SAR Contextual Target Tracking

Currently deployed radar systems have limited capability to maintain tracks on targets in a dense target environment. Dropped tracks and track switching difficulties often arise as a result of infrequent track updates, inadequate resolution, inaccurately extrapolated tracks and detection position errors that greatly exceed the distance between the target of interest and other vehicles in close proximity. Terrain cover and road information automatically derived from coregistered SAR, however, has the potential to dramatically reduce these problems. TSC proposes to implement and demonstrate techniques that exploit contextual information, automatically extracted from SAR imagery, to aid in tracking ground moving targets. TSC will: 1) execute existing software to automatically classify SAR images, 2) develop and test road identification algorithms, 3) implement a contextual tracker, 4) refine the techniques, quantify their performance improvement and 5) demonstrate them in key situations of conflict. TSC will leverage off existing SAR imagery, automatic terrain classification software developed for DARPA, SAR and MTI experience gained from numerous radar efforts, and extensive knowledge-based engineering and NCTI software and techniques. Existing SAR, together with simulated MTI data as well as co-registered SAR/MTI data obtained from either Northrop Grumman Joint STARS or Hughes Aircraft Co., will be employed. The contextual tracking algorithms to be developed on this effort have an immediate DoD application to time critical target tracking in several existing radar surveillance systems. Commercial applications of the technology include both air- and ship-tracking by ground-, air-, and space-based sensors. Applications of terrain classification and road identification algorithms include city planning, forest monitoring, and resource exploration.