Enhancement of Iris Recognition System using Deep learning

With the growing demand for security and accurate personal authentication, as well as the new dimensions in security issues confronting the world today, a dependable and secure authentication system is essential. The iris has become a popular biometric technique because it is a highly protected internal organ of the body. Furthermore, it is very hard to surgically change it without causing significant damage to the iris. The total accuracy of the iris recognition system is determined by the iris segmentation process's performance. This paper presents a deep learning technique for improving iris segmentation performance. The proposed method employs an efficient deep learning technique (SegNet), which performs joint semantic segmentation of ocular qualities (iris and pupil) with greater accuracy in unconstrained scenarios. These difficult circumstances limit the performance and dependability of ocular segmentation structures. Segmentation begins by denoising the pristine image with a deep convolutional neural network to address these issues (DCNN). The semantic segmentation of the iris and pupil is then accomplished with the help of a densely connected fully convolutional encoder-decoder network. Finally, for feature extraction and classification, the proposed system is implemented using a pre-trained convolutional neural network (Alex Net). The iris recognition system's performance is evaluated using five public databases: IITD, iris databases CASIA-Iris-V1, CASIA-Iris-V2 device 1, CASIA-Iris-V2 device 2, and MMU iris database. The results show that the proposed system has a high accuracy rate of 94.08 percent, 84 percent, 97.31 percent, 100 percent, and 97.7 percent, respectively and has time of execution of less than or equal to 2 minutes