The software could turn the thousands of CCTV systems installed around the UK from passive recorders into active preventers of crime. It would also reduce the chances of events being missed by tired or bored security staff.
But privacy advocates warn that it could be another step towards "total control society", in which acceptable behaviour is decided by machines.
One research prototype of the software, designed by Steve Maybank of the University of Reading and David Hogg of the University of Leeds, takes the "pixel" data which makes up the image from a TV camera, and analyses its contents and patterns of movement.
The software can identify long, low objects in pictures as cars, and tall, thin moving ones as people. A second element of the software can analyse the people's patterns of movement.
Normal car owners approach their car directly and get in with little delay. Would-be thieves tend to be more wary. The system can pick this out, and sound an alarm at the monitor station for the CCTV."It's unusual for someone [innocent] to hang around cars,"Dr Maybank told the New Scientist. He said deceiving the software was surprisingly difficult.
A similar product, developed separately at King's College London and being tested by London Underground, is able to identify potentially dangerous crowding on stations. "Cromatica" only raised a false alarm in 1 per cent tests.
However, Simon Davies of Privacy International, a human rights group, said in New Scientist : "The system reflects the views of those using it. Anyone who does act out of the ordinary will be more likely than now to be approached by security guards, which will put pressure on them to avoid standing out."Reuse content