printer friendly version

Face it

Facial recognition is nothing new. A number of facial recognition terminals have been available for several years. Unfortunately, this type of biometric reader is problematic in the African environment due to a combination of extreme light conditions and dark complexions.

Variations in background illumination have been one of the main challenges for practical face recognition systems. It is well known that variations of the same face due to illumination are generally greater than the variations of different face identities.

Many face recognition algorithms, such as illumination compensation, image preprocessing, illumination invariant feature extraction and modelling of the face and illumination have been created to compensate for this problem. However, these techniques cannot completely compensate for the large variation in the appearance of the same face due to changes in illumination.

Active illumination-based face recognition techniques are considered to be one of the most promising and practical methods for solving illumination issues in indoor applications. It uses an active sensing technology to create desirable ambient illumination, unaffected by uncontrollable surrounding illumination.

It must be noted, however, that this technology has its own limitations, based primarily on the variable distance between the user and the active illuminator. Over-saturation, partial illumination and the lack of illumination are three inherent issues. These can result from the variation of this distance and can significantly impact the characteristics of a face and, in turn, ultimately degrade the performance of a face recognition system.

Suprema’s Adaptive IR Illumination Technology solves this issue by controlling the intensity of the illumination based on the analysis of the image as well as various features of the face. The illumination is adaptively adjusted to create the ideal ambient environment for the capture of clear face images.

Non-intrusive and easy to use

Furthermore, the human face is one of the most common and non-intrusive biometrics used to identify individuals. It is more universal, acceptable and easier to access than a fingerprint.

Conventional face recognition technology contains inherent weak points brought about by pose variations. FaceStation’s face recognition technology overcomes such weak points with Smart Enrollment Technology, Advanced Face Recognition Algorithm and Adaptive IR Illumination Technology.

Conventional face recognition systems use just the frontal face and thus require the user’s active cooperation to establish the correct position. Different poses have the potential to drastically increase not only the false acceptance rate, but also the false rejection rate and therefore greatly affect the overall performance of a face recognition system. To compensate, conventional face recognition systems require the user to control their angle/pose to coincide exactly with how they were enrolled. This is very time consuming and extremely inconvenient for the user.

Suprema’s Smart Enrollment Technology solves this problem by using enhanced face analysis algorithms, combined with a systematic procedure to view key poses. The system begins by registering a few frontal face templates. These templates then become the reference for evaluating and registering the templates for the other poses: tilting up/down, moving closer/further, and turning left/right.

Strike a pose

Extensive experimental tests show that each set of poses contributed greatly to the overall performance of the system. With each pose, the facial information including eyes, nose and mouth, is automatically extracted and is then used to calculate the effects of the variation using its relation to the frontal face templates.

Algorithms were also created to automatically detect and reject improper face images during the enrollment process. This ensures proper enrollment and the best possible performance.

Although face recognition has many advantages, the concern against ‘Fake Face Attacks’ is highly warranted. Face recognition systems cannot be practically used without proper countermeasures to this threat.

Even when compared to a fingerprint, a face can be more easily imitated using printed papers and LCD displays. There have been many attempts to overcome this issue: hardware-dependent approaches such as facial thermogram, facial vein and 3D depth information require incredibly expensive hardware, making it infeasible for practical use. Conventional algorithm-based methods using 2D images, up to now, have not been accurate and required exhaustive calculations/computing power.

Suprema’s ‘Fake Face Detection’ technology combines cutting-edge technology with advanced proprietary algorithms. A dual camera system captures both visible and IR images, which are then processed using advanced image analysis techniques and intelligent machine learning-based classifiers. A fake face is detected by estimating specific features and their distribution is compared to reference models of real faces.

Share this article:

Further reading: