Greater security through improved facial recognition : Date:
With the FeGeb project of Bonn-Rhine-Sieg University of Applied Sciences, the Federal Ministry of Education and Research is funding the development of a new type of facial-recognition system that can reliably recognise artificial and falsified biometric features.
Because of its practical advantages, biometric facial recognition is a commonly used method for verifying the identity of a person. It is mainly used at locations with increased security requirements (e.g. border crossings) or as access control for the protection of critical infrastructure (e.g. high-security areas). A common problem of all biometric procedures is that body characteristics can be falsified and imitated – including facial features. The currently available systems for facial recognition can only detect and ward off deceptions by facial imitations, such as artificial facial parts and masks, to a limited extent.
As part of the FeGeb project, the research group led by computer science professor Norbert Jung is now developing a system for identifying people that is protected against attempts at deception. The work is based on research the group conducted on skin recognition using near-infrared sensors. Within the spectrum of visible light, human skin cannot be recognised beyond doubt by its colour and distinguished from skin imitations. Under invisible near-infrared light, on the other hand, it is possible to reliably differentiate skin – regardless of a person’s skin type, age or sex – from other materials based on the light reflection typical for it, its spectral signature.
Within the framework of the project, Jung is developing a near-infrared camera system together with his doctoral student Holger Steiner. This system emits short pulses of light in the near-infrared range, and then captures the light reflected from the skin and combines it into a multispectral image. With the help of special software, it is then possible to analyse the spectral signatures of the faces recognised in the image and thus differentiate real faces from fakes. Thus, attempts to mask the face, such as a fake nose, can be identified.
The Federal Ministry of Education and Research is supporting the research project with about 324,000 euros within the framework of the FHprofUnt funding line as part of the Research at Universities of Applied Sciences programme.