Iris verification method and system based on iris circumscribed rectangular graph

Abstract

The invention discloses an iris verification method and system based on an iris circumscribed rectangular graph, and belongs to the field of iris verification. The method comprises the following steps: a training stage: training an iris feature extraction network by adopting a training sample set formed by iris circumscribed rectangular diagrams of different individuals; an iris verification stage: acquiring an identity label and an iris image of a user to be verified, positioning the iris image to obtain an iris external rectangular graph, and inputting the iris external rectangular graph into the trained iris feature extraction network to obtain an iris embedding vector to be verified; determining an input iris embedding vector of the user according to the identity identifier of the user to be verified; and comparing the to-be-verified iris embedding vector with the input iris embedding vector, and judging whether the to-be-verified iris embedding vector and the input iris embedding vector belong to the same individual or not. The iris segmentation robustness problem is converted into the neural network feature extraction problem, and the neural network feature extraction robustness problem is solved through neural network training, so that the iris preprocessing complexity is reduced and the iris verification precision is improved by adopting the circumscribed rectangle positioning of the iris excircle.

Description

technical field [0001] The invention belongs to the field of iris verification, and more specifically relates to an iris verification method and system based on an iris circumscribed histogram. Background technique [0002] With the continuous development of global informatization and digital technology, technology brings convenience to society, but also brings information security issues. In order to protect the safety and reliability of information, more reliable authentication methods are needed. Traditional authentication methods use special items (such as personal documents) or special passwords (such as user IDs and passwords) to verify the legitimacy of users. These methods are simple and easy. realized, and thus widely used. However, these methods have many limitations. Special items are easy to be lost and forged, and special passwords are easy to be stolen. Therefore, traditional identification technology can no longer meet the needs of life. In order to solve th...

AI Technical Summary

Problems solved by technology

[0004] However, unlike other computer vision tasks based on deep learning, especially face recognition, there are not many works that make full use of deep learning to measure iris features. In recent years, some people have proposed to use deep neural networks for iris recognition, such as DeepIrisNet model, but these models require fine segmentation of the original iris area, and most of them use classification models instead of depth measurement models, which require a large number of data sets for support and are easy to overfit, difficult to train, and Not robust enough for complex common scenarios

Images