The future of CCTV is much more than looking for shoplifters, or even tracking all cars, at any time, in the UK with the National ANPR.
The future of CCTV will allow integrated systems to monitor and track people through behaviour patter. Initially the behaviours they are looking for relate to crime, violence and disorder, however this will move to crowding (and already has).
The question is, will it move to issues such as meetings of more than two? People engaged in heated debate? People speaking out against topics? (recently a person was summonsed for referring to Scientology as a “cult”).
Much of the European effort to develop intelligent surveillance stems from concerns with safety and efficiency on public transport. Closed-circuit television (CCTV) already plays a large part in airport terminals and metro stations. The general idea, then, is to capture video data from these cameras and program a computer to analyze it. A recent conference at the Institute of Electrical Engineers in London was dedicated to this task.
Francois Bremond from France’s INRIA (Institut National de Recherche en Informatique et en Automatique) described a system they have developed for metros in Barcelona. According to Bremond, this is capable of detecting fighting, vandalism, blocking of exits, overcrowding and fraud (i.e. jumping over the ticket barrier).
It achieves all this by using a mixture of techniques from artificial intelligence: Bayesian Networks and AND/OR trees. Data frames from multiple cameras are combined to form 3D images. These are then scanned to identify semantic classes like PERSON, OCCLUDED PERSON, GROUP, CROWD, TRAIN, SCENE OBJECT and NOISE or UNKNOWN. A tracker module can identify different kinds of motion and hence behavior.
Recent tests on camera data from Brussels and Barcelona revealed 100 percent detection on fraud, vandalism, blocking and overcrowding. Fighting was detected 20 times out of 24. “The next step consists in designing the video interpretation system to be operational (able to cope with any unpredicted real world event) and working on a large scale,” says Bremond.