Hi-tech tracker could close net on tube terrorists

Business Weekly
Aug. 27, 2005

Two Cambridge University engineers are “bracing themselves” for a deluge of worldwide interest in a revolutionary new artificial intelligence technology that could soon help police track down purse snatchers, prevent suicides at underground train stations and save valuable time when trying to retrace the movements of wanted or missing persons.

And, it has been revealed to Business Weekly, the innovation could also prove to be a vital tool for detectives investigating the kind of terrorist attacks that recently rocked London.

Professor Roberto Cipolla and Gabriel Brostow from the University’s Department of Engineering are working on a way of teaching computers how to detect and then track individuals in a large crowd, a feat that has eluded researchers around the world up to now.

The duo have combined lateral thinking, algorithms and probability theory in developing software that they hope will eventually lend CCTV cameras a potentially life-saving degree of intuition.

The technology could partially automate the surveillance of very large crowds, alerting police or security personnel to individual instances of a particular behaviour or accurately tracing the path someone takes across a large network of cameras.

The system could be programmed to detect anti-social or criminal behaviour as soon as it has happened, for example providing an early-warning at a security control room if a fight has broken out or if one person is chasing another, an indicator perhaps that a crime has been committed.

It could also potentially save police manpower and man-hours by using the information processing power of computers to identify one person among many, removing the need to comb through hours or even days of footage.

In order for a camera to track an individual in a crowd it has to first detect it – this is where Cipolla and Brostow made their first breakthrough. The best current method can track up to 33 people and this of course is of little practical use in places like the London Underground where crowds are considerably larger.

This technology scans camera footage to record all the points in the frame where light meets dark, with clusters of these points suggestive of an individual person.

While this is a groundbreaking method of detecting each individual, the computer still cannot tell with any degree of certainty that these clusters are indeed separate people.

This is where the movement and path of these clusters is taken into account. By comparing the different trajectories of the clusters, a computer is able to make a reliable distinction between each person in the frame.

Having achieved an accurate method of individual detection, Cipolla and Brostow are now working on the business of tracking those people, a target that will open up a host of different surveillance applications.

Having accurately detected and tracked an individual on one camera, the next objective is to recognise that person again when they move into the field of vision of another camera. This, Brostow says, will be achieved by bringing colour information, i.e the colour of someone’s clothes into play. This objective is still some months off, Brostow says.

Brostow, who is funded by the UK Government as a Marshall Sherfield postdoctoral fellow, told Business Weekly: “We started this project in November last year having met with London Transport and West Anglia Great Northern Railway (Wagn), which have their own reasons to need to detect and track people in crowd situations.

“London Underground use cameras at each of their stations to watch their passengers. The cameras are filtered to some extent; if no one is moving, those cameras are not shown on the monitoring screens. Hundreds of cameras are monitored by just a few staff watching the images, as they switch from one camera to the next.

“It is impossible to have the manpower to observe all these cameras closely enough to watch for all suicide attempts, which remains a problem on the Tube.

“Approximately two thirds of suicide attempts are stopped by Underground staff. Tracking individuals more efficiently in crowd situations could improve this figure.

“Wagn, meanwhile, needs information about when people travel. Planning the number of compartments on each train and when to run trains would be more accurate if detailed pedestrian-traffic information was available.”

Brostow stresses that the project was not conceived with any Orwellian Big Brother applications in mind, but concedes that in the future the work could prove invaluable to detectives conducting forensic work.

“Once it is possible to accurately detect and track people it is then possible to program the computer to screen for instances of pre-determined patterns of out-of-the-ordinary behaviours,

“Clearly this is only of use in detecting crimes if someone is displaying such behaviours. Most suicide bombers, for example will not,” he said.

“Where it will prove useful is in helping police track a criminal’s movements from camera footage, a vital part of the investigations into July‘s terrorist attacks in London.”

Having exclusively unveiled this technology in Business Weekly, Brostow says he is expecting organisations from around the world to beat a path to his door

“This is the first time we have made this work public and we are now bracing ourselves for the expressions of interest,” he said.