| Author: | Daniel Gutchess |
| Advisor: | Dr. Anil Jain |
| Email: | gutchess@cse.msu.edu; http://www.cse.msu.edu/~gutchess |
Video surveillance systems have become commonplace in modern society; they can be found in most retail stores, banks, and airports. Current systems have two major drawbacks: archival systems cannot issue alarms in real-time, and manual monitoring systems are costly in terms of manpower. These flaws would be overcome by a real-time system with the ability to automatically analyze digital video, recognizing people's identities and actions. We are developing a real-time automated surveillance system, which uses an omnidirectional video camera in combination with multiple active cameras. Tracking of multiple subjects in a scene is performed using omnidirectional video as input. While a relatively low image resolution makes omnidirectional video unsuitable for recognition of subjects, the wide field of view makes it well suited for the location and tracking of subjects. The world coordinates of each subject in the room are estimated in order to direct the attention of one or more pan-tilt-zoom cameras. The system automatically controls these cameras for the purpose of obtaining high-resolution images and video sequences of subjects. The system outputs trajectories of each subject onto a 2D map in real-time. Off-line processing is performed to detect all faces in the video.
Return to Workshop Listing