Abstract
In this paper we present a real-time traffic flow analysis approach that utilizes omnidirectional video networks. The approach includes novel means for video capture, transformation, segmentation, and classification. It also includes a graphical user interface, which allows a remote operator to view and interact with the video nodes deployed on the roadside. We describe the overall system as well as all of the associated algorithms. The system is fully developed and deployed at multiple sites on the campus of UCSD and on the Interstate 5. The paper shows the basic feasibility and promise of the approach using these nodes.