Pose-Invariant Face Recognition Security System

10.6412/AJHIS.200604.0101

Hui-Fuang Ng

face recognition ； security system ； pose-invariant ； similarity vector ； principal component analysis

Asian Journal of Health and Information Sciences

1卷1期（2006 / 04 / 01）

101 - 111

英文

Face recognition security systems have become important for many applications such as automatic access control and video surveillance. Most face recognition security systems today require proper frontal views of a person, and these systems will fail if the person to be recognized does not face the camera correctly. In this paper, we propose a novel approach for robust pose-invariant face recognition. Our recognition system is insensitive to viewing directions and it requires only one sample view per person. The proposed approach utilizes the similarities of a face image against a set of faces from a training set at the same view to establish pose-invariant representations of a person in different poses. Experimental results indicate that the proposed approach achieves high recognition rates even for large pose variations.