By John McHale
WASHINGTON - Researchers at the U.S. Naval Research Laboratory (NRL) in Washington are trying to develop a real-time automatic target recognition system for future pilotless combat aircraft by using hyperspectral sensors and algorithms running on Analog Devices SHARC digital signal processors (DSPs).
Hyperspectral sensors record reflected light in many narrow contiguous bands in the visual, near-infrared, and short-wavelength infrared regions of the electromagnetic spectrum. They provide fine-grain images, but pose complex signal processing problems.
The NRL scientists developed the signal processor and interface box for the hyperspectral sensor with a Quad SHARC DSP, and an Intel Pentium II processor to keep false alarms to a minimum, and increase the accuracy of ground target analysis, says Frank Bucholtz, flight systems integrator at NRL.
The system, tested aboard a Navy Lockheed Martin P-3 Orion four-engine turboprop aircraft, also uses pointing optics and autonomous cueing of a high-resolution imager, and designates targets with pointing optics and a pulsed laser, he says. The system integrates a sensor grating from Instruments S.A. in Edison, N.J., and a camera from David Sarnoff Research in Princeton, N.J., Bucholtz says.
NRL specialists supervised the work under the "Dark Horse" program, a four-year effort under sponsorship of the Office of Naval Research to develop and demonstrate real-time hyperspectral detection, cueing, target location, and target designation. NRL engineers developed the algorithms under the Joint Multi-Spectral Program.
The two algorithms operate simultaneously to verify information from one another and thereby reduce false alarms and improve accuracy.
The system demonstrated the ability to detect ground targets, as well as the capability to detect and cue in real time on aircraft flying 4,000 feet below the P-3 test aircraft on which the system was mounted.
A radio link communicates with a ground station when the system identifies a target.
This drawing represents how a hyperspectral imaging, cueing, and detection system on a future unmanned combat aerial vehicle might function. Hyperspectral imaging divides detected light into narrow light bands for fine-grain processing and data fusion.
