Video Analysis for Nighttime Surveillance and Situational Awareness

Recent events in both Iraq and Afghanistan have demonstrated the need for credible and timely situational awareness on the ground. Of key importance is the ability to locate specific individuals of interest and to quickly provide that information to central command so that appropriate action can be taken. An automated video-based surveillance system for covert monitoring of a small area of interest would be a key asset. Our proposed system is designed for day/night stakeout scenarios and detects the arrival/departure of individuals at a given location. Pedestrians arriving/departing in vehicles are detected and tracked. People in tight groups are counted as they split to enter/exit vehicles. Timestamped images of every individual and vehicle are stored in a database. Our system also profiles individuals based on size, appearance, and motion traits, building a model for comparison with subsequent instances of individuals fitting the same model. Repeated detections matching a sufficiently unique model can provide predictions for future arrival/departure times, which is essential for human-confirmed identification and/or apprehension. Our Phase I demonstrated solutions for nighttime-specific lighting artifacts (e.g., headlight glare), Phase II will solve the remaining algorithmic problems, specify hardware, collect data, and develop & test an integrated system.