Facial recognition: what you should know
Source: Prasanth Aby Thomas
Facial recognition systems are increasingly becoming popular these days, as it provides an accurate method to identify an individual and ensure security. As global security concerns rise on the back of geopolitical tensions and terrorist threats, the facial recognition market is expected to see massive growth.
According to the research firm MarketsandMarkets, the global facial recognition market size is estimated to grow from US$ 3.35 billion in 2016 to $6.84 billion by 2021, at a CAGR of 15.3 percent during the period. A growing surveillance market, increasing government deployment and increasing applications in several industry sectors are factors set to drive demand for this technology.
Facial recognition has a number of advantages over other biometric methods of identification. It is the least intrusive and the fastest. Unlike systems like fingerprint recognition where the user is required to place his finger on the reader, facial recognition systems can analyze an individual almost without their own knowledge, making it the most unobtrusive.
Understanding facial recognition technology
Generally speaking, there are mainly two methods that are widely used in facial recognition systems today. The first method gives importance to some of the features of a person’s face, calculating aspects like the space between the eyes, or the proximity of the mouth to the nose. These calculations are then fed to a database and used to compare later.
The second system is much more comprehensive and complicated. It records the whole image of the face and utilizes as much information as available. This information is then subjected to different kinds of computer algorithms like machine learning to create certain descriptive data. Such a database is naturally more dependable, although its costs can be higher.
Then there are other methods like 3-D modeling, skin texture analysis and thermal camera usage. However many of these are more resource-hungry and raises doubts on the price-performance balance.
Important factors to consider
Obviously, for the best performance of any facial recognition system, there should be images that are clear enough to be analyzed. And the clarity of the image relies heavily on the available light, resolution of the camera used, angle of the face and movement.
The first aspect, lighting, is perhaps the most crucial. Low light will cause the interference of noise to captured footage and this can hinder the chances of accurate facial recognition. For the facial recognition algorithm to read the features properly, the light should provide adequate contrast. According to Bob Mesnik, President of the systems integrator Kintronics, many of the systems require at least 300 to 500 lux of illumination, which is just about the kind of lighting that we see in an office setting. Moreover, there should be consistent lighting to avoid shadows and the negative effects they bring in.
Once the lighting is ensured, installers have to decide on the camera and its resolution. This largely depends on the kind of facial recognition system that is to be used and the total field of view. As a general rule, the required resolution will depend on the width of the field of view. Some solutions mandate specific number of pixels over certain facial features. Once you take a note of the requirements that the facial recognition system specifies, you can decide on the resolution and the camera.
Perhaps the biggest challenge in ensuring accurate facial recognition is making sure that the subject is actually facing the camera. In places like the immigration counter of an airport, officials can ask people to look straight into the camera but in most other areas it is all about making sure that the camera is fixed at an ideal location. A doorway or a gate where people are likely to look forward is an example of a good location.
For the camera to capture the face accurately, there shouldn’t be too much movement either. If installed in a site where people are moving too fast, the solution would require cameras with higher frame rates.
Where can it be used?
Modern facial recognition solutions are installed in a number of locations, but their usage can be broadly classified into two – to recognize certain specific individuals from a crowd and for access control.
Recognizing certain people by running them past a database is useful not just for security purposes, but also for business improvement. For instance, retail stores can use it to identify special customers. For access control, the usage is comparatively easier since users are expected to stand in front of the camera under specific lighting.
While these are the basics of facial recognition systems, there are more advanced systems on the market now that make use of more features, like a person’s body movements for authentication. However, these would naturally cost more than the usual systems. Eventually it’s up to the installer and the customer to reach a consensus on the requirements of a site and decide on the best solution.