Ear biometrics recognition using laser biospeckled fringe projection profilometry

Abstract

Anti-spoof touchless imaging of human ear has strong potential to be a useful alternative to traditional biometric sensors. The features in a human ear are more reliable than other physical traits like fingerprint, face, palm print, etc. In this paper, we propose an efficient ear pattern sensor by introducing a new imaging technique that is capable of simultaneous retrieval of the topographic details as well as its liveliness information. In the proposed technique, a collimated laser beam is made incident on a sinusoidal grating, and the resultant structured pattern is projected towards the ear specimen. For detection of topographic information and liveliness, time series bio-speckled fringe sequences are recorded using a CCD camera. For noise removal and shape reconstruction from the acquired information, a combination of windowed Fourier filtering and low pass filtering is used. Discrete cosine transform based fast biospeckle algorithm is introduced for assessment of genuineness (i.e. to create distinction between real and fake samples) of the specimen. The proposed sensor is simple, compact, full-field and robust against different spoof attacks. © 2018