Face Recognition or Detection

Sometimes you don’t want to know who it is, but only that it’s someone.
In that case, there is face detection.
This option is now standard in photo cameras and cameras of many modern smartphones.
In the image (in viewfinder or on screen) a (software-based) frame appears around the face.
The camera software then automatically focuses on that face.
Security cameras can also do this and therefore see that there is a person in the picture.
This makes it easier and much faster to filter or search for camera images with people.
Without violating the privacy of these individuals.

Restrictions

If you want to recognize faces, there are a number of things to take into account.
Changing faces While the bone structure of a skull (up to 16 years of age) will not change quickly, relatively large changes can take place on the surface of a face in a short period of time.
Think of someone who shaves their head or grows their beard.
Losing a lot of weight also sometimes affects the shape and structure of the face.
Just like certain cosmetic procedures.
In all these examples, the chance of recognition by the software decreases. Software can be fooled If you use facial recognition for access control, everyone understands that it only works if your face is uncovered.
If not, you just won’t get in or you won’t get through customs.
If there is recognition for investigative purposes, criminals gratefully make use of the opportunity to fool the facial recognition software.
This is child’s play and can be done with easily available items, such as caps, large sunglasses, fake beards, wigs, trendy Japanese face masks or scarves.
Or simply by shaving off a beard or shaving the head.
Even holding a hand in front of your face is enough.
In all these cases, the software will no longer be able to identify the face.
A fake face can also fool some recognition software.
The unlocking systems of Huawei and Samsung mobile phones have been ‘hacked’ several times with an almost perfectly recreated wax head or even with a photo.
Fortunately, professional facial recognition systems that are used for access control are less likely to go wrong.
These often use multiple cameras that measure many points of a face at different angles.
In addition, they have various techniques on board to detect whether there is a living person.
Perfectly recreating a head with a heartbeat is still only reserved for Tom Cruise in Mission: Impossible….
Unfortunately, security cameras do not have these smart techniques.
They have no 3D, no heat detection and just give a flat camera image.
These cameras are therefore often easy to fool with a photo.
Afterwards you can of course see on the recorded camera image that someone is walking with a photo in front of his head, but as said, that is afterwards… There are 3D scanners that are much more accurate than an average security camera. Privacy laws Biometric identification techniques are increasingly being used; Also in everyday applications, such as smartphones, tablets and laptops.
Another practical application is the Nest doorbell with camera.
The camera in this doorbell detects a face even before the doorbell is pressed.
When you assign names to previously recorded images, the doorbell will indicate who is at the door the next time, this can also be a spoken version via Google Home speakers.
China goes a few steps further. There, the Smart City concept has been implemented to such an extent that residents on public roads are continuously monitored by the government via facial recognition and can even be alerted to any legal missteps.
By linking to other techniques that can be used to identify people (Bluetooth, Wi-Fi and GPS scanning of mobile phones, purchasing behaviour, etc.), the Chinese government is meticulously mapping the behaviour of citizens.
Fortunately, this scenario is still unthinkable in the Netherlands.
We have to deal with strict European privacy legislation.
The introduction of biometric identification techniques without a legal basis will therefore be a costly divestment, not to mention the potential reputational damage.
The General Data Protection Regulation (GDPR) explicitly mentions biometrics.
In principle, the processing of biometric data to identify individuals is prohibited.
Unless data subjects give explicit and voluntary consent to it or it is necessary for authentication or security purposes (Article 29 of the GDPR Implementation Act).
If it is necessary to secure buildings or information systems in such a way that biometric identification is required, the basis of ‘legitimate interest’ can be invoked.
However, it is still necessary to substantiate this properly.
In addition, a number of other obligations under the GDPR must be met, such as informing data subjects and taking appropriate technical and organizational measures to protect the biometric data collected.
Companies with a Works Council must ask this body for permission to introduce a biometric system for authentication or identification purposes.
Also read our article and whitepaper on this topic.

Future

Facial recognition techniques are improving at a rapid pace.
For example, the technology requires much less processor capacity, which is why facial recognition is now also available as an add-on in existing cameras.
Furthermore, reliability has increased considerably thanks to improved measurement techniques.
Machine Learning (in the form of a large database of faces to ‘practice’ with) plays an important role in this.
Incidentally, the difference in recognition results between Western software and that from China and surrounding countries is striking.
Google is significantly better at recognizing Western individuals, probably because of the database offered.
With software from China, it’s the other way around.
In short, more and more is possible and the results are increasingly reliable.
However, you should always be aware of the legal framework for any form of biometric identification.
The covert use of facial recognition technology is hardly controllable, but the storage and processing of biometric data requires careful consideration and implementation.