Intelligent Retrieval of Surveillance Imagery

SET Corporation proposes to design, develop and demonstrate content-based retrieval tools and a user-friendly interface that enable users of video surveillance systems to easily conduct video-based forensics and IMINT from video imagery and display entire event histories for a given target or alerting event, regardless of whether the video data was tagged during acquisition. Our system integrates existing in house algorithms for color, appearance, face, gait, motion and shape-based retrieval of video content guided by a knowledge management architecture that allows non-monotonic reasoning and user interactions. The color and appearance based retrieval is based on a novel geometric transform; face based retrieval is implemented using probabilistic recognition of faces in videos; gait-based retrieval is implemented using a novel human gait signature for detecting the presence of humans and recognizing the activities they are involved in; motion and shape based retrieval is based on a novel application of factorization theorem. Residual analysis techniques are proposed for detecting anomalies in video. The user-interface is designed using extensions of quick-look profiles. The integrated system offers potential advantages for force protection and anti-terrorism in terms of automated alerts that improve the effectiveness of responses to threats and reduce manpower requirements for analysts.