Robert Linnehan

Multistatic scattering from moving targets in multipath environments

We are examining the feasibility of ground target detection and tracking in urban centers using ground moving target indicator (GMTI) radar when line-of-sight (LOS) coverage is intermittent, yet multipath responses from building walls are available. We enhance the standard ray-tracing propagation methods that assume specular wall reflections and point scattering target models. Our two-dimensional analysis includes near field reradiation from the illuminated wall to a target having a multistatic response. We also evaluate the degradation of power and coherency in the signal processing due to wall surface roughness. Results of this work provide insight for the development of processing techniques that may be required for multipath exploitation radar.

Dynamic Logic Applied to SAR Data for Parameter Estimation Behind Walls

Identifying and localizing targets within buildings using exterior sensors will offer superior advantages to the military and law enforcement communities. Research on wall-penetrating radar has produced significant advances in recent years regarding this topic. However, wall parameter ambiguities, multiple reflections, clutter, and measurement noise pose significant challenges to developing robust detection and estimation methods. In the present work we demonstrate can be mitigated using dynamic logic (DL), an adaptive method for iterative maximum likelihood.

Detecting multiple slow-moving targets in SAR images

Ground moving target indication (GMTI) radars can detect slow-moving targets if their velocities are high enough to produce Doppler frequencies distinguishable from the surrounding stationary clutter. However, no reliable technique is currently available to detect targets that fall below the minimum detectable velocity (MDV) of GMTI radars. Detecting slow moving targets in synthetic aperture radar (SAR) images has also not ceded a reliable solution. Reflected energy from the target is spread over many pixels in the image due to its motion, degenerating the detection process. The addition of clutter from surrounding stationary objects or ground features further complicates detection. Several techniques for SAR imaging of moving targets have been attempted. These techniques require pre-detection, which, in turn, requires sufficient signal-to-stationary ground clutter ratio (SCR). Other attempts such as adaptive, model-based approaches face exponential combinatorial complexity. Exponential computational cost results from having to consider a large number of combinations between multiple target models and the data. The dynamic logic algorithm (DLA) presented below detects multiple slow-moving targets simultaneously in SAR images with low signal-to-clutter ratio, no minimum velocity requirement, and without combinatorial complexity. The mathematics underlying the algorithm is based on biologically inspired signal processing concepts.

Detecting slow moving targets in SAR images

Ground moving target indication (GMTI) radars can detect slow-moving targets if their velocities are high enough to produce distinguishable Doppler frequencies. However, no reliable technique is currently available to detect targets that fall below the minimum detectable velocity (MDV) of GMTI radars. In synthetic aperture radar (SAR) images, detection of moving targets is difficult because of target smear due to motion, which could make low-RCS targets fall below stationary ground clutter. Several techniques for SAR imaging of moving targets have been discussed in the literature. These techniques require sufficient signal-to-clutter ratio (SCR) and adequate MDV for pre-detection. Other techniques require complex changes in hardware. Extracting the maximum information from SAR image data is possible using adaptive, model-based approaches. However, these approaches lead to computational complexity, which exceeds current processing power for more than a single object in an image. This combinatorial complexity is due to the need for having to consider a large number of combinations between multiple target models and the data, while estimating unknown parameters of the target models. We are developing a technique for detecting slow-moving targets in SAR images with low signal-to-clutter ratio, without minimal velocity requirements, and without combinatorial complexity. This paper briefly summarizes the difficulties related to current model-based detection algorithms. A new concept, dynamic logic, is introduced along with an algorithm suitable for the detection of very slow-moving targets in SAR images. This new mathematical technique is inspired by the analysis of biological systems, like the human brain, which combines conceptual understanding with emotional evaluation and overcomes the combinatorial complexity of model-based techniques.

