F3-G Distributed Video Analytics and Anomaly Detection

Abstract:This project investigates the development of automated tools for video monitoring using multiple cameras in support of security applications. The project has both experimental and theoretical components. On the experimental end, we outfitted a 2500 ft2 studio in RPI�s new EMPAC building with a permanent grid of ceiling-mounted, downward-facing cameras and a reconfigurable set of wall/ ceiling-mounted pan-tilt-zoom cameras. In this environment, we built a full-scale simulation of an airport carry-on baggage checkpoint. The movements of �passengers� through the simulation are recorded by approximately 20 video cameras. The research goal is to develop algorithms that determine which bags/ items belong to which people using automatic computer vision algorithms. We have also developed a novel theory for identifying anomalies in video, based on the temporal occupancy sequences of pixels in the video. The goal is to provide an automated capability for analyzing crowded videos. This can provide support to human operators monitoring crowds of people by focusing their attention on areas of interest containing anomalous behavior.

Faculty and Staff Currently Involved in Project:

Richard Radke
Associate Professor
Rensselaer Polytechnic Institute

Venkatesh Saligrama
Associate Professor
Boston University

Students Currently Involved in Project:

Eric Ameres, M.S.
Rensselaer Polytechnic Institute

Andrew Calcutt, M.S.
Rensselaer Polytechnic Institute

Ziyan Wu, PhD
Rensselaer Polytechnic Institute

Keri Eustis, B.A.
Asbury University

Jing Qian, Ph.D
Boston University

Erhan Ermis, PhD
Boston University