Even mask-wearers can be ID'd, China facial recognition
firm says
Send a link to a friend
[March 09, 2020] By
Martin Pollard
BEIJING (Reuters) - A Chinese company says
it has developed the country's first facial recognition technology that
can identify people when they are wearing a mask, as most are these days
because of the coronavirus, and help in the fight against the disease.
China employs some of the world's most sophisticated systems of
electronic surveillance, including facial recognition.
But the coronavirus, which emerged in Hubei province late last year, has
resulted in almost everyone wearing a surgical mask outdoors in the hope
of warding off the virus - posing a particular problem for surveillance.
Now Hanwang Technology Ltd, which also goes by the English name Hanvon,
said it has come up technology that can successfully recognize people
even when they are wearing masks.
"If connected to a temperature sensor, it can measure body temperature
while identifying the person's name, and then the system would process
the result, say, if it detects a temperature over 38 degrees," Hanwang
Vice President Huang Lei told Reuters in an interview.
The Beijing-based firm said a team of 20 staff used core technology
developed over the past 10 years, a sample database of about 6 million
unmasked faces and a much smaller database of masked faces, to develop
the technology,
The team began work on the system in January, as the coronavirus
outbreak gathered pace, and began rolling it out to the market after
just a month.
It sells two main types of products that use the technology. One
performs "single channel" recognition that is best used at, for example,
entrances to office buildings.
The other, more powerful, product is a "multi-channel" recognition
system that uses "multiple surveillance cameras".
It can identify everyone in a crowd of up to 30 people "within a
second", Huang says.
"When wearing a mask, the recognition rate can reach about 95%, which
can ensure that most people can be identified," Huang said, adding the
success rate for people without mask is about 99.5%.
LOSING FACIAL INFORMATION
A big customer, not surprisingly, is the Ministry of Public Security,
which runs the police.
[to top of second column] |
A software engineer tests a facial recognition program that
identifies people when they wear a face mask at the development lab
of the Chinese electronics manufacturer Hanwang (Hanvon) Technology
in Beijing as the country is hit by an outbreak of the novel
coronavirus (COVID-19), China, March 6, 2020. Picture taken March 6,
2020. REUTERS/Thomas Peter
Using Hanwang's technology, the ministry can cross-reference images with
its own database of names and other information and then identify and
track people as they move about, Huang said.
"It can detect crime suspects, terrorists or make reports or warnings,"
he said.
But the system struggles to identify people with both a mask and
sunglasses, he said.
"In this situation, all of the key facial information is lost. In such
cases recognition is tough," Huang said.
The company has about 200 clients in Beijing using the technology,
including the police, and expect scores more across 20 provinces to
start installing it soon, Huang said.
It is not immediately clear how Chinese citizens are reacting to this
new technology.
When it comes to other surveillance tools being used in the fight
against the coronavirus, there has been some grumbling on social media
but most people seem to be accepting extra intrusion, or even embracing
it, as a means to deal with the health emergency.
Although domestic customers have been driving Hanwang's business, Huang
also said he expected more foreign interest, as the virus spreads around
the world and more people wear face masks.
"It not only benefits Chinese people, but also, when the technology is
applied globally, it can benefit the world," he said.
(Editing by Brenda Goh, Robert Birsel)
[© 2020 Thomson Reuters. All rights
reserved.] Copyright 2020 Reuters. All rights reserved. This material may not be published,
broadcast, rewritten or redistributed.
Thompson Reuters is solely responsible for this content. |