Published on 10 March, 2012
|The facial recognition algorithm proved its worth during trial runs of the system
Boulvar is the largest café in Petropavlovsk-Kamchatsky, which is the regional hub of Kamchatka Peninsula and located in the easternmost reaches of Russia. The café is split into four distinct zones: a romantic zone (also referred to as the coffeehouse), an Italian zone, a Japanese zone, and a bar. As any restaurant or café grows and flourishes, it accumulates regular customers who require particular attention and a personal approach. By the same token, there may be unwelcome patrons best kept at arm’s length.
At Boulvar, the owners installed a system to automatically recognise faces in real time and classify incoming visitors. They started with two basic categories:
- "Unwelcome" – deadbeat diners who skipped out on their bill or behaved disruptively
- VIP – regular visitors and others requiring special care and service
Automation needed to enable timely decision-making: for example, signalling a waiter to escort a client to their favourite table.
The facial recognition system
needed to analyse the incoming
video stream and automatically
capture facial images, comparing
these with the photographs
in its database
The task from the very start was two-fold, as well as tricky: Unimax was tasked with installing an ambitious project, from an engineering point of view, while minimising system start-up costs. Due to Kamchatka’s distance from Moscow – nine time zones – installation works were assigned to local companies in Petropavlovsk-Kamchatsky. Thus Unimax was responsible for project design, manufacture and supply of equipment, consulting, and technical support.
The facial recognition system needed to analyse the incoming video stream and automatically capture facial images, comparing these with the photographs in its database. The system should then alert the operator after a match is found. Both administrators and security staff would be able to work more effectively, thanks to the database of VIP clients as well as deadbeat clients and hooligans.
Since the client’s selected contractor was inexperienced with such systems, Unimax developed a detailed technical solution and provided all necessary consultations. It also created guides for configuring cameras to work with the automatic facial recognition system and for writing Face Intellect scripts.
At the core of the system is Axxon Intellect Enterprise platform and the Face Intellect facial recognition module. To complement the VIDEOMAX-IP platform from Videomax, the AXIS P3346 – the flagship of the Axis lineup of professional fixed-dome cameras – was chosen. With 3 megapixels of resolution, high sensitivity, manual shutter controls, and advanced focusing algorithms, the Axis camera offered the consistent high image quality required for recognition.
The facial recognition algorithm proved its worth during trial runs of the system. The database began to store photographs of unwelcome clients who did not pay their bills. More recently, the client has decided to supplement the system with additional cameras and look into using panoramic cameras for facial recognition.