Research And Improvement Of Iris Localization And Recognition Algorithm

Compared with other biometric features, iris recognition has many advantages. By means of iris recognition technology, it has been widely concerned by the academic and business circles, and the research and improvement of the algorithm has become a hot spot in the new generation of biometric identification. The algorithm of position algorithm and feature point extraction and matching in iris system is the focus of this paper.In terms of the iris location problem, the algorithm currently used has algorithm Daugman, Wildes algorithm and the Chinese Academy of Sciences Wang Yunhong and academician Tan Tieniu proposed iris location algorithm. Each algorithm has its advantages, but there are some aspects that can be improved. In this paper, the method is based on the traditional level set model, the first improvement is to select the initial level set, this paper put forward in the selection of the initial level set with the Hough method produced, seemingly in trouble, but can greatly reduce the number of iterations of the algorithm, and some extent reduce the effect of light on the positioning results. The second improvement is the introduction of interference factors weighting factor in its iterative equation, the benefits of doing so is to make the anticipation and removing of the eyelashes and eyelids and other interference factors.In the feature extraction and matching of, this paper uses the SURF algorithm instead of the traditional feature extraction algorithm, using the algorithm of feature points can be extracted. Compared with SIFT, its speed is about 3 times faster, and is not sensitive to light, to a certain extent, reduce the influence of light on the iris image feature extraction. In this paper, and uses the sub region segmentation weighted fusion. And in the regional distribution of weight on the POS speed training method, the weight allocation is more reasonable. The segmentation method can be weighted feature point location of iris distribution more fully utilized.For above proposed two improved this paper through corresponding experiments and the commonly used iris location and recognition algorithm are compared. The results show that the method proposed in this paper compared to the previous methods in positioning effect is more ideal, the feature points matching the accuracy, robustness and matching speed has improved to a certain degree.