Single 2D pressure footprint based person identification

Abstract

Footprints carry many important human characteristics, such as anatomical structures of the foot, skin texture of the foot sole, standing or walking habits, and so on. They play vital roles in forensic investigations as an alternative biometric. In this paper, we propose an automatic footprint based person identification method using a single bare or socked footprint, which differs from the existing bare footprint based methods. An area rank filter is put forward to remove dust noises. Pressure distribution prior of the hind footprint is proposed to estimate the footprint direction. Both Geometrical Shape Spectrum Representation and Pressure Radial Gradient Map are proposed to represent a footprint in views of geometric shape, anatomical structure and one's standing or walking habits, which are also rotation and translation invariant. We also put forward a regional confidence value based method to compute the similarity values between two footprints. Additionally, we have constructed an evaluation dataset composed of 480 subjects and 19200 bare or socked footprints. Experimental results show that the proposed algorithm outperforms state of the-art algorithms, and its recognition rate reaches 98.75%.