john Gill technology header image

An Introduction to Biometrics - Face Recognition


What is face recognition?
Face recognition refers to an automated or semi-automated process of matching facial images. The image of the face is captured using a scanner and then analysed in order to obtain a biometric “signature”; different algorithms can be used for this and manufacturers have adopted various proprietary solutions [1].

Different types of face recognition
The term, face recognition, is used as though it refers to a single type of technology but in fact it constitutes a heterogeneous group of technologies which all work with the face but use different scanning techniques. Most common by far is 2D face recognition, using images captured by a standard camera. 2D face recognition is easier and less expensive compared to other approaches, but the technical challenges are greater (systems cope badly with variations in face orientation and lighting conditions) leading to lower accuracy rates. Research has also been carried out using 3D images resulting in reduced sensitivity to factors such as makeup and changes in illumination but with the disadvantage that the scanners are more expensive and the 3D images are not backwards-compatible with existing photo databases. An alternative approach is to use infra-red (IR) radiation to scan facial heat patterns though this is not a prime area of research.

How does it work?
There are four steps in face recognition. Steps 1 and 2 constitute the enrolment procedure. The information is then stored either in a centralised database or on a distributed storage medium such as a smart card. For identification or verification, steps 1 and 2 must be repeated followed by steps 3 and 4.

  1. Acquiring a sample
    The first step is generic for all biometric technologies; it consists of a sensor taking an observation. In the case of 2D face recognition, the sensor is a camera and the observation is a photograph or series of photographs. This acquisition can be accomplished by digitally scanning an existing photograph or by taking a photograph of a live subject. As video is a rapid sequence of individual still images, it can also be used as a source of facial images, though at present the standard of image quality makes this less suitable.
  2. Extracting features
    The generic second step is to extract the relevant data from the captured sample. For face recognition there is the added difficulty that first the face has to be located within the acquired image. This can either be done manually by marking the location of the eyes or through the use of software. Once this has been accomplished, the features of the face can be extracted. Algorithms used for this process are mostly proprietary and will depend on the manufacturer. The outcome is a biometric template, which is a reduced set of data that represents the unique features of the enrolled user’s face.
  3. Comparing templates
    The nature of the third step will depend on the application at hand. For identification purposes, this step will be a comparison between the biometric template captured from the subject at that moment and all the biometric templates stored on a database. For verification, the biometric template of the claimed identity will be retrieved (either from a database or a storage medium presented by the subject) and this will be compared to the biometric data captured at that moment.
  4. Declaring a match
    The face recognition system will either return a match or a candidate list of potential matches. In the second case, the intervention of a human operator will be required in order to select the best fit from the candidate list. An illustrative analogy is that of a walk-through metal detector, where if a person causes the detector to beep, a human operator steps in and checks the person manually or with a hand-held detector [2].
    Some systems automatically update the template periodically which means that the system is better at handling gradual changes in appearance. However, this may only be viable when the template is stored on a cantral database, and not in a secure token such as a passport.

Applications

Machine Readable Travel Documents (MRTDs)
The most important development for face recognition is the introduction of the face as the primary biometric on MRTDs. The ICAO’s reasons for recommending the face highlight the benefits of the technology [3]:

  • Facial photographs do not disclose information that the person does not routinely disclose to the general public
  • The photograph (facial image) is already socially and culturally accepted internationally
  • It is already collected and verified routinely as part of the MRTD application form process in order to produce a passport to ICAO Document 9303 standards
  • The public are already aware of its capture and use for identity verification purposes
  • It is non-intrusive – the user does not have to touch or interact with a physical device for a substantial timeframe to be enrolled
  • It does not require new and costly enrolment procedures to be introduced
  • Capture of it can be deployed relatively immediately and the opportunity to capture face retrospectively is also available
  • Many States have a legacy database of facial images captured as part of the digitised production of passport photographs which can be encoded into facial templates and verified against for identity comparison purposes. However, many of these legacy databases may store the photographs at too low a resolution to be reliably used for facial recognition
  • It can be captured from an endorsed photograph, not requiring the person to be physically present
  • It allows capture of children’s biometrics without the children having to be present
  • For watch lists, face (photograph) is generally the only biometric available for comparison
  • It always acquires
  • Human verification of the biometric against the photograph/person is relatively simple and a familiar process for border control authorities

The European Parliament has now voted in favour of introducing biometrics on MRTDs and it is foreseen that this application will be implemented within the next few years.

Existing face recognition applications
In 2001, the Tampa Bay Police used face recognition technology to screen the spectators that attended the Super Bowl game against a watch-list of known felons.

A borough of London was one of the first areas to introduce face recognition in 1998, in order to screen images from closed circuit television cameras (CCTV) for targeted offenders [4].

Face recognition has also been tested in airports around the world, including Keflavik Airport Reykjavik [5], Logan Airport Boston [6], Palm Beach International Airport Florida [7] and Sydney Airport [8] with mixed results.

Law enforcement
Face recognition offers certain facilities not available with other biometric technologies. One feature that appeals in particular to law enforcement agencies is the option of matching witness descriptions or artist-rendered images to databases of suspects, i.e. the capacity to compare biometric data with non-biometric data within the same system. Though the results are not precise enough to be admissible as evidence, they can provide the police with leads for further investigation [9].

Database mining
One of the touted advantages of face recognition technology is that it is compatible with existing databases of facial images. Many countries have databases of passport photographs, driver’s license photographs, mug shots, etc., and face recognition could be used to mine existing databases, checking for duplicates and multiple identities.

Acknowledgement: The above information was taken from Biometrics at the Frontiers: Assessing the Impact on Society report.


Standards
Published standards:

  • ANSI INCITS 385-2004: Information Technology-Face Recognition Format for Data Interchange
  • ANSI/NIST-ITL 1-2000: Information Systems-Data Format for the Interchange of Fingerprint, Facial, & Tattoo (SMT) Information
  • ISO/IEC 19794-5:2005: Information technology - Biometric data interchange formats - Part 5: Face image data


Further information


[1] Biometric-based technologies. OECD, Working Party on Information Security and Privacy, 30 June 2004, DSTI/ICCP/REG(2003)2/FINAL.
[2] National Institute of Justice 2003.
[3] Biometrics deployment of machine readable travel documents. ICAO TAG MRTD/NTWG, Technical Report, Version 2.0, May 2004.
[4] www.guardian.co.uk/g2/story/0,3604,736312,00.html
[5] Keflavik Airport, Iceland, archives.cnn.com/2001/US/09/28/rec.airport.facial.screening/
[6] Logan Airport
[7] Palm Beach www.usatoday.com/tech/news/2002/05/16/airport-face-recognition.htm


Acknowledgement

The information contained in this section was collected from the following source:


Previous page

 



John Gill Technology Limited Footer
John Gill Technology Limited Footer