Research And Implementation Of Fast Iris Recognition Algorithm

With the development of the society and the progress of technology, For the safety of information, the problem of identity authentication appeares in all aspects of life. Authentication is confirming the identity by a variety of techniques or non-technical means. Biometric technology plays an extremely important role in improving information security. It can be widely used in many fields. Among biometric identification techniques, the iris identification technique has been developing rapidly in recent years. Compared with other biometric identification techniques, iris recognition has more excellent characters, such as universality, stability, uniqueness, non-invasive, collection. So there is a strong potential market value of it which based on the identity of the iris identification techniques has got academia and the business community’s attention increasingly.In this thesis, the development of biometric recognition and some kinds of biometric recognition techniques are introduced. Then the development of iris recognition technique and the structure of iris recognition system are discussed in detail. The algorithm of iris recognition has also been researched deeply and all of the processes that include location, Pre-processing, feature extraction, and pattern match have been carried out. Location and feature extraction have been focused on researching since they are the key steps of them.In the iris location respect, the iris location of hough transform is large computation and positioning slow. So this thesis puts forward an improved iris location algorithm based on Hough Transform.shrinking the original images several times, making use of the Hough transformation to extract iris rings, and restoring the images.In the feature extraction of iris texture, several typical iris feature extraction methods are introduced. They are extraction algorithm based on2D Gabor filter feature, Extraction algorithm based on multi-channel Gabor filter feature and the method based on One-dimensional wavelet transform zero-crossing detection. And in the experimental analysis, the three ways has been comaperd.Iris recognition algorithm has been implemented by using C language and simulated in the CASIA iris image database. To test the new iris location method, the result shows that Iris images minification should be22n(n=1,2,3,…) times and when the iris images mignification is16times, the calculating speed and recognition performance can achieve the best balance. To test three kinds of feature extraction, the result shows that Gabor filter feature extraction method is better than the other two methods. And the using of the real and imaginary parts simultaneously encode is better than just using one alone.