A brand of Legrand

Facial Recognition

Facial recognition is a software application used to automatically identify people in photos or videos. It utilises a database and the principles of artificial intelligence. Increasingly used for practical everyday applications, facial recognition also plays an important role in surveillance and security. But how exactly does it work and what are its applications? What questions does it raise in terms of ethics and privacy?

What is facial recognition?

Facial recognition is a biometric technology used to identify people based on the characteristics of their face, notably the distance between the different facial features (eyes, ears, corners of the mouth, etc.). Unlike "human" facial recognition, which involves facial recollection profiles, the biometric technology for automatically recognising faces in photos or videos refers to an existing database to establish comparisons and parallels.
Emerging in the 1970s, facial recognition technology is now booming, with practical, commercial and security applications. Generally speaking, the idea is to recognise a person automatically based on their traits. More specifically, it's used to identify a person or authenticate them, i.e. to verify their identity; for example, by comparing their image with the picture in their passport, such as at the Roissy and Orly airports in Paris.
Facial recognition and facial detection technology are sometimes confused. The latter is designed to detect the presence of a human face in a photo or video, rather than to identify it, as is the case with facial recognition.

How does facial recognition work?

A facial recognition system relies on several complementary technologies: a sophisticated image capture system (photo or video via a surveillance camera), artificial intelligence and machine learning, also called automatic learning.
Using static images or a video, facial recognition software converts the characteristics of a face into digital biometric data. An algorithm then compares these recognition elements, either with a model corresponding to the presumed identity in the case of authentication (a passport photo for example) or with the elements already contained in a database if the goal is to identify an individual. In the latter case, the algorithm searches through the "templates" already listed to find those with the highest similarity score. Identification takes place based on these facial matches.
Facial recognition can take place in two dimensions, when it uses the shape and measurements of facial features (eyes, nose, etc.), or three dimensions when several angles of the face are used (front, profile, three quarters, etc.) to create the model from photos or a video recording. Discover all our products

How is facial recognition used?

There are a wide variety of applications for facial recognition. The most commonly mentioned are surveillance and security uses: the identification of criminals by the police, checking the identity of travellers at borders or passengers at airports in a special area fitted with CCTV monitors, etc. This system provides greater security and saves passengers time. Facial recognition technology can also be used to control access to sensitive sites where visitors' identity needs to be verified.
In practical terms, this biometric technology can also be used to restrict access to events to individuals who have been invited and identified as such. Similarly, facial recognition authentication can be used to unlock camera-controlled entry points or start up a vehicle equipped with biometric recognition software.
Identification through facial recognition also has many benefits on the internet or when using certain applications. When installed on a mobile device equipped with one of these cameras (smartphone, tablet or computer), this type of system can identify the user quickly without the need for codes or passwords, or as a complement to these if a more sophisticated security system is required (in the case of sensitive data for example). Facial recognition can even be used to secure online transactions: the camera in the mobile device or computer compares the video or photo of the user with a reference image kept in a secure storage area. Used alongside a security code, this system ensures safer transactions and limits fraud.
Although more trivial but just as useful, facial recognition technology plays a role in indexing images and videos on the internet. It can be used to tag photos and social media or in certain mobile apps by identifying each registered user.
More recently, commercial applications have been developed. In a sales outlet, a camera equipped with facial recognition software can recognise a customer who is already in the database and provide suitable sales offers on interactive screens or panels. It can also determine what category the person belongs to (man, woman, child) so that it can offer suitable content.

The effectiveness and limitations of facial recognition

Systems that incorporate facial recognition technology aren't without their faults. Various factors are involved in obtaining reliable results. Image quality obviously plays a role and varies depending on the type of camera used, the distance of the subject whose image is being captured and also their cooperation. The same results won't be achieved with faces captured close up, with the individuals' consent, versus an image taken from a surveillance camera or using a low-quality smartphone.
Other decisive factors are the performance of the algorithms used and the quality of the reference database. The more material there is to compare, the easier it is to find relevant matches and the more reliable the facial recognition will be.
To improve the reliability of this innovative technology, 3D sensors have been designed to more effectively identify faces in motion or seen from different angles. The police even use software that "ages" faces so they can compare models captured at different times. Despite these constant improvements and the ever-more sophisticated cameras, mistakes can still be made. The CREOGN (the research centre of France's college for national gendarmerie officers) puts the error rate in facial recognition applications at up to 20%.

Facial recognition, ethics and legislation

Thanks to its ease of implementation, which is contactless, non-invasive and remote and can even identify individuals in a crowd, facial recognition has proven popular with the police and security services. Now used in many airports, border areas and public spaces, this technology nevertheless raises ethical and security questions.
Microsoft's legal director himself recently stressed the importance of establishing a clear legal framework for this identification technology in order to prevent abuse. If misused, whether by an authoritarian government, criminals or unscrupulous companies, facial recognition could threaten individual freedoms.
Some countries like China use this technology to excess. In France, the CNIL (national commission for data protection and liberties) governs the use of facial recognition in the public arena. One of the key measures of the system is to always notify individuals whose image has been captured. As such, the person is never unknowingly filmed and a data protection impact analysis must be carried out whenever cameras are installed in a public space.
In addition, the European directive 95/46/EC on data protection also applies to facial recognition systems. As the technology evolves and its use becomes more widespread, it therefore becomes necessary to manage how it's used in order to prevent any abusive use of facial recognition. This also means ensuring that the biometric data stored is secure. Lastly, the error rate in the algorithms also needs to be taken into account.
The protection of systems and the implementation of a legal framework are therefore key issues in facial recognition, and only these measures can guarantee that effective security systems will be implemented without violating each individual's basic freedoms.

Discover all our products

Smart Home Weather Station

Smart Thermostat

Smart Indoor Camera

Smart Outdoor Camera