Algorithms for IR data

Persistent, autonomous monitoring from infrared surveillance and staring systems is necessary for early missile warning/defense, battlespace awareness, and technical intelligence. Specifically, the ability to accurately detect, track and geo-locate events of interest in known hostile regions enables the Air Force to provide countermeasures to potential threats from adversaries. Toyon Research Corporation is proposing development and feasibility demonstration of advanced real-time algorithms for exploitation of high-frame-rate overhead persistent infrared (OPIR) imagery. The algorithms being developed are for the purpose of detecting and geo-locating dim targets (due to system noise and background clutter), both moving and static, with a variety of potential signatures. The geo-location algorithms are based on the coupling of image registration to external geo-referenced satellite maps, frame-to-frame registration using feature point detection and tracking, and sensor orientation bias estimation. The algorithms for dim target detection are implemented via a nonlinear Track-before-Detect (TrbD) particle filter and are designed to work in conjunction with statistical clutter estimation and rejection algorithms to near-optimally integrate information to provide improved signal-to-noise ratio (SNR) at the detection stage. A variety of target/event signatures are modeled, and fusion of external geo-spatial information (if available) provides reduced false alarm rates over the entire field of view.