Multicamera surveillance automated

Imagine a bank of cameras trained at different angles on many objects, such as people at an airport. It takes time for an operator to look at the images in each camera, see if a particular object of interest is in that image, then manually zoom in to take a closer look.

Researchers from the University of Toronto have developed a system that allows the user to indicate an object in one view and automatically zoom to that object in all other views. The system sorts all the camera views according to the quality of the view and zooms into all the images that contain the object in about half a second in a system containing as many as 200 cameras, according to the researchers.

The system can be used anywhere where multiple cameras are used for surveillance, such as airports or casinos. It can also be used to speed the process of parsing through a large number of still pictures of a scene, such as archeological and crime scene pictures.

The researchers are working on adding real-time object tracking to the system so it could to follow an object chosen by the user.

The system coordinates the multiple views using the three-dimensional spatial location of the object of interest, which it finds using the visual input plus sound localization information from an array of microphones.

The basic method is ready for practical applications now; real-time object tracking will be ready for commercial use in about a year, according to the researchers. The work appeared in the proceedings of the 2004 IEEE International Conference on Systems, Man and Cybernetics held in October in behavior, Netherlands.