[Reader-list] the Next Step in Surveillance Video ?

[Shuddhabrata Sengupta]

Dear All,

Here is a link to a story in the New Scientist that reports new 
developments in the intelligent analysis of surveillance video. The 
next step, so to speak, after face recognition. Notice how the 
'Simhastha Kumbha Mela" is cited as a likely site of the immediate 
implementation of this technology.

This story was forwarded by Coco Fusco on the 'Undercurrents List'.

Shuddha


http://www.newscientist.com/news/print.jsp?id=ns99993918
Smart software linked to CCTV can spot dubious behaviour


19:00 11 July 03
Jenny Hogan


It could be the dawn of a new era in surveillance. For the first 
time, smart software will help CCTV operators spot any abnormal 
behaviour.

If the trial due to go live in two London Underground stations this 
week is a success, it could accelerate the adoption of the technology 
around the world. The software, which analyses CCTV footage, could 
help spot suicide attempts, overcrowding, suspect packages and 
trespassers. The hope is that by automating the prediction or 
detection of such events security staff, who often have as many as 60 
cameras to monitor simultaneously, can reach the scene in time to 
prevent a potential tragedy.

If the technology takes off it could put an end to a long-standing 
problem that has dogged CCTV almost from the beginning. It is simple: 
there are too many cameras and too few pairs of eyes to keep track of 
them. With more than a million CCTV cameras in the UK alone, they are 
becoming increasingly difficult to manage. And besides being a 
tedious task, watching TV monitors also demands a higher level of 
concentration than many people can manage.

The new software, called the Intelligent Pedestrian Surveillance 
system (IPS), could change all that, says London Underground's Peter 
Tollington, who is overseeing the installation at the stations. "CCTV 
itself is fantastic, but this puts it on another plane. It means that 
you don't have to look at a screen all of the time," he says. Indeed 
it seems likely this is just the beginning. Other companies are 
working on similar technologies.


Behavioural oddities


Not that long ago similar claims were made for face recognition 
software, which has still to make its mark, particularly in crime 
prevention. But the computer scientist behind IPS is convinced that 
his software will not suffer a similar fate.

Sergio Velastin at Kingston University, London, has spent a decade 
developing his algorithms (New Scientist print edition, 11 December 
1999). The trick, he says, is to keep it simple. Face recognition 
software by its very nature involves complex image analysis and tries 
to do too much. As a result, operators are often disappointed. Others 
have argued that recognising faces on CCTV of known offenders, for 
example, is only of limited use. You can't arrest people just for 
being caught on camera.

Velastin's software compares the images from each CCTV camera with 
pictures of the empty station, analysing each pixel to pick out those 
that differ between the two images. Clusters of pixels which could be 
objects or people are tracked in real time.

For example, a blob that remains on the platform even after several 
trains have passed could be a loiterer, or someone who is lost. It 
doesn't matter that the system cannot distinguish between the 
possibilities, as long as someone is alerted who can investigate. Its 
job is not to make evaluations but merely to draw attention to 
behavioural oddities.

Similarly, when people are building up the courage to attempt suicide 
on the tracks they will often let several trains pass before actually 
jumping. By spotting this behaviour it should be possible to dispatch 
someone to the platform or warn approaching trains.


Million pilgrims


"I think CCTV can be tremendously effective in improving health and 
safety and reducing crime. But it can be even more effective if we 
can use technology to supplement the operators," says Adam Wiseberg, 
chairman of the CCTV group at the British Security Industry 
Association.

IPS was developed using actors to simulate passenger behaviour in 
London Underground stations. It was the product of 10 years of 
research and collaboration between six European countries, in a 
European Union funded project. Now Velastin has set up a company, 
Intelligent Pedestrian Surveillance and Observation Technologies 
(IPSOTEK), to market the technology.

Already, he says, organisers of the Simhastha Kumbh Mela festival in 
India have contacted him. They want to use CCTV with automatic 
surveillance to help manage congestion and prevent people being 
crushed or injured next spring when more than a million pilgrims are 
expected to descend on the town of Ujjain in the state of Madhya 
Pradesh. Velastin thinks it will also be useful in art galleries, 
schools and public spaces.


Invasion of privacy


But the two London Underground stations that have installed the IPS 
software - Liverpool Street and Mile End - will be the first to test 
it for real. When it is switched on Mike Chatterton of Capital 
Projects Consultancy will be gauging the response of the people using 
it. Over the next six months, he will evaluate the system with staff 
in the two stations' control rooms.

Customers in the two stations have not been told about the project, 
says Tollington, because "we want to get the technology right before 
we make any song or dance about its effectiveness". Although some 
campaign groups see CCTV as an invasion of privacy, customers are 
likely to welcome technology that makes the stations safer, says Ian 
Brown, from the Foundation for Information Policy Research, a 
London-based IT think tank .

The total number of CCTV cameras in Underground stations across 
London is set to increase from 6000 to 9000 over the next few years, 
and they are also being installed on trains. That will make the need 
for a system like IPS all the more acute.

This pattern is being repeated in cities around the world, says Joe 
Freeman, who runs a consulting company for the security industry in 
Newtown, Connecticut. In a survey of hundreds of US security 
executives, he found that systems which could process the video from 
the spiralling number of