An 3D face recognition approach based on facial curve analysis

Xuexian Hou, Xinzhi Zhou, Yinjie Lei

COMPUTER MODELLING&NEW TECHNOLOGIES 2014 18(12B) 633-641

School of Electronic Information, Sichuan University, Chengdu 610065, China

In this paper, we present a novel 3D face recognition approach based on the analysis of facial curves, which are extracted from the semi-rigid facial regions. Our approach excludes the facial region which is most affected by facial expressions (non-rigid region) resulting in a set of indexed open geodesic curves. A novel open curve analysis algorithm combining the geodesic and Euclidean distances is used to match the same level pairs of open geodesic curves of a probe and the gallery faces. In order to increase the accuracy of face recognition, a curve ranking and weighting algorithm is also developed to select, during a training phase, the most reliable curves and to assign different weights to the selected ones. During the testing phase, the selected reliable curves and their corresponding distances are weighted fused to perform face recognition. The proposed approach has been tested on the Face Recognition Grand Challenge (FRGC v2.0) dataset via a number of experiments and a superior recognition performance was achieved.