Tracking Multiple Unresolved Targets in Noisy, Cluttered Environments

ABSTRACT: This project addresses the difficult computational challenges that arise in space-based infrared object detection and tracking. We propose to implement a particular particle filtering technique which been demonstrated to robustly handle multiple, closely spaced, unresolved targets of varying intensities in high noise environments. This technique will be combined with signal processing methods specifically designed to suppress both stationary and rapidly changing clutter. The proposed technique will be implemented on massively parallel Graphics Processing Units (GPUs) to allow real-time data processing, and they will be evaluated using the Fast Line-of-Sight Imagery for Target and Exhaust Signatures (FLITES) simulation package. BENEFIT: The problem of tracking multiple poorly resolved targets in cluttered environments is extremely important and has a large number of applicationsboth military and civilian. Applications with a very high potential for commercialization include space-based infrared surveillance, surveillance and target tracking for unmanned aerial vehicles (UAVs), and video surveillance for domestic security.