A Multi Biometric System Based On The Right Iris And The Left Iris Using The Combination Of Convolutional Neural Networks

摘要

Biometrics to revolutionize the world of IT security. Many companies and governments and multinational corporations are opting for this technology to secure their data. In recent years we have seen the appearance of the biometric passport, the biometric driver’s license. Researchers and industrialists in the field use multimodal biometric recognition systems to increase the security and robustness of the system. Single-mode biometric systems have made their contributions, but the use of only one modality makes the system vulnerable. Biometrics professionals opt for a multimodal system. The combination can do between the following methods: facial recognition, fingerprint, iris, voice recognition, signature, etc. The biometric identification system for people in India (launched in 2009) [39] includes several steps for authentication. First, he asks for recognition of the iris, then fingerprints of the ten fingers, and finally facial recognition. In this article, we propose a system of iris recognition by the classification and the combination of the right iris and the left iris. All this done through the use of Deep Learning technology, the experiments carried out on three architectures of convolutional neural networks (CNN): VGG16 [31] [34], DenseNet169 [32] [35], Resnet50 [33] [36] with the MMU1 [37] database. We obtained excellent results with an accuracy of 100 % for the combination of the architecture ResNet50 of right iris and DenseNet169 of the left iris.