Biometric Iris Authentication Using a Segmentation Network with Attention and a Key-Points Detection Algorithm

Biometric Iris Authentication Using a Segmentation Network with Attention and a Key-Points Detection Algorithm

연구 분야: Analysis

학회: International Conference on Artificial Intelligence and Soft Computing

Authorization of an individual within a given system stands as a pivotal element of access control. Hence, in this paper, we propose a system based on biometric features such as the iris of the eye. For this purpose, the user takes a photo of the eye, which is further processed as a graphic image. The first step is the extraction of the iris itself, which is carried out via the U-Net network with the attention module. Then, using the generated mask, the iris is extracted and subjected to comparative analysis. An algorithm for finding key points is used to this end. Depending on the overlapping number of detected features, the system grants or rejects authorization. Both the quality of the segmentation performed and the authorization possibilities in the proposed system were examined.