Infrared Search and Track (IRST) Systems are one component of a multisensor suite which can meet the technical challenge of the timely detection/track/identification of low signal-to-(noise+clutter) ratio targets. Cruise missiles over land and sea cluttered backgrounds as well as ballistic missiles are serious threats to IRST. These threats are stealth in both the infrared and radio frequency bands. That is, their thermal infrared signature and their radar cross section can be very small.

In this project, CAMS investigators have developed new technology that allows for the timely detection, tracking, and identification of such threats. This technology is based on nonlinear filtering and adaptive sequential change-point detection techniques for track before detect, detection and tracking of targets. The developed algorithms are computationally efficient and, yet, almost optimal from the statistical point of view.

Related Publications:

These two movies illustrate the performance of the spatial-temporal clutter suppression filter and the results of the detection and tracking for IR data sets recently collected by Dr. John Barnett  at SPAWAR Systems Center, San Diego, CA. The data sets provided for processing by our algorithms are from an Amber High Speed Midwave infrared camera, which collects data  in the 3-5 micron waveband at the rate 30 frames/sec. The data in these sequences were taken from a height of 80 feet  above mean sea level, off Point Loma CA, looking west over  the ocean. The data sets contain strong sea glint and cloud clutter. The cloud borders produce random pulsations. These pulsations generate rather strong outliers. The first  movie  shows an input cluttered, jittered and noisy frame, the second movie shows the residuals at the output of the filter along with the results of tracking, and the statistic movie  shows the  behavior of the detection statistic. The detection is declared when the statistic crosses the threshold. It can be seen that clutter is completely removed, and the residuals look like  noise. The algorithms successfully detected and tracked even low-SNR targets (in these particular experiments pixel SNR was -7.5dB for the 3-30-555 data set (the 128-bit data set); while it was -3dB for the sailboat1 data set (the 64-bit data set.)).