Research On The Key Technology Of Dynamic Iris Image Feature Extraction

In the current society,iris recognition technology has become a very important branch in the field of biometrics.Iris recognition technology research in social life,information security and other fields has great significance.At present,the research object of iris recognition is mostly static iris image under the ideal state.However,in daily life,iris image is not acquired under the static state.The iris image acquired by the iris collection device in motion may be blurred due to movement,defocus and other factors,and may also be interfered by sundries such as light spots and eyelashes.Therefore,the feature extraction is much more difficult than the ideal state.In this paper,dynamic iris image is taken as the research object,and the key parts of dynamic iris image feature extraction,such as dynamic iris image definition evaluation,dynamic iris image location and dynamic iris feature extraction,are studied and improved.This paper mainly focused on the three key stages to do the following work:(1)In order to study the key technology of dynamic iris feature extraction,a dynamic iris collection system was built to collect dynamic iris images and a small dynamic iris image library was established.(2)In the definition evaluation stage,the traditional image definition evaluation method based on gradient is first introduced.Then,aiming at the problem that dynamic iris image is affected by defocus blur and dynamic blur,a gradient-based evaluation method for the resolution of dynamic iris image without reference is proposed: The resolution value of the dynamic iris image is obtained by using the method of local region segmentation to evaluate the resolution of the unreferenced image.And the iris image with the highest resolution is selected as the object for subsequent processing.Experiments show that this algorithm can accurately evaluate dynamic iris resolution and is more efficient.(3)In dynamic iris localization stage,several traditional iris localization methods are first introduced,then distance regularized level set model is applied to the iris localization.A method for locating the inner edge of iris based on geometric feature-distance regularization level set evolution is proposed: in the coarse positioning phase,according to the physiological characteristics of iris,a binary image secondary projection method is proposed and carries on the coarse positioning within the iris on edge.Then,an improved distance regularization level set model is used to further locate the inner edge of iris.For iris outer edge localization,the method of Canny edge detection and Hough transform is adopted.Finally,the elastic model is used to transform the positioned annular iris into a rectangular region of the same size and enhance the image to makes the iris texture clearer.The experimental results show that the accuracy and efficiency of the dynamic iris interior edge localization algorithm proposed in this paper are greatly improved compared with the traditional methods;the algorithm of iris outer edge location adopted in this paper is also a feasible and effective algorithm.(4)At the feature extraction stage of dynamic iris,everal traditional feature extraction methods are introduced at first.Then three Sn-LBP feature extraction algorithm is proposed.After experimental analysis,the Sn+equivalent model LBP algorithm which can be used for feature matching is selected for feature matching experiment,and cosine similarity is used to evaluate the similarity between different iris features.The results show that this algorithm is more accurate than the traditional LBP algorithm.