Georegistration of Imagery with Target Tracking

Dynamic Scene Analysis - a real time implementation of an image registration, target detection, tracking, and geo-location software product.  Registration is based upon Mutual Information, a more efficient and robust technique than traditional feature-based methods.  Reference imagery and live imagery are utilized as needed to refine the estimate of the sensor position and attitude in the geo-registration solution.  The algorithm automatically and adaptively performs the required computations to maintain the desired level of accuracy.  Moving ground targets are automatically detected using an algorithm based on Optical Flow; it is ideally suited to separate object motion from perspective-generated scene motion, greatly reducing false alarms.  Tracks are formed in geo-registered coordinates using an Associative Tracker, enabling tracking of irregularly moving or intermittently visible targets and fusion of target information from multiple sources.  Image Supplemented Tracking further distinguishes targets moving close together.  Tracking algorithms are adaptively executed as needed to maintain accurate track while minimizing computational burden.  The efficiency, flexibility, and adaptability of our processing stream lends well to implementation in a variety of computing environments. BENEFIT: This research will create a system that automatically detects, tracks, and geo-locates ground targets from airborne imagery at video rate, and determines the uncertainty of geo-locations.  The processing is efficient and intended to be run on board the aircraft.