Research On Related Technologies Of Iris Recognition

With the development of Internet technology,more and more people are aware of the importance of information protection.Therefore,identity has also entered all aspects of society.Traditional password authentication,such as digital passwords,character passwords,and even numeric and character combination passwords,can't satisfy people's requirements for privacy protection.Therefore,biometric identification has become more and more applicable.Biometric technology is unique to everyone and has reliable identity information.Common fingerprint recognition,face recognition,and iris recognition include palmprint recognition and human vein recognition.Among them,iris recognition has more obvious advantages than other biometrics,fast speed,strong stability,non-contact non-invasiveness,and more unique biometric information of other biometrics.Iris recognition consists of image acquisition,quality evaluation,pretreatment,pupil iris positioning,normalization,feature extraction and matching.This paper is based on existing research,expands or improves some algorithms,and changes according to specific algorithm tasks.Therefore,the innovative work is as follows:The iris image is pre-processed before the pupil positioning,and the approximate position of the pupil is determined by the pixel peak and valley.The centroid is combined with the statistical characteristics to cut the iris and remove the unwanted noise.In addition,through the upper and lower eyelids of the eye,the eyelashes,eyebrows and other noises are removed to prepare for pupil positioning.In the aspect of the coarse hole positioning,the maximum inter-class variance algorithm is used to obtain the pupil points of the pupil boundary after binarization and the morphological method to obtain the pixel points of the pupil boundary,so that the relatively accurate pupil can be fitted according to the least squares method.Accurate positioning of the pupil Based on the previous work of the pupil,using the characteristics of the pupil roundness,a movable,circular radius window variable filter is designed to filter the iris image.When the filter detects a pupil and the filter window radius is the same as the pupil radius,the pupil center position can be determined.The method is faster,more efficient and more accurate after pre-processing such as row and column cutting and down-sample.It is a reliable method of pupil positioning.In iris localization,this paper proposes a simple pupil-based iris localization method that uses the pixel curve trough to determine the iris radius.In addition,in the accurate pupil positioning,the improved Viterbi algorithm is used to calculate the optimal path of the iris closed loop.According to the acquired pixel points,the iris circle is accurately fitted by the least squares method,and the precise positioning of the iris edge is realized.This paper also proposes a normalized expansion method based on iris expansion and contraction.The method uses the exponential and logarithmic functions to simulate the stretching and expansion of the iris,and it is carried out at intervals to realistically simulate the physiological condition of the iris.In terms of iris matching,based on the gray image of iris,this paper extracts 12 feature blocks from the normalized iris image,and then uses the normalized correlation coefficient matching method to match the segment feature image with the database image.Use the matching threshold to determine if the match is successful.This method has a high accuracy because of the many features extracted.And because the feature block is smaller,the matching rate is higher.This paper also proposes a Weight-based Feature Fusion(WFF).The algorithm is developed based on Gabor filtering feature extraction,encoding and recognition.It uses multi-view,multi-scale and multi-modal multi-data features formed by Gabor parameters to solve multiple texture features generated by encoding.The weights are such that redundant information and noise are eliminated,and valid classification information is retained.The core idea of the model is to classify similar samples as much as possible,with different samples being separated as much as possible.The particle swarm optimization algorithm is used to iteratively solve the WFF model.Finally,the algorithm converges to obtain a set of optimal solutions.Then using this set of weight combinations can effectively extract the features of the most classified information.Thereby determining matching parameters,improving matching efficiency and recognition accuracy.