|
Multisensor Threat Detection
Law enforcement and military officials place a high priority on protecting citizens and forces from concealed
human-carried explosives carried by suicide bombers. Required detection ranges extend beyond 100 meters; however,
modern hidden weapon detection technologies use portal configurations that necessitate close-in sensing, putting
security professionals at risk. SET is developing portable, low-cost detection systems that integrate heterogeneous
sensors to extend operating ranges beyond the blast zone of the bomb, and expand the system performance envelope
ver a wide range of threats and operating environments, including crowded pedestrian areas. Through the synergistic
use of multiple sensors, our prototype systems have been shown in independent government experiments to reduce
operator workload and increase operator performance duration. 
Compact, Lightweight Dual Mode SAR/GMTI Radar
SET has designed a compact lightweight SAR/MTI radar system for the Micro Air Vehicle (MAV) that will provide significant
technical advances and operational flexibility for this class of aircraft. The radar consists of a sub-miniature electronically
scanned array that combines all RF and digital electronics into a single integrated package. The radar is built with standard
microelectronic fabrication methods. It is inherently mass producible, yielding very high reliability, low weight, and low cost.
The integrated design uses efficient waveforms that require low power, while providing high stability, low MDV, and low target
location error. The design provides 90? of coverage that is readily expandable to a full 360? by adding more ESA modules.
Unlike mechanically scanned approaches, the SET design provides a compact lightweight package that has no moving parts, and
is mountable anywhere on the exterior surface of the MAV.
Wall Penetrating Radar
SET is developing an advanced RF exploitation workstation for detecting, tracking, and characterizing human signatures
using building-penetrating (through-wall) radar. Our tracking algorithms cancel stationary background clutter, localize
people in 3D through exterior-grade walls, and account for wall-propagation delay and multi-path. Our 3D imaging algorithms
are designed to compensate for through-wall signal distortions and provide continously-updated situation awareness for the
user. Human micro-Doppler signatures are exploited to identify characteristic motions and to support feature-aided track
maintenance. These capabilities are combined with indoor activity models to provide a virtual rendering of human activities
within the building.
SAR Automatic Target Recognition
Synthetic aperture radar makes all-weather, stand-off target recognition feasible. Unlike conventional
ATR approaches that rely on large repositories of expensive data exemplars, model-based approaches hold
the promise of making SAR ATR more robust, accurate, and affordable through the use of real-time computational
electromagnetics engines. SET researchers have been at the forefront of model-based ATR technology development,
and are currently advancing state of the art model based techniques to take advantage of next-generation multidimensional
SAR sensing capabilities. These techniques focus on improving SAR ATR robustness and accuracy under challenging
operating conditions, including line-of-sight obscuration, background layover, and target adjacency.
3D Shape Representations for LADAR ATR
The core of many modern Automatic Target Recognition (ATR) approaches consists of matching sensor data against a model base of
known objects. We have developed a testbed for LADAR ATR and implemented multiple 3D shape representation and matching techniques.
The testbed includes multiple LADAR data sets and multiple 3D vehicle and object model sets. Principal measures of performance are
match accuracy, speed of matching data to the model base, and compactness of the model base. We have chiefly focused on compact
shape representation and matching approaches. A novel technique we developed takes less than a second to match a LADAR image to
the top candidates in a model base of dozens of vehicles, making this approach feasible for real-time seeker applications.
In addition to fundamental shape representation and matching technology, we have developed capabilities to automatically detect
and segment the ground, visualize and explore LADAR images, and visualize the shape representations. The core testbed code runs
in batch mode to find aggregate experiment statistics and in a GUI that supports interactive and automatic play modes.
|
