Multi-scene iris recognition method based on deep learning

Abstract

The invention discloses a multi-scene iris recognition method based on deep learning, and solves the technical problems that the robustness of a current iris recognition algorithm is poor, and the recognition capability of developing iris recognition under complex scenes and non-ideal imaging conditions is low. The recognition method comprises the following steps: shooting the whole eye of a person to obtain a to-be-recognized eye image; carrying out iris positioning segmentation to obtain an iris image to be identified; performing iris feature extraction to obtain a to-be-recognized feature matrix formed by iris features as elements; and carrying out iris feature matching, and if matching succeeds, passing identity recognition. In the iris positioning segmentation stage, jump connection is added and a UNet semantic segmentation model of cavity convolution is used to obtain a separated iris image, the robustness of iris positioning segmentation is improved, some tedious preprocessing steps needing a large number of experiments to determine parameters are omitted, meanwhile, upper and lower eyelids can be directly removed, two-step processing is not needed, and the generalization of the algorithm is high.

Description

technical field [0001] The invention relates to a recognition method, in particular to a multi-scene iris recognition method based on deep learning. Background technique [0002] The application of biometric identification technology has penetrated into every aspect of everyone's life under the situation that the trend of informatization is becoming more and more intense. This kind of technology that uses the physiological characteristics of some human beings to distinguish different individuals for identity authentication is extremely convenient. At the same time of our life, it also protects our privacy and property security. Common biometric technologies include: fingerprint, face, iris, vein, voiceprint, palmprint, etc. Among them, iris recognition technology is widely used in the field of biometrics because of its strong security, good stability and high accuracy. [0003] The iris is a pigmented connective tissue formed during human embryonic development and located b...

AI Technical Summary

Problems solved by technology

These algorithms mostly use the characteristics of the approximate circular shape of the pupil and the gray level difference between the two in terms of iris segmentation and positioning. Some parameters need to be obtained through repeated experiments, and the accuracy of recognition depends heavily on specific imaging conditions and image quality. , poor adaptive ability, weak robustness

In terms of iris feature extraction, the traditional method is highly dependent on manually extracted features. It needs to use human prior knowledge to process the original data and then classify the features. The classification results are highly dependent on the features and have poor robustness.

Images