Video Surveillance System in Street Environment
Using Heterogeneous Cameras

Sponsor: National Science Council (NSC), Taiwan
Time: 2008/8/1 to 2009/7/31
Grant No: NSC 97-2221-E-147-005
Budget: 418,000

Project Descriptor

An Image Annotation Tool with Relevance Feedbacks
There exist a huge number of cameras for their different goals in streets of a city. Our goal is to integrate these cameras, not to construct a new and perfect system, to build a surveillance system. That does not only reduce the cost of the video surveillance system but also make the system more flexible in the insertion or deletion of cameras. This project is based on a city map to integrate cameras located in the map area. First, we apply the calibration for different cameras to get their intrinsic parameters. We then solve the localization problem to compute the position in the city map of a point in the camera image plane. We design a mapping between the city map coordinate and image coordinate of cameras. The mapping aims to transform the position coordinate from the subject tracking in a camera to the presentation in the city map. Moreover, we design a tracking method crossing multiple cameras that may not have overlapping views. Finally, the tracking path of a subject can be visualized in the city map for surveillance.
This project is cooperated with Prof Wee Kheng Leow, Department of Computer Science, National University of Singapore.

Members

Publications

  1. Wee Kheng Leow, Cheng-Chieh Chiang, and Yi-Ping Hung, "Localization and Mapping of Surveillance Cameras in City Map", in Proceedings of ACM International Conference on Multimedia, ACMMM, Vancouver, Canada, 2008.