الفهرس | Only 14 pages are availabe for public view |
Abstract An automatic fingerprint identification/authentication system has been developed in the present work. It consists mainly of four stages: fingerprint image acquiring, image preprocessing, feature extraction and fingerprint matching. The application of appropriate image preprocessing techniques has enabled important information from the images to be restored and recognized: it is asserted that the use of dynamic block size window and combining inverse and Gabor filtering procedure makes it possible to deal with fingerprint images having distorted regions. To extract the fingerprint minutiae, binarized image versions were used. A set of directional morphological filters have been utilized for preprocessing the binarized image, then a thinning procedure was applied. The FVC2000 standard database was used to evaluate the developed system. For the matching stage a minutiae-based matching technique was adopted. The results have shown that the Equal Error Rate (EER) reaches 3.1%, 1.5%, 3.5% and 5% for DB1, DB2, DB3 and DB4 respectively, when using the inverse and Gabor filtering method with dynamic block size compared to 4.9%, 3.3%, 6% and 7% respectively, when using only Gabor filtering with constant block size. In an attempt to improve the system performance, two different data fusion techniques have been used. The results demonstrate that combining two impressions of the same finger or combining the images captured from two different sensors for the same fingerprint improve the system accuracy. It is concluded that the adopted image processing and pattern recognition methods are highly relevant to the fingerprint images and that the developed system is quite effective in identification and verifying person identity. |