Research On Iris Recognition Algorithms Based On Deep Learning Network Model

In today's information explosion society,more and more personal information is uploaded to the network or mobile phones,personal information security has received more and more attention,people pay more and more attention to the security of personal information,and people have higher and higher requirements for the security and accuracy of identity recognition.Traditional identity authentication technologies,such as password login,have been unable to meet the current requirements for personal information security.In such a social context,biometric recognition technology stands out with its security,uniqueness and accuracy,and becomes the focus of the security identification field in the era of information explosion.In the field of biometrics,iris recognition technology has become a key research direction in this field because of its advantages in various aspects.In recent years,with the rapid development of electronic hardware devices,in-depth learning has become a widely used technology in the field of image processing.The deep learning network model becomes deeper and wider with the development of hardware devices.Iris recognition essentially belongs to the field of image recognition,so the application of deep learning network model to iris recognition has theoretical basis.Iris recognition system includes iris image acquisition,iris image quality evaluation,iris image location,iris image normalization,iris image enhancement,iris feature extraction,iris feature recognition.This paper focuses on the research of iris feature extraction and iris feature recognition,combining with in-depth learning network model.Based on the theory of deep learning network model,this paper proposes a novel deep learning network model,MultiModel_DenseNet,which is suitable for iris recognition.This paper improves the traditional DenseNet network model,changes the traditional input model of DenseNet network model and the detailed structure of the network model,makes it more suitable for the field of iris recognition,and then proposes the MultiModel_DenseNet network model.The network model integrates the iris feature extraction and iris feature recognition of traditional iris recognition system.More importantly,the traditional iris recognition algorithm only uses the spatial or frequency domain information of iris image,while the MultiModel_DenseNet network model not only inputs the spatial information of iris image,but also adds the frequency domain information of iris,which enlarges the amount of input information of the network,improves the flow of iris information in the network model,and is conducive to improving the ultimate iris recognition.Accuracy.In the experimental part,this paper compares the effect of applying the traditional DenseNet network and the newly proposed MultiModel_DenseNet to iris recognition.The experimental results show that the training time of the MultiModel_DenseNet network model is obviously less than that of the traditional DenseNet network model;in terms of stability,the MultiModel_DenseNet network model is more stable,while the oscillation phenomenon of the traditional DenseNet network model is very obvious;in terms of the final accuracy,the accuracy of the MultiModel_DenseNet network model is obviously higher than that of the traditional DenseNet network model.Network model.By comparing the experimental results,it can be concluded that the new network model proposed in this paper has more advantages than the traditional DenseNet network model in iris recognition.In this paper,not only a new iris recognition algorithm is proposed,but also a new idea of applying the deep learning network model to the direction of iris recognition is proposed.