Watching live CCTV footage of thousands of people, trying to pick out any sort of noteworthy activity... it sounds like a very tedious, difficult job for a human being. That’s why researchers from Germany’s Fraunhofer Institute for Applied Information Technology are working on an electronic system that uses the principles of human motion vision to do the same job. It is part of the EU’s SEARISE project, which stands for Smart Eyes: Attending and Recognizing Instances of Salient Events.
Fraunhofer’s Smart Eyes system consists of one fixed camera, two active stereo cameras, and software modeled after the visual processing system of the human brain. The system ascertains the average degree of movement for each pixel in the fixed camera’s shot, and stores those motion patterns as typical models. It can then identify and classify certain patterns within those models, such as people waving flags in a soccer stadium. These patterns can be noted, or the system can even be programmed to ignore them.
More importantly, Smart Eyes notices when pixels are moving atypically. This tells the system that something unusual is going on, at which point the other two cameras zoom in to get a closer look.
The developers of the system state that it is superior to human observers because it can simultaneously monitor every part of a large scene (such as a stadium full of soccer fans), and it never gets tired or distracted. It can reportedly be added to existing security camera systems with no adjustments.
A somewhat similar system, ISIS, is currently being developed in the UK.
