Intel Systems & Networking Laboratory

People:

John Kassebaum, Nirupama Bulusu, Wu-chi Feng

In order for smart camera networks to perform vision-based tasks, such as subject detection and tracking, each camera’s position and orientation relative to a 3D coordinate frame must be accurately determined. Manually measuring positions and orientations is prone to error, and is impractical in large networks that cover wide areas. Researchers have proposed methods to compute camera positions and orientations based on a camera’s available visual data, but these methods are biased towards deployments with large view overlaps and incur high computation and communication costs that are problematic for battery-powered, processing-constrained smart camera networks. Also, these methods only compute camera positions to within an unknown scale factor of a real-world coordinate frame. If we show the cameras a feature-point rich 3D target, though, a cheaper and more efficient method can be used to determine camera positions and orientations, and that gives positions relative to a real-world coordinate frame while also reducing overlap constraints. This new method is not only available for resource-constrained networks, but is a viable option for any smart camera network.

Publications:

John Kassebaum, Nirupama Bulusu, and Wu-chi Feng, "Smart Camera Network Localization Using a 3D Target," ACM SenSys Workshop on Applications, Systems, and Algorithms for Image Sensing, 2008. (PDF)

Sponsored by:

This material is based upon work supported by the National Science Foundation under CISE grants 0514818 and 0722063. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).