Método de reconhecimento pessoal através da íris usando funções geoestatísticas

Abstract

Biometrics identification methods are gaining applications each day and this has motivated a lot of research in this area. This work presents a proposal for a method to identify people through iris texture analysis using geostatistics functions and their combination. To achieve this work objective, it is considered the following phases: automatic localization of the iris, features extraction and classification. In the localization phase, it is used a combination of three techniques: Watershed, Hough Transform and Active Contours. Each technique has an essential function to achieve a good performance. Within the extraction phase, there were used four geostatistics functions (semivariogram, semimadogram, covariogram and correlogram) and a combination of them to extract this features with a good precision. Finally in the phase of classification it is used a Euclidean Distance to determine the similarity degree between the extracted features. The tests were realised for the phases of localization and classification using an iris database called CASIA that has 756 images. The results achieved by the localization method are about 90%. For the classification method, considering the tests realized with the authentication mode, the obtained results has reached a success rate of 97.02% for a false acceptance rate equal to zero and 97.22% for a false acceptance rate equal to a false rejection rate. The tests realized with the identification mode have reached a rate of success of 98.14%.