Fast vertical-pose-invariant face recognition module for intelligent robot guard

Abstract

Based on a semi-3D cuboid face model, this paper proposes a simple but practical preprocessing method to recover the vertical pose variation for the face recognition module of the intelligent robot guard. The proposed Fast Semi-3D Vertical Pose Recovery method evaluates the angle of the vertical pose variation simply from a single 2D image view and thereby recovers the vertical-rotated face to the nearly frontal view. Consequently, the recovered face image can be identified by the original face recognition module more precisely and effectively. In the experiment, this paper adopts Gabor Wavelet transform as the feature extraction core of the original face recognition module. Experimental results show the proposed method can significantly raise both recognition accuracy and recognition confidence of the face recognition module under various vertical pose conditions.