Blog: Tracking progresses far beyond facial recognition - Mobile World Live

Blog: Tracking progresses far beyond facial recognition

29 MAR 2017

Hitachi this week announced it developed detection and tracking technology using artificial intelligence (AI) which can recognise an individual in real time using features from more than 100 categories of external characteristics including age, hair style, skin colour and type of clothes.

The Japan-based company said the technology can “immediately detect and track the sought after individual” by analysing the entire aspect of a person detected using camera images, even where only the rear-view is captured and the face cannot be seen, or the person is at a distance.

Hitachi’s timing is uncanny, coming on the same day as reports photos of about half of all American adults are stored in facial recognition databases which can be accessed by the FBI without their knowledge. The pictures are gathered from databases of drivers’ licenses and passports around the country, with only about 20 per cent of those in the photos identified as criminals, according to The Guardian.

The US Government Accountability Office (GAO) analysed the FBI’s use of facial recognition technology in 2016 and found the agency lacking in accountability, accuracy and oversight, and made recommendations on how to address the problems.

Accuracy issue
The algorithms used by the US government to identify matches are said to be inaccurate about 15 per cent of the time and are five years behind what companies in the commercial space are using now, The Guardian said.

There is certainly little doubt of the lag. Hitachi said it is now possible to detect a suspicious individual or a lost child by gathering information from eye-witness accounts and then finding the person fitting the description using images from public security cameras. The company is applying the technology to wide area security and surveillance systems in public areas, such as airports, bus and train stations.

Since it is impossible to screen all images with a limited number of staff, technology to detect and track people was developed using facial images or other physical features such as colour of clothing gathered at the point of entry. However, it is a challenge to detect and track a person if there were limited eye-witness accounts, other people wearing similar colour clothing, or if the surveillance camera was unable to capture a clear facial image due to the angle or lighting.

To overcome this, Hitachi tapped AI to identify and track an individual in real time using footage from wide-area surveillance cameras. AI can recognise an individual as the same person by analysing and converting the image of the entire body into a numerical description so the person can be identified regardless of any changes in body angle or lighting, the company said.

By recording the result of the AI analysis in a database with high speed vector processing developed by Hitachi, an image of the person can be extracted from among several tens of thousands of recorded images in less than 1 second.

It is a huge advancement and certainly an amazing feat. But the question remains how private firms will coordinate with government authorities to share the information and if that sharing will be limited to actual immediate security and safety concerns such as a lost child or suspicious individual, or merely gathering data wholesale for use in the future.

The editorial views expressed in this article are solely those of the author and will not necessarily reflect the views of the GSMA, its Members or Associate Members.


Joseph Waring

Joseph Waring joins Mobile World Live as the Asia editor for its new Asia channel. Before joining the GSMA, Joseph was group editor for Telecom Asia for more than ten years. In addition to writing features, news and blogs, he...

Read more