One of the challenges of using synthetic aperture radar (SAR) to detect and classify an object behind a wall
consists of determining the amount of signal attenuation introduced by the signal's propagation through the wall.
This attenuation is difficult to determine because the electromagnetic properties of the wall, along with its
thickness are normally not known a priori. We describe a procedure for determining the necessary parameters
given that the SAR has high enough resolution such that the front and the rear surfaces of a uniform wall or cinder
block wall can be determined from the SAR image. In addition, we provide a procedure for estimating the signal
level behind the wall, or equivalently the attenuation due to the wall, from measured returns from its front and rear
surfaces. We demonstrate the effectiveness of this procedure using data generated by XPATCH simulations.
One of the challenges of using synthetic aperture radar (SAR) to detect a man behind a wall is determining the
amount of signal attenuation introduced by the signal's propagation through the wall. This attenuation is difficult
to determine because the thickness and the electromagnetic properties of the wall are normally not known a priori.
We describe a procedure for estimating the relative permittivity, conductivity, and thickness of the wall that minimize the error between physics-based predicted values of wall return and the corresponding values of the
SAR image. The accuracy of the prediction is a function of the resolution of the SAR image relative to the
thickness of the wall-the SAR image must have sufficient resolution such that the locations of the front and rear
surfaces of a uniform wall can be estimated from the SAR image. The signal level behind the wall, or equivalently
the signal attenuation by the wall is then determined from the estimates of the thickness and electromagnetic
parameters. We demonstrate the effectiveness of this identification procedure using data generated by XPATCH
simulations of three different wall materials.
KEYWORDS: Sensors, Global Positioning System, Unattended ground sensors, Monte Carlo methods, Autoregressive models, Error analysis, Target detection, Time metrology, Sensor networks, Analytical research
We present a procedure wherein unattended ground sensors (UGSs) that are not equipped the GPS can locate their own positions by transmitting pulses and receiving retransmitted pulses from UGSs that are equipped the GPS. T The payoff of this approach is reduced cost for the network of UGSs. We show through simulation that the implementation of this procedure locates the sensors that do not have GPS with sufficient accuracy for the network of UGS to detect and locate moving targets.
We present a procedure for classification of targets by a network of distributed radar sensors deployed to detect, locate and track moving targets. Estimated sensor positions and selected positions of a target under track are used to obtain the target aspect angle as seen by the sensors. This data is used to create a multi-angle profile of the target. Stored target templates are then matched in the least mean square sense with the target profile. These templates were generated from radar return signals collected from selected targets on a turntable. Probabilities of correct classification obtained by a simulation of the classification procedure are given as functions of signal-to-noise ratios and errors in estimates of target and sensor locations.
The Army Research Laboratory (ARL) has examined single-polarity, synthetic aperture radar (SAR) data collected in spotlight mode at X band as part of an effort to identify land mines in radar imagery. This data set consisted of several single-polarization, extremely high-resolution, spotlight-mode SAR images from multiple passes over a common target area that included various reference reflectors as well as the landmines. In particular, certain mines were placed in the scene for only a portion of the data collection, creating an opportunity for the investigation of various change detection algorithms. We describe a sub-optimal, yet effective, change detection algorithm based on a Mahalanobis-like distance metric, and we apply it to a recently collected SAR data set.
The Army Research Laboratory (ARL) has recently examined single-polarity, synthetic aperture radar (SAR) data collected in spotlight mode at X band as part of an effort to identify land mines in radar imagery. This data set consists of several single-polarization, extremely high-resolution, spotlight-mode SAR images from multiple passes over a common target area that includes various reference reflectors as well as the landmines. In earlier investigations at Ku band, we observed that a multi-look averaging scheme could enhance the contrast between mines and background clutter. In the most recent investigation, we hypothesize that a similar behavior would be present in the X band imagery, and we demonstrate how the enhanced contrast from multi-look processing leads to improved prescreener performance under certain conditions. Results are presented in the form of receiver operating characteristic (ROC) curves for different several different prescreener parameter settings.
When a synthetic aperture radar, employing an ultra-wideband impulse waveform, forms an image of an area on the ground by using a backprojection algorithm, artifacts occur due to any strong reflectors in the scene. These artifacts can obscure or influence the image of the area containing low-level reflectors that may be of interest. Herein, we examine the factors of the imaging geometry that leads to these artifacts or influences and propose a procedure for reducing them. We demonstrate the artifacts and the procedure to reduce them with the data collected by the ultra wideband radar of the U.S. Army Research Laboratory.
