| In the universal information era, the personal identity gradually becomes digitized and networked, which indispensably requires a reliable and convenient authentication scheme. Identification based on biometrics attracts more and more attentions by the various countries and many other fields in recent years, among them iris recognition has been widely accredited for both high accuracy and security. The iris recognition system includes four parts, image acquisition, iris localization, feature extraction and pattern matching. This paper puts more emphases on feature extraction and pattern matching. In this paper, we introduce the below aspects:1. Introduced the principle and process of local image characteristics matching algorithm, and analyzed the possibility of the local image characteristics used in iris recognition. Considering the particularities of iris image, we applied a lot of improvement and innovation to local image characteristics matching algorithm. These modifications not only overcame the limitation of time consumption and space occupation, but also kept its invariability advantages, which successfully introduced the algorithm to the real-time iris recognition system, feasibly.2. Introduced the online training concept for iris feature descriptor, made the iris recognition system self-optimizable and self-adaptive during continuous usage.3. Implemented the proposed algorithm in MATLAB, and ported it to cross-platform C++code, with considerable optimization. Having evaluated the proposed algorithm, its variants, and classic2D-Gabor algorithm with CASIA V3databases detailedly, the proposed algorithm is superior to the classic2D-Gabor algorithm, especially the excellent FMR1000performance nearly20times better.4. Designed a real-time iris recognition system based on the proposed algorithm for demonstration in C/S architecture, with realistic situation and user experience taken into account. The system was cross-platform and highly extensible. |
PDF Full Text Request