Research On Iris Feature Extraction And Recognition Algorithm Based On Improved Particle Swarm Optimization

Iris recognition technology has received increasing attention in academia and engineering because iris feature information have characteristics of uniqueness,stability,inviolability,high anti-counterfeiting and reliability.Iris recognition products have gradually been applied in the fields of access control system,airport security system,and confidential anti-counterfeiting system authentication,which have gradually become a widely used technology for identifying people after fingerprint recognition and face recognition.The traditional iris recognition system mode is mainly divided into hardware modules and software modules.In most cases,the software module is divided into the steps of iris quality evaluation,iris localization,iris normalization and enhancement,iris feature extraction and iris certification.In addition,the iris acquisition environment is complex and variable,so we divide the iris acquisition image into constrained iris and unconstrained iris.In order to improve the performance of iris recognition system under traditional cases,on the basis of the hardware acquisition conditions have been determined,it is necessary to improve each step algorithm of software module of iris recognition system for different situations.In this paper,we take the constrained ideal state to shoot iris as the research object.Based on the particle swarm optimization algorithm,we improve iris feature extraction and iris recognition algorithm for software modules in iris recognition system with common coding recognition mode.At the same time,based on the improved iris feature extraction and recognition algorithm,we propose improved algorithms for iris quality evaluation and iris localization.Improve the overall performance of the iris recognition system while improving the performance of a single step.The main content of this paper is as follows:1.Aiming at iris quality evaluation,we propose a sequence iris fast quality evaluation algorithm based on morphology and grayscale distribution.Perform sharpness,live detection,effective iris area detection,squint and super boundary detection for the constrained iris video stream collected by the infrared camera.It is judged whether the tester's iris can be used for the next iris recognition process to improve the image quality of the iris recognition information input.2.Aiming at iris localization,we use the methods of partitioning search,based on discipline change of gray scale and convolution amplitude in sub-blocks,the inner circle and the outer circle of iris are roughly positioned and finely positioned.And perform the normalization and enhancement operations.The algorithm is mainly used to accurately segment the iris,retain the characteristics of the iris as much as possible,and improve the ability of iris feature extraction.3.For the recognition of small scale iris library,we propose an iris feature extraction algorithm based on mutated particle swarm optimization Gabor filtering.Processing images with Gaussian Laplacian operator to reduce image noise and redundant interference.The mutation particle swarm optimization algorithm is used to optimize the parameters of Gabor filtering,and improve the ability of Gabor filtering to extract iris features.Finally,the Hamming distance is used to accurately identify the category of the test iris in the scale iris library.4.For the recognition of medium scale iris library,we propose a secondary iris recognition algorithm based on AM-PSO which consists of ant colony system and mutation particle swarm optimization.With the increase of the number of iris samples,the traditional Hamming distance iris recognition algorithm can not improve the accuracy of recognition.On the basis of the first recognition composed by Gabor filtering + Hamming distance,Haar wavelet+BP neural network are used for secondary recognition.The iris recognition range is reduced by the first recognition process,and the iris category is accurately authenticated by secondary recognition.Gabor filtering and BP neural network are adaptively optimized by ant colony mutation particle swarm optimization to improve the optimal value search efficiency.In this paper,we use the irises collected by the Institute of Automation of the Chinese Academy of Sciences and the Biometrics and Information Security Technology Laboratory of Jilin University for performance analysis.With the correct recognition rate,equal error rate,ROC curve and running time as the evaluation indicators,while analyzing the performance of a single algorithm,we also analyze the improvement effect of the overall performance of the iris recognition system,and then demonstrate the performance advantages of each single algorithm proposed in this paper and the overall iris recognition system in the traditional case.