Facial recognition technology can provide a streamlined contactless approach to on-site attendance tracking, says Chris Coughlin.
How does it work?
Facial recognition analyses live images of faces using AI to match a person’s face to facial images stored in a database. It works in two parts:
- Creation of a database of facial images, usually built with passport-style photos and using AI. A biometric map is created for each facial image and marks out specific features, such as size of eyes and length of nose to build a profile that’s specific to that person.
- A camera captures live images of faces. Using the AI, the same markers on a ‘live’ image of a face are cross-referenced against the database to find a match.
Why use it?
Manually tracking employees’ attendance and visitors’ whereabouts on site tskes time and lacks immediacy and accuracy. Electronic access control, visitor management, and time and attendance systems are used in varying degrees to manage this.
These systems use facial recognition technology to give a contactless approach.
Authorised personnel can enter and exit site areas without touching devices or doors – safer in a pandemic.
Cameras located at strategic points around a site remove the need for access control and time and attendance readers. The person gaining access doesn’t require a pin number, swipe card or fob – their face is their passport. Security is improved but it’s more cost-effective to businesses or employees as there are no longer cards/fobs to be issued, and environmentally friendly because of reduced hardware and plastic consumables.
Facial recognition in action
An example of this technology being put to use for attendance tracking can be seen in a recent warehouse installation. The client wanted to improve security in its warehouse of high-value stock including electronic devices, but did not want to affect the flow of staff in and out of the premises.
Automated turnstiles were installed at the entrance of the warehouse linked to facial recognition access control terminals to create ‘speed gates’ – colleagues just walk towards the gate and it opens automatically with no delay.
Linked in to this system is a ‘randomiser’, which randomly selects colleagues coming out of the warehouse for a security check, without having to stop everyone. It’s a highly effective security solution that can be readily adopted for most buildings, even those with large numbers of people.
Accuracy and cost-effectiveness
Advances in processing power and deep learning algorithms have dramatically improved automatic facial recognition’s accuracy. However, accuracy is something that is widely contested. Advocates will say it is up to 97 per cent accurate whereas critics put that figure at less than 25 per cent. In reality, a system is only as good as its setup.
For the best outcome, consider the following:
- Create a well-lit area for a higher-quality image but don’t point lights directly into the camera as this can obstruct the view;
- Install cameras on a fixed structure, such as a wall or a post so they don’t move and change the field of view;
- Mount cameras at between 1.5 and 2.5 metres from the ground to prevent sharp angles of faces; and
- Upload and maintain an accurate database of faces, to make sure the facial recognition is fully used.
One barrier to uptake has previously been cost, but it isn’t as expensive as you might think now. Cloud-based automatic facial recognition, which allows you to stream standard CCTV images to the cloud for analysis, enables you to use existing hardware and overlay analytics “on the edge”, so you do not have to invest in equipment and labour.
If you are planning to improve your on-site employee tracking, be sure to look at the viability of incorporating automatic facial recognition.
Chris Coughlin is technical product manager- CCTV, Intruder and Monitoring at Stanley Security