Research On Iris Feature Expression And Recognition Algorithm Based On Wavelet And CS-LBP Fusion

With the introduction of new information technologies,high-tech conveniences can be seen everywhere.However,the following information security issues have also received widespread attention.Traditional identity encryption methods have been unable to meet people's needs for high security,so society urgently needs to obtain a fruitful identity authentication method.The new technology for identity authentication based on the unique biological characteristics and behavioral habits of the human body is called biometric identification,which has very high security characteristics.Compared with other biometrics,iris recognition has been widely used in biometrics because of its unique high accuracy,easy collection and good stability.In recent years,more and more iris recognition products have been used in airports,access control,banks and other places that require high security and confidentiality.Iris recognition has two working modes: one-to-one authentication mode and one-to-many recognition mode.In this article,we only research on one-to-one authentication.The iris recognition system is mainly composed of a series of processes such as preprocessing,feature extraction and recognition.The images used in this thesis are all qualified images,and this article mainly studies the feature extraction and recognition part in the later stage,the main purpose is to improve the performance of iris recognition system.In order to compensate for the instability of single iris feature recognition and enhance the versatility of the iris recognition system,the center symmetric local binary pattern(CS-LBP)and Hall wavelet were used to extract the iris image texture in feature extraction The features are binary coded,that is,valid information is collected in the spatial and frequency domains of the iris image,and the similarity distance corresponding to the two features is calculated using the Hamming distance.In recognition,XGBoost is used to achieve multi-feature similarity fusion.Since XGBoost is a machine learning model based on boosted trees,it has a good classification effect for non-linear problems.At the same time,the model itself supports parallel computing,making it suitable for large data A large number of iris samples can be identified quickly.In the paper,Harvard wavelet and CS-LBP are used for feature extraction and binary coding,and intra-and inter-class similarity obtained after Hamming distance matching is input into the XGBoost model as feature vectors,and the intra-class label values are used.The difference is 0 and the inter-class value is 1.The XGBoost model is used to train the sample data,and the test iris is identified based on the trained model.In this paper,the iris image database of CASIA-V1 and JLU-IRIS 6.0 of the Chinese Academy of Sciences is used to evaluate the algorithm performance,and the correct recognition rate(CRR),equal error rate(EER)and ROC curve are used as evaluation indicators.Through experimental results and comparative experiments,it is proved that the iris recognition algorithm proposed by XGBoost and implemented by multi-feature fusion of Haar wavelet and CS-LBP in this paper has obvious advantages in applying to a large database of iris samples.It not only improves the accuracy of iris recognition,but also reduces the instability of single iris features,and has good performance.