Face Recognition

Face recognition uses the spatial geometry of distinguishing features of the face. It is a form of computer vision that uses the face to identify or to authenticate a person.

An important difference with other biometric solutions is that faces can be captured from some distance away, with for example surveillance cameras. Therefore face recognition can be applied without the subject knowing that he is being observed. This makes face recognition suitable for finding missing children or tracking down fugitive criminals using surveillance cameras.

Independent of the solution vendor, face recognition is accomplished as follows:
  1. A digital camera acquires an image of the face.
  2. Software locates the face in the image, this is also called face detection. Face detection is one of the more difficult steps in face recognition, especially when using surveillance cameras for scanning an entire crowd of people.
  3. When a face has been selected in the image, the software analyzes the spatial geometry. The techniques used to extract identifying features of a face are vendor dependent. In general the software generates a template, this is a reduced set of data which uniquely identifies an individual based on the features of his face.
  4. The generated template is then compared with a set of known templates in a database (identification) or with one specific template (authentication).
  5. The software generates a score which indicates how well two templates match. It depends on the software how high a score must be for two templates to be considered as matching, for example an authentication application requires low FAR and thus the score must be high enough before templates can be declared as matching. In a surveillance application however you would not want to miss out on any fugitive criminals thus requiring a low FRR, so you would set a lower matching score and security agents will sort out the false positives.
 

Difficulties that often arise with face recognition are

  • Variable image lighting and background make it more difficult for software to locate the face in the image.
  • Parts of the face are covered, e.g. long hair, makes it more difficult for the software to locate the face in the image and to recognize the face.
  • Subject does not look directly into the camera, when the face is not held in the same angle the software might not recognize the face.
  • Using different types of cameras (with different lighting, resolution, etc.) makes it more difficult for the software to recognize the face.
  • The face of a subject changes with ageing.
  • It is difficult to make face recognition secure enough for authentication purposes.

As you can see there are some important constraints for using face recognition. Different vendors work on resolving these issues.

3D face recognition solves some of the above issues. Using 3D images the actual 3-dimensional form of the face is evaluated, this is not affected by lighting and does not change with ageing. Also different viewing angles can be better compared when using 3D images. Of course the hardware for 3D face recognition is more expensive.

Application of face recognition

Face recognition can be used together with surveillance cameras to automatically identify missing children, unwanted subjects in casino’s or fugitive criminals for which a picture is registered in a central database.

Different solutions exist for both small and large businesses, as well as for private use, that apply face recognition for access control to computer systems. Software requiring only a webcam to use face recognition instead of passwords to secure your computer or laptop is around for multiple years. Sensible Vision and Keylemon have developed software suitable for small companies and home users:

Sensible Vision has developed a face recognition software that uses such standard cameras to authenticate users on their computer. When an individual sits in front of the computer, the camera acquires an image which the software uses to identify the individual. If the individual in front of the computer is a registered user, then he will be automatically logged on to the computer. As soon as the individual leaves the computer, this is detected by the software and the user is logged off. We did not yet review this face recognition software.
A very similar software has been developed by KeyLemon, however we succeeded in hacking KeyLemon software easily even with its strongest security settings activated. Read our review of KeyLemon face recognition.

Furthermore a lot of system integrators develop custom solutions for border control and physical access control.

Suitability of face recognition

How suitable is face recognition as a biometric solution? We use the following 7 criteria to evaluate the suitability of face recognition:

Universality For some people face recognition might not work as well as for others. For example long hair or a beard might give face recognition systems extra difficulty, and not all marketed solutions will deal with this equally well.
Uniqueness Face recognition cannot distinguish identical twins.
Permanence As you age your face will most likely change. Also injury, plastic surgery or more temporary changes such as sunglasses, make-up or growing a beard might have an impact.
Collectability Faces are easy to collect, direct contact with the biometric device is not required and the subject might not even know that an image of his face is being collected.
Acceptability There certainly are privacy concerns when using a surveillance system to track people’s whereabouts. However applying face recognition for access control will be easier accepted than other biometric solutions because no direct contact is required with a reader, and in general people do not consider taking a photograph as being intrusive as might be the case with biometric solutions such as iris recognition or fingerprint recognition.
Circumvention This is very much dependent on the technical implementation, much depends on the quality of the camera, the control of the surroundings (e.g. background) and on the matching algorithm. Some biometric authentication applications based on face recognition include a liveness check, for example by requesting the subject to blink with his eyes.
Performance Speed might be an issue for surveillance systems, imagine having the matching algorithm verifying the faces of travelers on an airport: a high number of verifications with images that are taken without subjects looking directly into the camera.

We can conclude that face recognition is most interesting because the subject is not necessarily aware that his identity is being verified, this is very useful for surveillance applications. Circumvention is an important factor to consider when choosing face recognition for authentication purposes, furthermore permanence and uniqueness of the face might remain a limitting factor of this biometric solution.

Search