Multipath analysis of dismount radar responses

This work continues from our article published in RadarCon 2010 titled “Validating Multipath Responses of Moving Targets Through Urban Environments”, where we describe the collection and analysis of multipath radar responses from vehicles passing in front of a two-story brick building at Wright-Patterson AFB. In May 2010 we performed a similar experiment that included line-of-sight (LOS) and multipath measurements of dismounts (humans) moving at various speeds and directions in front of the same two-story building. Examination of the data at X-band shows strong multipath responses of the dismounts from specular regions on the background wall, as well as from other structures present in the scene, e.g., a light pole and metal downspouts. Micro-motion analysis of the multipath signatures yields features similar to those extracted from the corresponding direct path responses, suggesting that dismounts could be characterized even when they are not in LOS of the radar. Furthermore, preliminary results show the potential to enhance dismount classification and localization when both direct path and multipath signatures are available.

Validating multipath responses of moving targets through urban environments

This work proceeds from the paper we published in RadarCon 2009 titled “Multistatic scattering from moving targets in multipath environments”, where we explored the potential to track moving ground targets with radar as they enter urban areas and become obscured by buildings. An X-band radar data collection was performed which validates the predicted multipath response, and the received multipath power in relation to the line-of-sight (LOS) response. Results from a bistatic experiment are used to examine the spatial coherency of energy reflecting from a large, rough surface, and the power distribution in angle that illuminates a target as it traverses in front of a building. This experiment may inspire knowledge-based methods to coherently process multipath returns, beyond that of standard GMTI processing, i.e., free-space matched-filtering (FFT) and CFAR detection.

On the design of SAR apertures using the Cramer-Rao bound

The Cramer-Rao inequality is applied to the likelihood function of the synthetic aperture radar (SAR) scatterer parameter vector to relate the choice of flight path to estimation performance. Estimation error bounds for the scatterer parameter vector (including height) are developed for multi-dimensional synthetic apertures, and quantify the performance enhancement over a limited sector of the image plane relative to standard-aperture single-pass SAR missions. An efficient means for the design and analysis of SAR waveforms and flight paths is proposed using simulated scattering models that are limited in size. Comparison of the error bounds to those for standard-aperture SAR show that estimates of scatterer range and cross-range positions are accurate for multi-dimensional aperture SAR, even with the additional estimator for height. Furthermore, multi-dimensional SAR is shown to address the layover problem

Object pattern recognition below clutter in images

We are developing a technique for recognizing patterns below clutter based on modelling field theory. The presentation briefly summarizes the difficulties related to the combinatorial complexity of computations, and analyzes the fundamental limitations of existing algorithms such as multiple hypothesis testing. A new concept, dynamic logic, is introduced along with an algorithm suitable for pattern recognition in images with intense clutter data. This new mathematical technique is inspired by the analysis of biological systems, like the human brain, which combines conceptual understanding with emotional evaluation and overcomes the combinatorial complexity of model-based techniques. The presentation provides examples of object pattern recognition below clutter.

Synthesis of formal and fuzzy logic to detect patterns in clutter

Recognizing patterns in data often relies on rules, or exploits simple features in the data. However, when noise or clutter obscures these features in the data, one must consider a number of different features to determine the best match. This often leads to combinatorial complexity manifested in either of two ways, complexity of learning or complexity of computations. Adaptive model-based approaches potentially offer better computational performance than feature-based methods and may lead to extracting the maximum information from data. These techniques still often relied on using formal logic to compare library models to incoming data. Neural networks are usually not easy for implementing model-based approaches. Fuzzy logic bypasses using formal logic, but it provides solutions that often are heavily influenced by the initial degree of fuzziness. We are developing a technique for detecting patterns below clutter based on the neural network modeling field theory. Modeling field theory (MFT) using fuzzy dynamic logic to overcome combinatorial complexity is introduced along with an algorithm suitable for the detection of patterns below clutter. This new mathematical technique is inspired by the analysis of biological systems, like the human brain, which combines conceptual understanding with emotional evaluation and overcomes the combinatorial complexity of model-based techniques.

Evaluation and Tuning of a SAR Detector using Sparse-Array Spotlight Mode Simulations

We describe a method of tuning a simple detection process of stationary targets in SAR images. The tuning metric accounts for the squared-error performance of prescient estimators of target location and reflectivity. The efficiencies of these prescient estimators are compared to their Cramer Rao bounds (CRBs). The off-line tuning is performed by collecting statistics of scatterer estimation in SAR images created using sparse-array spotlight mode simulations