CCTV cameras could soon be able to spot any suspicious behaviour, thanks to smart software that automatically detects unusual activity.
The statistical approach used should mean the technology can be used in many more situations than current systems, which are usually designed to detect specific objects or events.
The new system, developed by Roke Manor Research in Hampshire, UK, aims to overcome a common problem with CCTV - having too many cameras and not enough pairs of eyes to watch them all. Roke Manor estimate there are approximately 25 million CCTV cameras in the world.
Their Video Motion Anomaly Detection (VMAD) system will help by alerting a security operator when something unusual is happening on screen. This would include, for example, someone running through an airport or trying to enter a restricted area.It might even be able spot when a baggage handler was interfering with a passenger's bag.
Rather than attempting the difficult task of recognising specific objects and their particular behaviours, VMAD simply looks at how groups of pixels move within the image. It uses these movements to construct a statistical model of "normal" activity. Movements that are out of the ordinary can then be flagged up.
Existing "intelligent CCTV" systems tend to be more limited in their applications. For example, a system currently being tested at London Underground train stations also monitors pixel behaviour, but only flags up a lack of activity, to indicate crowd congestion or loitering.However, statistical systems do also have drawbacks. A key one is that by not identifying specific objects, they cannot account for the context of an unusual behaviour. For example, lots of people wearing bright clothing might set off an alarm, whereas someone entering an airport with a rocket launcher slung over their shoulder might be ignored