The objective of the proposed program is to develop a long term image-based tracking camera system. Our goal is to allow fewer personnel to analyze and respond to threats detected in aerial video, making large format video useful in real time analysis. We propose to exploit long term contextual information gathered over long periods of time over wide areas to improve tracking performance. Further, we provide identification and classification of thousands of simultaneous targets using both tracking information and previously gathered identity data. In the first phase, we propose to study three algorithms short term tracking (0-30 seconds), medium term tracking (30 seconds 1 hour), and long term tracking and identification (1 hour 24 hours). ArguSight personnel have significant experience in large format imagery and video analysis, and are well qualified to address all parts of the work effort. We will be working with subcontractor Carnegie Mellon University (CMU, a premier computer vision research center.
Keywords: Gigapixel Video Camera, Large Format Video, Context Based Tracking, Tracking Through Occlusion, Gis Integration, Computer Vision, Automated Target Recognition
