Citation
Teo, Min Er and Chong, Lee Ying and Chong, Siew Chin (2024) Fusion-Based 2.5D Face Recognition System. Journal of Telecommunications and the Digital Economy, 12 (1). pp. 19-38. ISSN 2203-1693 Full text not available from this repository.Abstract
Face recognition is the dominant biometrics system used to authenticate an individual’s identity in various applications. Most commercial face recognition systems rely on 2D face images, but the changes in the environment lighting and a person's posture affect the accuracy of the 2D face recognition systems. Hence, the 2.5D face recognition system arises as the solution to eliminate the drawbacks of the 2D face recognition system. The depth feature in the 2.5D data (depth image) provides additional information that can help to improve the accuracy and robustness of 2.5D face recognition systems, particularly in challenging scenarios. This paper proposes a fusion-based approach for the 2.5D face recognition system to enhance the system’s performance, where feature fusion involves the combination of features extracted from the depth image. Gabor-based Region Covariance Matrices (GRCMs) that serve as face identifiers combine the depth and texture images in the structure of a covariance matrix. Several experiments on different fusions have been conducted in the Face Recognition Grand Challenge version 2 (FRGC v2.0) database. This paper shows that the max-min fusion applied to the surface normal (y-direction) and the mean curvature has achieved the best accuracy rate of 93.66% among the other fusion approaches used.
Item Type: | Article |
---|---|
Uncontrolled Keywords: | fusion-based approach, depth image, 2.5D data, Gabor-based Region Covariance Matrices, 2.5D face recognition |
Subjects: | Q Science > QC Physics > QC350-467 Optics. Light |
Divisions: | Faculty of Information Science and Technology (FIST) |
Depositing User: | Ms Nurul Iqtiani Ahmad |
Date Deposited: | 02 May 2024 04:13 |
Last Modified: | 02 May 2024 04:13 |
URII: | http://shdl.mmu.edu.my/id/eprint/12398 |
Downloads
Downloads per month over past year
Edit (login required) |