AB081. Multi-unit ocular biometric system based on corneal shapes

Background: Our goal is to build a multi-unit ocular biometric system based on the fusion of left and right corneal shapes for identity authentication.

Results: For each individual, we had eight feature vectors (eight measures in two sessions) of size 36 (Zernike polynomial coefficients) from their corneal topographies. The experimental results on our cornea database constructed for this study showed encouraging performance of the proposed ocular biometric system with Equal Error Rate decreasing to 1.38% with the weighted-sum rule compared to the analysis of the left (4.5%) or right (3.7%)cornea alone.

Conclusions: The objective of this work was to investigate corneal topographyas an accurate biometric modality using shape discriminating features. Our idea was to propose an ocular multi-unit system based on the fusion of the left and right corneal shapes. The corneal feature extraction was done by Zernike polynomial decomposition. Multi-unit cornea fusion was performed at the matching score level to generate a unique score. This allowed a significative decrease of the EER to 1.38%.

Keywords: Ocular biometric; left cornea; right cornea; fusion; multi-unit system