CAREER: Distributed Active Processing on Hundred-Camera Networks (DAPHNE)

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

It is proposed to develop a new framework for the distributed processing of images from networks of literally hundreds of cameras, for which no calibration is known in advance and in which cameras may freely appear, disappear, and move about. Conventional computer vision methods are not adequate for such networks, because they assume a small, fixed number of stationary cameras that can be centrally, simultaneously processed. We will study distributed solutions for both calibration (computing absolute and relative camera positions) and computer vision problems (such as view synthesis) in large dynamic camera networks. This research has broader impacts in the design and effectiveness of the camera networks that will be ubiquitous in the city streets, airports, and battlefields of the 21st century. Such networks will be crucial for automatically monitoring and extracting information from environments where the safety and security of human lives are at stake. Our goal is to develop the models, algorithms, and applications that will form the foundations of the new field of dynamic camera networks.