Homeland Security Project


Abandoned object detection

Not detecting people as objects

This project for the Department of Homeland Security set out to develop a real-time, vision-based, human behavior monitoring system to automatically watch for suspicious terrorist-related behaviors in a very large urban transportation center. This is a very large project, with many people working on different aspects of the final system. The system is comprised of over 100 cameras and monitor for over a dozen specific behaviors.

I personally developed several modules for this project. One of these was to detect abandoned objects in areas where people could sit for a long time. Another detected when people crossed a virtual perimeter in one direction only. In addition to these, I worked on a module to detect loitering individuals and a module to analyze crowd behaviors in aggregate.


  • Nathaniel Bird, Stefan Atev, Nicola Caramelli, Robert Martin and Nikolaos Papanikolopoulos, "Real-Time, Online Detection of Abandoned Objects in Public Areas," Proceedings of the IEEE International Conference on Robotics and Automation, pp. 3775-3780, May 2006.