VISION: Video Identification of Structures, Intentions, Objects, and Networks

Decision support tools for target identification and alert generation depend upon the classification of static and moving objects. Existing methods for object classification rely primarily on static information available in single-frame images. The data from long-range video surveillance assets are usually not sufficient to distinguish the objects from one another, and even if high-resolution data were available, the visual features alone would not allow determination of the intent or purpose of the objects that are of high interest to military intelligence analysis. Persistent video surveillance is a source of motion and temporal activity data that promises to enable automated object intent classification. Aptima proposes to develop the VISION decision support system to automatically recognize the intents and functions of potentially interdependent static and moving objects. The data currently acquired by video exploitation technologies (object detection, tracking and motion analysis, for example) will be inputs to VISION. VISION technology will be based on high-performance, empirically validated probabilistic vision-based activity recognition and temporal-relational pattern matching algorithms developed by the Aptima team.