the results of a study of several polarimetric target detection algorithms. Fully polarimetric measurements (HH, HV, VV) are processed into intensity imagery using adaptive and non-adaptive polarimetric whitening filters (PWFs); then a twoparameter CFAR detector is run over the imagery to detect the targets. Non-adaptive PWF processed imagery is shown to provide better detection performance than either adaptive PWF processed imagery or single-polarimetric-channel HH imagery. In addition, non-adaptive PWF processed imagery is shown to be visually clearer than adaptive PWF processed imagery.