"'Smart cameras' to tackle abandoned luggage alarms"

Friday, May 11, 2007

'Smart cameras' to tackle abandoned luggage alarms 
By Mark Trevelyan
REUTERS


LONDON - A suitcase lies abandoned in a busy airport terminal. Was it
planted by a bomber, or carelessly left for a couple of minutes while the
owner went to buy coffee? 

One of the commonest headaches facing security staff may soon be remedied
with the help of 'intelligent security cameras' developed by European
scientists. 

A newly concluded research project relies on formulae known as algorithms to
enable computers to analyse video images and spot potential threats, from
abandoned baggage to people loitering suspiciously. 

'It's (about) developing solutions so that computers can detect abnormal
behaviour,' said project coordinator Jean-Marc Suchier of French group Sagem
Defense Securite, a unit of Safran. 

For security staff at airports or railway stations, often monitoring images
from dozens of surveillance cameras at once, the new technology offers the
promise of picking out dangers that might otherwise be missed. 

'The idea is to automatically analyse and intelligently filter all of that
video, but also to add a next level of intelligence,' said James Ferryman, a
specialist in 'computational vision' at the University of Reading in
England. 

'We're talking about smart cameras which go to the next level of proactive
detection.'

THREAT SCENARIOS 

Mainly funded by the European Union, the two-year, 2.3 million euro ($3.11
million) project involved 10 European companies and research institutes and
is known as ISCAPS (Integrated Surveillance of Crowded Areas for Public
Security). 

It focused on several key threats: abandoned bags; erratic movements or
loitering by individuals; suspicious vehicle movements; and 'drop dead
scenarios' in which people fall to the ground, possibly affected by smoke or
some kind of attack. 

To define 'suspicious behaviour', the researchers conducted extensive
interviews with security experts. They used actors to play out scenarios and
mapped these into computer algorithms. 

They set up cameras to monitor public areas over weeks or months, enabling
the system to build statistical models showing busy and slack times and
typical patterns of movement - and then to spot anything that diverges from
the pattern. 

TECHNICAL HURDLES 

The scientists had to battle a host of technical challenges, including
variable weather and lighting and the need to quickly analyse dense, complex
images of people and objects. 

They stressed that while the system can flag something suspicious, it will
still fall to a human operator to make the final call - for example, whether
someone is running for a train or sprinting to escape the scene of a crime. 

'To fully detect if this person wants to catch a train or it's an anomaly,
the decision has to be taken by the person monitoring,' said Louis-Marie
Cleon, scientific director at French railways SNCF, which took part in the
research. 

Sagem's Suchier cautioned that the technology was still young and it could
take 10 years before robust systems were in place to monitor large crowds
reliably for a full range of threats, without triggering excessive false
alarms. 

But more limited solutions could be introduced much faster to the specific
problem of abandoned baggage. In a follow-up to ISCAPS, scientists will look
at how to automatically link video from all available cameras to show
immediately who left the bag and trace their subsequent movements. 

'In the coming years the technology will be able to find who is the person
who has dropped the bag and where has this person gone,' Suchier said. 'If
this works, then it will be a major benefit for the operators.'

Other companies involved in ISCAPS were BAE Systems Plc, Spain's GMV and
Italy's Elsag Datamat, part of the Finmeccanica group, together with
research institutes from France, Britain and the Netherlands.

 Do you have an opinion about this story?
Share it with other readers in our CAA Discussion Forums

http://www.californiaaviation.org/dcfp/dcboard.php


*****************************************