CHARLES RIVER ANALYTICS, INC. — Department of Defense SBIR Phase I: A19-158

Description

Advances in inexpensive COTS cameras have made them critical to distributed security surveillance at Army installations; however, the result has been an overwhelming amount of visual data that must be processed by human operators to distinguish possible threats. The current Installation Notification and Warning System (INWS) attempts to reduce this overload, but generates too many automated cues for suspicious objects or behaviors, subjecting operators to alarm fatigue. To address these challenges, Charles River Analytics proposes to design and demonstrate the feasibility of a Criminal and Terrorist Indicator and Alarm System (CRITICAL) that leverages computer vision and deep learning techniques along with pattern of life analysis algorithms to detect suspicious and anomalous objects, people, and behaviors with a low rate of false alarms. CRITICAL is designed to be implemented easily within existing distributed surveillance installations, and can accommodate any number of cameras with a simple setup process. CRITICAL distills the vast amount of visual data from surveillance into targeted, actionable alerts without requiring operators to be system experts.