Wednesday 07 May 2014 @ 13:20
Technology Strategy Board
Technology Strategy Board
Printable version |
Sucess Story - Aurora faces new opportunity
The 327,000 British
football fans who used an app called My
Player Twin between May and July 2013 to see which Barclays
Premier League player they most resemble probably didn't think much about
the technology they were using when they pinged their photos over their
phones.
But the software that
instantly matched a player to the fan's photo came about because of a
project backed by the Technology Strategy Board to create an innovative facial
recognition system that could offer a step change in surveillance
systems.
It
did that and more: one of the by-products of the project was a new line of
business for Aurora Computer
Services. ‘My Player Twin was an unexpected spin off,'
said Hugh Carr-Archer, chief executive officer of the UK
company.
Aurora has become a market
leader in the business of providing facial recognition systems that are used to
verify a person's identity since it was founded in 1998.
Its ClockFace+ and FaceRegister systems use facial recognition
to record and verify the identity of people checking in and out at work by
matching an image on a database with the image on the camera in the entry
system.
Its
Passport ID system matches the colour photos in biometric passports with a
photo of the person trying to use it. ‘We have sold more systems of this
type to UK government and police than all of our competitors combined,'
said Hugh.
Its
current systems use an innovative technology it developed that uses infrared
flash combined with the Aurora Imaging Recognition (AIR) engine to eliminate
the problems with lighting that make matching a colour image from a camera
difficult. ‘We use colour technology to open the passport picture but
match it against the live person with a quick infrared flash which isn't
affected by lighting. So it's much quicker and much more accurate,'
said Hugh.
But
all of these systems, and the others on the market, rely on the person standing
in a particular place and looking straight at the camera. Aurora knew there was
huge potential market for a product that could match live photographs, such as
those taken by CCTV cameras, with a database of images. ‘Surveillance is
widely recognised as the "nut to crack" for facial recognition
systems. These systems could be used anywhere there is a watch list. It might
be at a football ground to pick up banned supporters or at an airport to pick
out terrorists,' said Hugh.
Seeking Smart funding
But
such systems are highly complex. Not only do they have to deal with the
variations in lighting that come with live photos, but they have to cope with
different facial angles. Aurora did not have the resources to develop such a
system so it applied for a Smart award for funding to back the
development of a prototype for a system for facial recogntion from multiple
angles (FIRMA). The Smart programme provides cutting edge innovation based
small and medium sized enterprises (SMEs) with high growth potential and
ambition like Aurora with funding awards to enable them to assess potential
markets and invest in R&D and innovation.
The
£234k award that Aurora was given in 2012 was essential. ‘We would
not have embarked upon this project without Technology Strategy Board funding
because it would have been too high risk,' said Hugh.
The
risk paid off. Aurora set out to develop a facial recognition system that could
match colour photos of faces taken at any angle, and it has done that.
‘We're really pleased. Technically we were able to achieve what we
set out to do. Now we believe we have the most accurate facial recognition
system in the world,' said Hugh.
Getting to this point did not
always go quite as planned. ‘There are three steps to creating a colour
facial recognition system. The first step is finding a frame in which there is
a face. We believed that would be relatively straightforward and it was. We
were able to do it in roughly the time we had predicted.
‘Once you find a face you
need a reference point. All facial recognition systems use two eye points; in a
millisecond the system has to find two eye points. If two eyes are looking
right at the camera, it's fine. If not, it's really, really
difficult,' said Hugh.
To
cope with different facial angles, Aurora found it needed to add more reference
points. ‘We had to identify more points like chin, and nose and jaw. It
took us about five times longer to do this than we had anticipated,‘ said
Hugh.
Finally, it had to come up with
the algorithms to recognise the same face whether it was looking up, down,
sideways or at an angle. ‘Our ambition was to test it against an
internationally recognised database and we've done that,' said
Hugh.
Suppliers of facial recognition
systems like Aurora can have their systems ranked against a database called
‘Labelled faces in the wild' by the University of Massachusetts in
the US. In its most recent test, Aurora expected its sytems to be ranked
first in the world.
New creative opportunities
But
getting the technology right was only the beginning. Aurora needed to find a
way to make its system commercial. The approach from an advertising agency that
led to My Player Twin has led to entirely new line of business. ‘We
realised it could be used for sports fans
everywhere.'
It
has also brought in revenue. ‘The income we've already derived from
it this year has allowed us to retain the three people we hired for the
project.'
And
it has inspired Aurora to set up Aurora Creative, a website
aimed at creative companies, said Hugh. ‘We realised there were several
other features of our product that might be of interest to companies trying to
be creative.
‘As a side effect of the
Technology Strategy Board product we have developed other tools so that we get
92% accuracy on gender.' (It is worth pointing out that, when tested, human
beings only achieve 94% accuracy, according to Hugh.) One use for the
technology is interactive billboards – they could display advertisements
suited to the gender of the person approaching them.
That would never have happened
without the Technology Strategy Board funding for the FIRMA project.
‘Without it we would never have been able to move into new markets,'
said Hugh.
It
has also brought the promise of business in more familiar markets. ‘We
have been asked to run a trial at Orly Airport in France in the first or second
quarter of 2014. The product we've created enables the person to be checked
against a colour photo. We needed the technology the Technology Strategy Board
allowed us to develop to recognise that colour image,' said
Hugh.