Introduction to facial recognition
With emerging markets and global advancements, the facial recognition market is predicted to expand. It’s estimated to reach $6.5 Billion by the end of 2018.
With 2D, 3D, thermal, emotion, forensic and mobile facial recognition (similar to the technology deployed in the iPhone X), popularity has increased in the corporate world.
Prime systems adopting facial recognition include:
- access control
- time and attendance systems
These systems all benefit from increased hit rates and low false acceptance rates.
Let’s delve deeper to understand the technology, how it works and how to use it to best effect.
What is facial recognition?
Facial recognition is a process whereby a CCTV camera and some server software are able to identify a person from a live moving image.
The system identifies the face of a person whether they are static or moving.
This is different to the type of 2D still facial image matching used at airports for Border Control.
These systems use a static digital image captured from the chip embedded into the passport and match it against the person standing at the gate.
How it works
Facial recognition systems start by looking for peoples faces when observing a moving scene or image.
Once the system finds an image, it will adjust the image by resizing or reorientating the image to clearly view a face.
The system will then observe the physical layout of the face, such as the shape of the face, location of the eyes, nose and mouth.
It will use this layout information to build a template for that face, sometimes also referred to as a faceprint.
Once the template has been created, the system will then compare it to the templates that have been enrolled into the system to attempt to identify a person.
If the system identifies an enrolled person, it will positively identify them and pass this data to an operator terminal or screen.
It may also pass a unique identification number to another system, such as the access control system, instead of a card number.
If the system cannot identify the person against those enrolled in the database, it will log that it detected a person in the scene, but could not positively identify them.
Unlike CCTV which records images whenever someone passes in the field of view of a camera, facial recognition requires a database of people enrolled before it can start recognising people.
Unregistered people will appear as unknown.
Data such as counting the total number of people, or the amount of time people takes to move from one point to another can be recorded anonymously.
Facial recognition systems use cameras that are very similar to CCTV cameras, therefore people are less aware of their presence.
We are very used to seeing cameras in our daily lives and cannot easily tell what the camera is being used for.
It’s easier to move a camera, or alter the field of view of a camera and the area it is monitoring.
Most cameras use a single data cable for both power and data.
In comparison, moving an access control installation would require more extensive engineering work, such as moving controller equipment and recabling readers.
The organisation could also choose to deploy additional temporary cameras for a specific event, incident or investigation.
No physical credentials to issue
With most access control systems, you need to issue a physical credential such as a card.
The process of getting the card programmed, printed and issued is complex.
Card issuance involves a number of people and departments to get it into the hands of the authorised person.
In comparison, a simple process of capturing their facial data the first time they visit site is both faster and less costly to the organisation.
For facial recognition systems to work, it needs to view a persons face.
If people do not look in the correct place, wear hats or sunglasses or obscure their faces, they will not be identified.
Anyone not willing to look at a camera or obscuring their face will need to be monitored and tracked by a member of the surveillance team.
In most corporate organisations, this a standard procedure.
Installation is more complex
Installation of cameras at the correct viewing angle to view someone’s face requires careful planning and is vital.
Lighting should be of a sufficient level to illuminate the scene, to allow the system to recognise a person’s face and then identify them.
Systems are still expensive
Facial recognition is gaining in popularity and the technology is getting better.
However, the costs are still, fairly high in comparison to video content analytics and CCTV.
Organisations may choose to retain card or biometric-based access control in most areas.
Facial recognition can be deployed in addition in specific locations to supplement card or biometric solutions.
Reliance on infrastructure
Facial recognition systems require constant access to a central database of enrolled persons to operate successfully.
This makes the design and availability of both the servers and the network are vital.
This requires careful design and a significant IT investment to ensure that the database is always available.
In comparison, card-based access control systems store the data in the controller at the edge.
This eliminates a single point of failure and reduces reliance on the server.
Facial recognition is being used for more than just security, it’s also used for hospitality.
Imagine a casino environment where high-rollers can be greeted by name.
The casino’s VIP Concierge team can make them feel appreciated and welcomed.
Football match safety can be enhanced where the behaviour of known hooligans can be monitored.
Systems can monitor if they sitting still, standing or moving around.
Security staff and police can react when necessary, resulting in less confrontation, when compared to blocking entry at the turnstiles.
I would like to thank Oliver Lacey at Nedap Security Management for his assistance in helping to create and check the accuracy of this article.