Algorithm Development for WFOV Mission Data Processing

ABSTRACT: Toyon Research Corporation is proposing to develop and demonstrate the feasibility of a suite of real-time algorithms for front-end exceedance generation for WFOV OPIR data. Persistent monitoring from infrared surveillance and staring systems is necessary for battlespace awareness and technical intelligence. However, persistent OPIR surveillance produces massive amounts of WFOV image data. To detect and track events of interest to the Air Force the data must first be pre-processed to remove focal plane array (FPA) noise, remove sensor motion jitter, remove background clutter, and reduce false alarms. Pre-processing is necessary for both automated and/or man-in-the-loop analysis methods. Additionally, it is vital that pre-processing be performed in real-time to enable the Air Force to respond to potential threats. Large-format data collected for persistent WFOV applications poses a greater challenge for real-time processing. Toyons proposed algorithm suite includes novel, GPU accelerated, implementations for jitter suppression, clutter suppression, noise suppression, exceedance formatting, image windowing, and buffering and thresholding. The accelerated algorithms take advantage of parallelizable components in each process. To further increase efficiency, improvements in feature detection and tracking algorithms that make use of OPIR data constraints and accelerated versions of state-of-practice clutter suppression routines are proposed.; BENEFIT: The successful completion of this research and development effort will result in advanced, real-time algorithms for front-end exceedance generation of large-format image data. A direct application will be the insertion of these algorithms within baseline OPIR software for real-time processing at Air Force ground stations (such as Space and Missile Systems Center's Falcon Shield algorithm research facility), or technical intelligence labs (such as NASIC). The benefits of the proposed R&D effort would be a comprehensive pre-processing algorithm suite which would increase the efficiency of detecting and tracking targets of interest as well as shorten the timeline needed for detecting and reacting to threats observed with OPIR sensors. This software could also be optimized for implementation by other agencies that use satellite sensors for surveillance to detect and locate new classes of potential threats in real-time. Furthermore, the algorithms described have the potential to be used onboard future platforms carrying SWIR sensors for persistent surveillance.