Is facial recognition technology the end of the security guard?
When a three-year-old was recently kidnapped in Shenzhen, China, it took a matter of hours for the offender to be tracked down and the child to be reunited with his parents. Unknown to the kidnapper, her journey to a town more than 1,000km away was being automatically tracked by "intelligent" cameras.
"Through a facial-recognition video system the police could quickly identify the three-year-old and the kidnapper," Huawei Enterprise global public safety expert Peter Goulding says.
"The kidnapper was identified in another part of the city, and we then saw a video of her in a building collecting a bag.
"We picked up another video of her going to a railway station, where she checked through with her ID," Goulding says.
By the time the kidnapper arrived in a town about 2½ hours’ flying distance away, police were waiting to arrest her.
"It was all done by city-wide closed-circuit TV (CCTV) using facial recognition. It’s about prevention and also about prosecuting and finding offenders and victims," Goulding says.
Facial recognition technology is ubiquitous in China, which monitors its citizens closely.
Now, in SA, a growing number of companies are turning to facial recognition technology to secure their premises and stamp out crime, says Colleen Glaeser, MD for Southern Africa and the SADC region at AxxonSoft.
Moscow-headquartered AxxonSoft sells its facial recognition software to shopping malls and housing estates, and to airport, mining, banking and transport operators.
Glaeser says it’s a modern substitute for private security guards, who account for large chunks of companies’ costs but are often an impotent solution.
"The private sector is spending huge amounts of money on the employment of guards and on security. Companies could use that money to create other jobs — rather than spending R200,000/month on security, they could employ more staff."
By applying artificial intelligence to its growing database of images, the company can automatically notify a shopping centre when, for instance, a recognised criminal is shown on a video screen entering the premises.
The retail sector is a prime target market, as crime syndicates tend to hop from mall to mall across the country, says Glaeser.
AxxonSoft has run a facial recognition pilot programme from a facility in the KwaZulu Natal Midlands that monitors 3,500 cameras across SA, some of which are in SA’s larger shopping malls.
In partnership with a JSE-listed company whose name Glaeser says she is not at liberty to disclose, the firm is moving its "incident management centre" to a full-scale facility in Centurion.
It will monitor about 60,000 cameras that identify faces and number plates, she says. "AxxonSoft has implemented more than 150 ‘safe city’ projects globally, and this is a private-sector replication of that model."
Glaeser says that while facial recognition technology is used mainly to curb crime, fitness chains and other private sector operators are interested in using it for access-control purposes.
AxxonSoft applies its software to existing CCTV cameras that are provided by Huawei, Samsung, Bosch, Hikvision, Dahua and other technology companies.
According to Goulding, Huawei’s video surveillance hardware is also used in Kenyan cities.
The company is working with SA cities and government departments on "safe city solutions" that encompass crime, crisis and disaster management, Goulding says.
Meanwhile, Glaeser says AxxonSoft has also developed software to reduce theft in the retail sector.
If security guards who man storefronts and patrol aisles can be seen as the last line of defence that retailers have in their fight against shoplifting, cashiers are the second last. The problem, says Glaeser, is that this safety net is porous: retailers bleed money through their tills because tellers could pass goods through the point of sale without scanning them.
Some cashiers work with friends who pose as customers to siphon goods out of stores, and there’s little that retailers can do because their security cameras can’t tell that the cashier is faking a scan, she says.
A cashier might, for example, hide the barcode on an item while it is being swiped.
AxxonSoft has developed an algorithm that prevents "sweethearting", or theft by employees who work at cash registers.
The algorithm allows a camera to "continuously record a cashier’s workspace and combine this information with point-of-sale transaction data to prevent fraud or theft by the cashier or customer", Glaeser says.
While some retailers employ security personnel to check customers’ till slips when they exit, Glaeser says these staff members are sometimes in on the crime or fail to notice when slips do not reconcile with the contents of shopping bags
Since the product was launched nine months ago, "arrest rates have been extremely high," says Glaeser.