Research On Gender Classification Of Iris Images Based On Deep Learning

Biometric identification technology is the identification of people using high-tech about computer and biosensors.With the continuous maturity of technology,biometric identification has attracted more and more people's attention and favor in the fields of information security,security,and payment.Among several commonly used biological characteristics,iris recognition has aroused great interest of researchers.Compared with other biometric identification technologies,iris has a series of advantages such as more complex and better anti-counterfeiting.Recognizing a person's gender from iris images has broad application prospects in various fields such as identity verification and security monitoring.However,according to previous research work,the accuracy of gender classification based on iris images alone is not high.This paper is based on deep learning to classify the gender of iris images.And the feature of the eye image is extracted and fused with the feature of the iris image,and then the features of the fusion are classified by SVM to improve the accuracy of iris classification.The research content and innovations of this article are as follows:Aiming at the difficulty of iris image acquisition,the previous research work caused the problem of low recognition accuracy due to insufficient iris data.This paper proposes a data enhancement method to flip and crop the iris image to expand the iris data set.,Which solves the problem of small sample training prone to overfitting.Aiming at the problem of low recognition accuracy due to insufficient iris data in previous research work,this paper proposes a data enhancement method to flip and crop the iris image,expand the iris data set,and solve the small sample training The problem of over-fitting is prone to occur.Using Res Net-Iris network,combined with transfer learning,an iris image gender classification model is trained on the preprocessed iris image data set.Finally,the gender to which the iris belongs is predicted,and the accuracy rate of gender prediction reaches 94.6%.According to previous researchers' work on iris gender classification,the accuracy of gender classification using only iris images is not ideal.This paper proposes to fuse the features extracted from the eye image into the iris image features,through feature fusion,and then classification,and finally get The accuracy of gender classification is 98.9%.A set of large field of view iris gender classification system is designed.Based on the trained iris and the fusion classification model,the acquired images are classified by gender in real time through the camera.Due to external disturbances,the accuracy rate of iris classification in the real-time test of gender classification on the system is 90%,and the accuracy rate of image classification after feature fusion is 96%.