Iris Segmentation And Recognition Algorithms For Biometric Certification

With the rapid development of mobile devices and social networks,more and more identity authentication and information security-related activities are involved.Traditional authentication methods cannot meet the growing demand for information security due to the drawbacks including being forgotten,lost,forged and stolen easily.Therefore,bio-certification technologies based on biometrics such as fingerprint,face,etc.have emerged.Compared with other biometrics,iris contains rich and unique texture,which is unique,stable,and easy to collect,serving as an important role in biometric authentication systems.However,there are still some problems with the existing iris segmentation and recognition algorithms.For example,many iris recognition systems are designed based on the iris images with good visual quality from the same database,which fail to consider the difficulty in heterogeneous iris segmentation and recognition.Based on this,the thesis focuses on the iris physiological structure characteristics and studies iris segmentation and recognition algorithms towards biometrics,showing important research significance and reference value in practice.In this thesis,some challenging problems in iris segmentation and recognition are studied.The main contents are as follows:(1)Aiming at the problem that the existing iris segmentation methods can be applied to specific iris databases,this thesis studies the traditional segmentation methods and proposes a heterogeneous iris image segmentation method based on the prior noise assessment and active contour model.The proposed method firstly classifies iris images into different groups according to their noise types.After that,the active contour model together with the "divide and conquer" strategy are utilized to segment iris images with diverse noises,improving the accuracy and robustness of heterogeneous iris segmentation.(2)Due to occlusion of eyelashes and the light reflection in iris images,some existing iris recognition algorithms fail to match the iris features successfully.To target this problem,an iris recognition algorithm based on local sector and code optimization is proposed in this thesis.By considering the occlusion caused by eyelids and eyelashes in iris images,the proposed algorithm utilizes the local sector sampling and encoding to deal with the noise disturbance.Following this,the recognition accuracy is improved by combining the iris characteristics and iris coding optimization.In summary,this thesis focuses on the challenging problems facing by existing iris segmentation and recognition and proposes novel iris segmentation and recognition algorithms based on traditional algorithms and iris characteristics.The proposed algorithms overcome the disturbance of noises and occlusion and achieve the effective segmentation and recognition of heterogeneous iris images.