Iris Image Classification Based On Deep Features Of Iris Texture

With the rapid development of the Internet and information technology,people pay more and more attention to the public security and biometrics.Iris is the annular part between the black pupil and white sclera of human eyes,which possesses rich texture information.The iris texture is highly discriminative and stable,which makes iris an important part of the human body for biometric identification.As the application population continues to expand,the size of the iris database is also dramatically increasing.Large-scale database can slow down the system response.Iris image classification is one of the main methods to solve the problem of large-scale iris recognition.So it is important to study iris image classification.Iris image classification methods are classified to several specific categories,like Iris liveness detection and race classification.There exist two problems in iris classification methods.On one hand,the existing methods for iris liveness detection mainly use manual designed features,while the type of fake iris images in practical applications is not completely known.On the other hand,the iris race classification mainly focus on the classification of Asians and non-Asians and cannot solve the problem of sub-ethnic classification.So this requires the algorithm have good generalization performance and the ability to learn better texture features from iris data.Therefore,this study proposes a novel iris image classification method based on deep features of iris texture.Firstly we feeds preprocessed iris images to a convolutional neural network to extract deep features as low-level features.We use a Gaussian mixture model to cluster the features to obtain iris texture textons,and the model is then used with Fisher vector to extract high-level features.A support vector machine is used for classification.In this paper,the features more suitable for different iris classification task are learned from the training data in a data-driven way.Over 99% accuracy was achieved on different single fake patterns datasets and hybrid fake pattern datasets.Experimental results show that the proposed method can achieve 91.94% on the Asian and non-Asian dataset.For the Han–Tibetan dataset,the proposed method can obtain 82.25% accuracy.At the same time,a new iris image database,which is suitable for sub-ethnic classification based on iris images is also established.It proved the feasibility of sub-ethnic classification based on iris images for the first time.