A Fast And Accurate Biometric Identification System Based On Retinal Vasculature Network

The advancement in automation technologies and the internet of things have completely revolutionized the concept of smart cities.While technological advances in automation have made life easier,it has introduced major security threats as well.In order to cope up with the increasing rate of e-commerce fraudulent activities and identity theft,the need for highly secure automated identification systems has now become inevitable.Though the conventional biometric systems have provided the medium level of security,they all are subject to variations and are susceptible to forgery.Hence,the permanence required for biometric identification can be breached.We developed a high performance fast and accurate person identification system based on the retinal vasculature network for the multi-sample data set.The retinal vasculature pattern is unique in all the individuals.Even the pattern varies within the eyes of the same individual.To achieve a high recognition accuracy,unlike the existing techniques that use a single segmentation method,a novel hybrid segmentation approach is proposed.The proposed hybrid approach used different segmentation techniques to deal with the variations in width of the retinal vessels.The thin and thick vessels are dealt independently with no compromise between the two.Consequently,when both the vessels are retained,segmentation accuracy is improved.As the proposed method is reliant on the retinal vasculature,so the improved segmentation ultimately enhances the recognition accuracy of our proposed retinal identification system.The recognition accuracy is further improved by using a combination of end points,bifurcations,and crossing over.This proposed feature set provided the maximum discriminant power and completely eliminated the overlap between different classes.Unlike the existing methods that focused only on the recognition accuracy,the proposed method also worked on identification process acceleration and computation time reduction.As biometric systems are to be deployed in real time,so they must be computationally accelerated.To work on this research gap,our method used Principal Component Analysis(PCA)based feature processing approach.PCA projects the extracted features into a subspace.The dimensionality reduction attained by PCA significantly accelerates the identification process.Hence,the proposed method is not only accurate,but is also time efficient and is computationally inexpensive.Experimental evaluation is performed using DRIVE,STARE,VARIA,RIDB,HRF,Messidor,DIARETDB0,and a large muti-sample per subject database created by the images provided by Dr.Chen(Shanghai Jiao Tong University Affiliated Sixth People Hospital).The experimental evaluation demonstrated that our method outperformed the existing techniques by achieving a recognition accuracy of 99.40% with an Equal Error Rate(EER)of zero.The segmentation accuracy of 99.65% is achieved.