Research On Identity Recognition Technology Based On Face And Iris Fusion

At present,traditional authentication and recognition methods(such as identity cards or passwords)have risks of forgetfulness and theft.Biometric identification methods use fingerprints,iris,face and voice to achieve identity authentication and recognition.These personal characteristics of physiology or behavior do not exist problems such as forgetting and easy to steal,which are safer,more convenient and more reliable than traditional methods.However,single biometric recognition is affected by some disadvantageous factors such as environment,noise and deception attacks in practical applications,which can not meet the needs of high security occasions.It is considered as an effective solution to integrate multiple biometric characteristics for authentication and recognition.Face and iris,as typical characteristics,play an important role in the field of identity recognition.However,the quality of iris image acquired by most iris acquisition devices is usually poor,including more noise and lower clarity,and part of the iris is often occluded by eyelid and eyelashes,resulting in unsatisfactory iris recognition effect.Face is vulnerable to changes in quality,posture,expression and illumination in adverse environments,and face acquisition equipment is more universal,compared with other biometrics.There is a high risk of duplication and imitation of physical characteristics.Therefore,in this paper,the fusion recognition of face and iris is the key research direction.The following two fusion recognition methods based on score and feature levels are studied and presented.Firstly,this paper studies and presents a dynamic weighted fusion recognition method based on quality assessment.First,combining a variety of algorithms to extract face and iris features can make up for the single algorithm to extract features from a single facet,and the defects of mining incomplete feature information.Second,biometric image quality is used to evaluate the quality of feature matching.Identity authentication information hidden in high-quality feature matching scores is fully excavated,which weakens the interference of low-quality feature matching scores on fusion recognition results and achieves dynamic weighted fractional fusion.The experimental results show that the proposed method enhances the robustness and accuracy of the system,and then improves the overall recognition performance of the system.Secondly,a feature level fusion recognition method based on deep learning model is studied and presented.First,aiming at the blindness and difference of traditional artificial design features,convolutional neural network is used to automatically mine the essential information of features,so as to carry out subsequent authentication and recognition.Second,parametric t-SNE algorithm is used to solve the problem of high feature dimension after multi-feature series fusion,and support vector machine is used to classify and recognize,so as to achieve high-quality fusion recognition.The experimental results show that the proposed method effectively improves the accuracy and robustness of the fusion recognition system.