Scientific Computing Seminar

Our group's work is focused on wireless networks of video cameras. It is envisioned that such networks will replace today's antiquated and expensive camera installations used in surveillance, monitoring and traffic awareness. Low cost intelligent wireless camera nodes are readily available and can be easily networked. However, the very high data rate of video cameras presents new challenges beyond today's WSN. Transmitting all the data (even in a compressed format) would consume too much energy and limit network scalability, and performing complex computer vision algorithms locally to reduce communication would require too much computational power at each node. Accordingly, we are exploring task-driven wireless camera networks with simple local and collaborative processing.

In this talk, we discuss two aspects that are essential to the realization of such networks: Local processing and camera selection. The task we consider is occupancy reasoning, which is useful in applications such as tracking, surveillance, monitoring, counting the number of people in a room, etc. Cheap local processing that is driven by this task is explained. The camera selection problem is formulated and solved using algorithmic and analytical utility metrics, which is also driven by the task of occupancy reasoning.

Biography:

Mr. Ercan received his BS in EE from Bilkent University, Turkey, and M.S. in EE from Stanford University. He is currently a Ph.D. student at Stanford University, working under supervision of Prof. Abbas El Gamal. His research interests include wireless sensor networks, wireless camera networks, distributed and collaborative algorithms, statistical signal processing and CMOS image sensors.