Research And Implementation On Retina Vessel Recognition Technology

With the development of e-commerce,mobile payment and online shopping,the requirement of virtual data security is increasing for people.The traditional verification technology such as account and password can no longer satisfy the need for security,and benefit from its reliability and difficulty to counterfeit biometric technology has been applied to the authentication and payment system of many kinds of civil electronic equipments gradually such as telephone,computer and etc.Among the biometrics,retina is one of the most reliable,stable and difficult to forge biometric technologies,so it is pretty suitable as a identification.And in the near future,it would probably be applied to civil fields that demand high security,such as online payment,access control and ATM.Therefore,the research of retina recognition has a great value and broad prospects.This thesis has researched into many dissertations about retina and b iometric technology at home and abroad,especially in image blood vessel segmentation,feature extraction and matching.And we picked,integrated and improved the methods from those paper,designed an accurate and stable retina recognition system.The main work and achievement as follows:(1)In retina image segmentation,we have researched and compared many kinds of blood vessel image segmentation algorithms.Finally,we used the Gaussian filter to enhance retina blood vessel image by incorporating the feature of retina blood vessel,and using two-dimensional maximum entropy threshold to segment blood vessel.Experimental results show the method has better performance in blood vessel segmentation than other frequently used edge segmentations,and has good anti-noise ability.(2)In feature extraction,according to the research of blood vessel,we choose branch points and cross points as feature points.And tinning retina blood vessel images by using the knowledge of morphology.Then we have analyzed the Harris corner feature extraction algorithm and neighborhood feature extraction algorithm in detail,and confirmed neighborhood feature extraction is the best way to extraction method for the skeleton image by compared the results of these two algorithms.Besides,we used further extraction to filter out the pseudo-feature and improved accuracy of extraction.(3)In feature matching,we have compared several common used feature matching methods,and weighed the pros and cons of these algorithms by standards of stability,accuracy,efficiency and so on.Finally,we designed a fusion algorithm based on analyzing and comparing.It combine the triangle matching algorithm with two-dimensional clustering algorithm to make full use of the advantages of the two methods and make the matching algorithm have stability and efficiency.Besides this,we have improved the efficiency of matching by fiddling with selection method of template triangle and triangle retrieval algorithm.