UK riots: Facial recognition software to convict suspected criminals

Within hours of the rioting that took place across London and other cities, facial recognition software swung into action in an attempt to track down those who took part and whose images were caught on CCTV footage.

One of the most common methods used is photoanthropometry, which uses ‘proportionality indices’ to compare a picture of a suspect to those held on police databases. Key points on the face, such as the edge of the nose, the brow line, the centre of the top lip and the chin, are marked and measured.

In theory, the expert carrying out this procedure will be able to judge whether there is a match. In practice, however, the poor quality of the majority of CCTV footage makes this procedure almost impossible.

Lyndon Smith, from the Machine Laboratory of the University of the West of England, Bristol, says of the software currently used “Changes in lighting, image quality, changing background and orientation – even make-up can fool them.”

However, all is not lost. Smith and his colleagues are working on a new system called Photoface, which takes a number of 2D images of a face and then recreates it as a 3D image where the lighting can be altered and the face viewed from different angles.

Smith believes that Photoface will be used on CCTV footage in police control rooms to better indentify perpetrators.

Meanwhile, Alessandro Acquisti of Carnegie Mellon University in Pittsburgh, Pennsylvania illustrated how current facial recognition software is still a powerful weapon.

To demonstrate, a hundred student volunteers were asked to look into a webcam. After just three seconds, facial recognition software linked 31% of the students to their Facebook profile.