| dc.description.abstract | The human eye is one of the identifying characteristics of an individual, especially the iris part. Human iris has a consistent pattern and more reliable than the other biometric subjects such as face and fingerprint which change over time. This research uses the CASIA iris image data. The system uses an automatic segmentation based on threshold to localize the iris collarette and normalize the results to constant dimension using Daugmans rubber sheet model by remapping each point within the iris region to a pair of polar coordinates. The features are extracted using the 1D log-Gabor filter produce a template which will be used as features for the classification process. The template data will be classified by using One-against-all Support Vector Machine (SVM) method. The data are split into five subsets and then alternately used for training data and testing data. These method used three kinds of kernel: Linear kernel, Polynomial kernel, and Radial Basis Function (RBF) kernel. The tests are performed on a dataset of the left eye, a dataset of the right eye, and the combination of both eyes. The best iris recognition accuracy of the dataset for the right eye is at 99%, for the left eye is at 100%, and for the combination of both eyes is at 100%. According to these results, it can be concluded that if there is only one iris image for each individual, then the SVM method is appropriate to be used for iris recognition. The training time required for the dataset of the left eye, the dataset of the right eye, and the combination of both eyes are 0.3368 seconds, 0.3366 seconds and 0.6857 seconds, respectively, while testing time required for a dataset of the left eye, a dataset of the right eye, and the combination of both eyes are 0.0361 seconds, 0.0364 seconds and 0.0638 seconds, respectively. | en |
