Research And Implement Of Iris Recognition Algorithm Based On Deep Learning

Biometrics technology is an identity authentication technology based on analyzing and studying the differences in physiological characteristics or behavioral characteristics between different biological individuals.Iris recognition technology is considered to be the most promising biometric technology due to its uniqueness,stability,and anti-counterfeiting properties.This thesis analyzes and studies the problems encountered in the iris region location,iris authenticity identification,iris feature matching and related algorithms.The main researches are summarized as follows:(1)Aiming at the problems that the existing iris positioning algorithm is easily interfered by information such as spot and glasses edge,resulting in the failure of iris positioning,an improved edge location algorithm is proposed.By weighting the Hough transform results and establishing the weight calculation model,the influence of the irrelevant edge information on the positioning process is reduced.The pupil spot and the glasses interference are effectively reduced,and the iris positioning accuracy is improved.(2)Aiming at the problems of iris recognition system being deceived by counterfeiting iris information,a multi-block threshold classification convolution neural network model for iris authenticity identification is proposed.By segmenting the iris data to sub-blocks,and individually judging each sub-blocks for authenticity,the authenticity identification of the iris is performed according to the threshold value and the authenticity judgment result of the sub-block.The recognition accuracy of the neural network model is improved.(3)For the iris feature recognition problem,an automatic learning model of iris features based on convolutional neural network is proposed.In order to reduce the cost of neural network training,the convolutional neural network model is trained using the extracted ROI information.After training,a network model that can automatically learn the iris features is obtained,combined with the Euclidean distance matching algorithm for recognizing iris features.Proposed algorithm has reliable guarantee on the recognition accuracy and stability,can effectively reduce the cost of network training and recognition.(4)Combined with the algorithm proposed in this thesis,a prototype system for iris recognition is designed and implemented,and the system function is tested and verified.The results show that improved algorithm for iris recognition in this paper can accurately distinguish real iris and fake iris,effectively guarantee the safety of the iris recognition system,and improve the performance of the iris feature recognition.