Abstract:
Biometrics are the personal physiological and behavioural characteristics which are mostly used for personal recognition. In our project, we present a novel system based on vascular pattern of human retina which can be used for security purpose as well, along with identification. We have also introduced our own database which is composed of 100 images from 20 different individuals named as RIDB for retinal recognition. Our proposed system consists of three stages; i.e. Preprocessing, Feature Extraction and finally the Matching Process. In Preprocessing, it extracts the vascular pattern from input retinal image by first enhancing them with Gabor wavelets next they are segmented using Recursive Supervised Multilayered Thresholding Technique. Morphological Thinning Operation is applied in order make these vessels uniform, by reducing the width of every blood vessel equal to single pixel. Chosen features are Vessel Ending and Bifurcation Point. Second stage extracts all these possible feature points using Crossing Number Method and validates them by keeping the correct ones while discarding the false features through Windowing Technique. Then system forms feature vector for each feature point, which is calculating the orientation and distances with four nearest neighbours. These feature set are saved in the database. The proposed system matches the template feature vectors and input query image feature vector by calculating Feature Distance using Mahalanobis Distance. In Person Recognition, DRIVE, STARE, VARIA and RIDB databases are used. It is clear by the experimental results that high accuracy is obtained for vascular pattern segmentation and recognition by the proposed system. Validity of the proposed system is also proven by the results.
