(50-3) 15 * << * >> * Russian * English * Content * All Issues

  Full text (PDF)

DOI: 10.18287/COJ1730

Article ID: 1730

Language: English

Abstract:
Nowadays, biometric authentication techniques are used widely and an autonomous solution requiring on-site decision-making without transferring data to a server or cloud for processing is increasingly required. In this paper, three algorithms for authentication based on scleral vessels are proposed. The algorithms are based on segmentation of the scleral region and scleral vessel pattern using the lightweight HalfUNet neural network. A distinctive feature of the first two algorithms is the use of Dense-SIFT and BRIEF descriptors for feature extraction and the k-nearest neighbor method for their comparison. Both algorithms have high classification accuracy and recognition probability for a given level of erroneous admissions of unauthorized persons by the system. The classification accuracy for the first two algorithms is AUC = 0.986 and 0.981 and the average error probability is EER = 0.038 and 0.033. These algorithms are not demanding on computing resources and can be used in devices with energy resource limitations. In the absence of limitations on the computing power, the third algorithm can be used, which is distinguished by the use of neural network algorithms SuperPoint and SuperGlue for searching for descriptors and subsequent comparison of features and, due to this, has a higher classification accuracy and a lower probability of errors (AUC = 0.998; EER = 0.022).

Keywords:
biometric authentication system, feature vector, scleral vessel pattern, scleral segmentation, scleral vessel segmentation, Half-UNet neural network.

Acknowledgements:

Citation:
Chernyadyev SY., Kharina NL, Medvedeva EV. Segmentation of sclera vessels for a biometric authentication system. Computer Optics 2026; 50(3): 1730. doi: 10.18287/COJ1730.

References:

1. Doke K, Ingle DR. A closer look at sclera: emerging trends in biometric security. In: Proc 2nd Int Conf Comput Commun Control (IC4); 2024. doi:10.1109/IC457434.2024.10486207.
2. Lucio DR, Laroca R, Severo E, Britto AS, Menotti D. Fully convolutional networks and generative adversarial networks applied to sclera segmentation. In: Proc IEEE 9th Int Conf Biometrics Theory Appl Syst (BTAS); 2018. doi:10.1109/BTAS.2018.8698597.
3. Das S, De Ghosh I, Chattopadhyay A. An efficient deep sclera recognition framework with novel sclera segmentation, vessel extraction and gaze detection. Signal Process Image Commun 2021; 97: 116349. doi:10.1016/j.image.2021.116349.
4. Vitek M, Štruc V, Peer P. GazeNet: a lightweight multitask sclera feature extractor. Alexandria Eng J 2025; 112: 661-671. doi:10.1016/j.aej.2024.11.011.
5. Lee S, Low CY, Kim J, Teoh ABJ. Robust sclera recognition based on a local spherical structure. Expert Syst Appl 2022; 189: 116081. doi:10.1016/j.eswa.2021.116081.
6. Rot P, Vitek M, Grm K, Emeršič Ž, Peer P, Štruc V. Deep sclera segmentation and recognition. In: Advances in computer vision and pattern recognition. Cham: Springer; 2020. pp. 395-432. doi:10.1007/978-3-030-27731-4_13.
7. Vitek M, Bizjak M, Peer P, Štruc V. IPAD: iterative pruning with activation deviation for sclera biometrics. J King Saud Univ Comput Inf Sci 2023; 35(8): 101630. doi:10.1016/j.jksuci.2023.101630.
8. Dimauro G, Camporeale MG, Dipalma A, Guarini A, Maglietta R. Anaemia detection based on sclera and blood vessel colour estimation. Biomed Signal Process Control 2023; 81: 104489. doi:10.1016/j.bspc.2022.104489.
9. Maheshan MS, Harish BS, Nagadarshan N. On the use of image enhancement technique towards robust sclera segmentation. Procedia Comput Sci 2018; 143: 466-473. doi:10.1016/j.procs.2018.10.419.
10. AlRifaee M, Almanasra S, Hnaif A, Althunibat A, Abdallah M, Alrawashdeh T. Adaptive segmentation for unconstrained iris recognition. Comput Mater Continua 2024; 78(2): 1591-1609. doi:10.32604/cmc.2023.043520.
11. Das A, Kaur A, Zhang J, et al. Sclera segmentation and joint recognition benchmarking competition: SSRBC 2023. In: Proc IEEE Int Joint Conf Biometrics (IJCB); 2023. doi:10.1109/IJCB58213.2023.10332519.
12. Rajan G, Sharma S. Sclera segmentation techniques. Int J Recent Technol Eng (IJRTE) 2020; 9(2): 1072-1076. doi:10.35940/ijrte.B4066.079220.
13. Vitek M, Rot P, Štruc V, Peer P. A comprehensive investigation into sclera biometrics: a novel dataset and performance study. Neural Comput Appl 2020; 32: 17941-17955. doi:10.1007/s00521-020-04782-1.
14. Chernyadyev SY, Kharina NL, Prozorov DE, Zemtsov AV. Segmentation of the eye sclera for biometric authentication systems based on the lightweight neural network HALF-UNET [in Russian]. Ekonomika i Kachestvo Sistem Svyazi 2024; 4: 1-10.
15. Lu H, She Y, Tie J, Xu S. Half-UNet: a simplified U-Net architecture for medical image segmentation. Front Neuroinform 2022; 16: 911679. doi:10.3389/fninf.2022.911679.
16. Taghanaki SA, Abhishek K, Cohen JP, Cohen-Adad J, Hamarneh G. Combo loss: handling input and output imbalance in multi-organ segmentation. Comput Med Imaging Graph 2019; 75: 24-33. doi:10.1016/j.compmedimag.2019.04.005.
17. DeTone D, Malisiewicz T, Rabinovich A. SuperPoint: self-supervised interest point detection and description. In: Proc IEEE/CVF Conf Comput Vis Pattern Recognit Workshops (CVPRW); 2018. pp. 224-236.
18. Sarlin PE, DeTone D, Malisiewicz T, Rabinovich A. SuperGlue: learning feature matching with graph neural networks. In: Proc IEEE/CVF Conf Comput Vis Pattern Recognit (CVPR); 2020. pp. 4938-4947. doi:10.1109/CVPR42600.2020.00499.
19. Iandola FN, Han S, Moskewicz MW, Ashraf K, Dally WJ, Keutzer K. SqueezeNet: AlexNet-level accuracy with 50× fewer parameters and <0.5 MB model size. arXiv preprint 2016; arXiv:1602.07360.

---

151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail: journal@computeroptics.ru; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20