Abstract
Face recognition has a broad range of uses for business and law enforcement, such as access control, security monitoring, and video surveillance. This paper proposes an effective algorithm for Infrared face recognition using a Local Binary Pattern (LBP) for extraction of features and a Canonical Correlation Analysis (CCA) for fusion and classification of features. The facial characteristics are extracted using LBP. The extracted characteristics are then converted into different domains for transformation. For dimensionality reduction, the two-Dimensional Principal Component Analysis (2DPCA) approach is used in the proposed algorithm to generate more lightweight, robust and discriminatory features, which are then combined using the CCA classifier. The spatial relationship between adjacent pixels is also maintained by 2DPCA, increasing the overall accuracy of recognition. In addition, the paper introduces a comparative study between infrared facial recognition systems using the proposed technique and previous work. Based on the recognition rate and time usage, the output is evaluated. The analysis of the findings shows that the technique proposed is the most efficient and the least time compared to previous techniques. The experimental results are tested on a dataset acquired by Equinox Corporation. The proposed technique achieves a recognition rate of 99.26% at 0.45 seconds.