Supervised Learning for Automated Detection of Events Navy SBIR FY2012.1 Sol No.: Navy SBIR FY2012.1 Topic No.: N121-017 Topic Title: Supervised Learning for Automated Detection of Events Proposal No.: N121-017-1362 Firm: Barron Associates, Inc. 1410 Sachem Place Suite 202 Charlottesville, Virginia 22901-2496 Contact: Michael DeVore Phone: (434) 973-1215 Web Site: http://www.barron-associates.com Abstract: Barron Associates proposes to develop and demonstrate algorithms for detecting events involving one or more objects in a sensor data stream and mechanisms for post-deployment training of those algorithms by analysts. The algorithm training approach will be highly efficient, requiring minimal interaction with analysts, who need only provide an example of the desired event and possibly answer a small number of questions clarifying their intent. The resulting system will allow analysts to save and share event definitions, will run in real-time, and will be built around a service oriented architecture for ease of integration into existing and forthcoming systems. The project will leverage significant existing capabilities in the area of automated behavior analysis developed by Barron Associates. Benefits: In asymmetric conflicts, it is frequently not the class of a target that is most relevant to an analyst but its behavior and interactions with other targets. Imagery analysis activities are often highly time-critical, and failures to rapidly and accurately detect events of potential interest can severely hamper mission effectiveness. This can result in a number of problems, including failures to collect critical intelligence, failures to maintain situational awareness, failures to detect and disrupt insurgent activities, and failures to engage targets of opportunity. However, manually detecting such events, which may occur anywhere at any time, is both time consuming and cognitively demanding. As a result, the process is prone to missed detections, particularly when wide-area video and/or multiple sensor data streams must be processed. The proposed effort therefore addresses an acute need to develop adaptive tools that can automate detection and characterization of events. By doing so, it will help produce actionable intelligence that enhances the safety and effectiveness of U.S. warfighters. Additionally, the technology will be of significant value to the intelligence community, the Department of Homeland Security, law enforcement agencies, and a variety of civilian organizations. Return