“We try to explore how you can get information from a lot of sensors,” says Professor Parham Aarabi, of the Artificial Perception Lab he manages at U of T’s Department of Electrical and Computer Engineering. A large touch screen at the far end of a room in the Bahen Centre lab is the most visible part of the Artificial Perception Lab’s latest offering-a new 3D camera system.
The system can monitor a large number of cameras-one or two hundred-and understand some of what it sees. Starting from a large image that shows many people, operators can click on a particular person and let the computer sort out which camera angle will capture the clearest image of his or her face. This eliminates the difficult and tedious job of going through hundreds of slightly different shots one at a time. With this system, a set of security cameras that now requires ten operators could be run by only two or three.
The team that developed the system was led by graduate student Sam Mavandadi and Professor Aarabi. Ideas for the project were part of the proceedings of the 2004 Institute of Electrical and Electronics Engineers (IEEE) International Conference on Systems, Man and Cybernetics, but more specific details have not yet been published.
Aarabi says that the system is particularly novel in two ways: first, the way it ranks images from best to worst, based on what the operator wants to look at; and second, the interface, where more relevant images are larger. It appears quite simple to use.
Aarabi and Mavandadi don’t seem concerned that their invention might be misused.
“There are situations in which having cameras isn’t a bad thing,” says Aarabi, citing TTC platforms at night and airports as examples. “Any technology could be used out of context,” he adds.
“You are reducing the number of people…who see this,” says Mavandadi, suggesting that by decreasing the number of operators who work with the images, you might also decrease the chance that the images are misused.
The system is only a prototype now, but it will be commercialized. Exactly how hasn’t been decided, though it will probably involve starting a new company. The new systems will not cost more than those being used now, because they require fewer operators.