Trained algorithms are required for detecting stationary targets with practical real-beam radars. The parameters of these algorithms are unique to each site or clutter class. A problem arises when an algorithm trained on one clutter class is applied, perhaps inadvertently, to another class. In this case, the performance of the system can degrade to an unacceptable level. We have developed a system that adapts, online, the parameters of the algorithm to the
encountered clutter type. This system consists of two neural networks - one for adapting the coefficients of the algorithm and the other for adapting the threshold level.
The U.S. Army Research Laboratory has investigated the relative performance of three different target detection paradigms applied to foliage penetration (FOPEN) synthetic aperture radar (SAR) data. The three detectors - a quadratic polynomial discriminator (QPD), Bayesian neural network (BNN) and a support vector machine (SVM) - utilize a common collection of statistics (feature values) calculated from the fully polarimetric FOPEN data. We describe the parametric variations required as part of the algorithm optimizations, and we present the relative performance of the detectors in terms of probability of false alarm (Pfa) and probability of detection (Pd).
The Army Research Laboratory has investigated various phenomenology-based approaches for improving the detection of targets in wide-angle, ultra-wideband foliage penetration synthetic aperture radar (SAR) data. The approach presented here exploits the aspect-dependent reflectivity of vehicles, by filtering the SAR image data to obtain sub-aperture images from the original full-aperture radar image. These images represent the images of the target as seen by the sub-aperture SAR from two different locations (squint angles). We present a straightforward approach to extending an existing collection of features for a quadratic polynomial discriminator with features calculated from these two, lower-resolution sub-aperture SAR data images. We describe a method for generating the modified features and assess their potential contribution to improved probability of detection.
Many common clustering algorithms, such as the fuzzy C-means and the classical k-means clustering algorithms, proceed without making any assumptions about the form of the detector that will use the parameters that they determine. We compare the performance of a radial basis function (RBF) network with parameters that are determined using a modified fuzzy clustering procedure to that of an RBF network with parameters that are determined using a least-mean-square- error (classical) clustering procedure. As part of the fuzzy clustering procedure, we assume a particular functional form for the fuzzy membership function. We train and test both of the networks on simulated data and present performance results in the form of receiver operating characteristic curves.
Many radar automatic target detection (ATD) algorithms operate on a set of data statistics or features rather than on the raw radar sensor data. These features are selected based on their ability to separate target data samples from background clutter samples. The ATD algorithms often operate on the features through a set of parameters that must be determined from a set of training data that are statistically similar to the data set to be encountered in practice. The designer usually attempts to minimize the number of features used by the algorithm -- a process commonly referred to as pruning. This not only reduces the computational demands of the algorithm, but it also prevents overspecialization to the samples from the training data set. Thus, the algorithm will perform better on a set of test data samples it has not encountered during training. The Optimal Brain Surgeon (OBS) and Divergence Method provide two different approaches to pruning. We apply the two methods to a set of radar data features to determine a new, reduced set of features. We then evaluate the resulting feature sets and discuss the differences between the two methods.
The quadratic polynomial detector (QPD) and the radial basis function (RBF) family of detectors -- including the Bayesian neural network (BNN) -- might well be considered workhorses within the field of automatic target detection (ATD). The QPD works reasonably well when the data is unimodal, and it also achieves the best possible performance if the underlying data follow a Gaussian distribution. The BNN, on the other hand, has been applied successfully in cases where the underlying data are assumed to follow a multimodal distribution. We compare the performance of a BNN detector and a QPD for various scenarios synthesized from a set of Gaussian probability density functions (pdfs). This data synthesis allows us to control parameters such as modality and correlation, which, in turn, enables us to create data sets that can probe the weaknesses of the detectors. We present results for different data scenarios and different detector architectures.
We have applied the Optimal Brain Surgeon (OBS) pruning strategy to a polynomial discriminator in order to reduce the number of coefficients it employs. The polynomial discriminator multiplies various combinations of test features by the respective coefficients and then sums the products to obtain a discriminant that is compared to a threshold. The test features are derived from the radar data associated with the cell under test, while the coefficients are determined a priori by minimizing the mean-squared error (MSE) between the actual and the desired value of the discriminant over the training set. The OBS pruning strategy examines the Hessian matrix of a network's error surface-- derived from the training data--to determine which coefficients can be eliminated without adversely affecting the MSE. Besides simplifying the network, such a reduction may also allow for improved network performance when an unseen test data is input. We present the application of the OBS pruning strategy to reduce the dimensionality of a polynomial discriminator and show that the reduction in dimensionality does not adversely affect performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.