Department of Information Systems Engineering
Permanent URI for this collection
Browse
Browsing Department of Information Systems Engineering by Author "BOSTAN, Atila"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item ENHANCEMENTS IN FINGERPRINT AUTHENTICATION(2017-06-02) Alsubaihawi, Mohammed Abdulraheem Taqi; BOSTAN, AtilaFingerprint identification and verification tasks are among the most challenging tasks in image processing and machine learning domains. Fingerprint processing presents a key issue in the biometric technologies and information security. According to the fraction of the people population based on the complete detection of the biometric fingerprint feature such as ridge structure, incomplete (portion) fingerprint image identification and verification task is very difficult to be accomplished. The main challenge in this problem is that the partial loss of the ridge structure in the incomplete fingerprint image. In this thesis, we studied the effectiveness of global feature approach in fingerprint identification and verification task that can deal with the partial image loss or incomplete fingerprint image. Global feature vector extraction is the main global approach that we contribute in this thesis. In this case, we implemented global geometrics based feature extraction for fingerprint identification and verification task. A set of global features (seven-moment values) were extracted from the partial fingerprint (incomplete fingerprint image). The study shows that global feature vector can more efficiently deal with incomplete fingerprint recognition problem when compared with the classical approach to the fingerprint identification and verification problem which is based on extracting minutia features from the fingerprint rides as well as the pores in different feature extraction levels. The studied system has been tested using a database that was randomly generated out of some random incomplete fingerprint images. Randomly generated incomplete fingerprint images were sorted into 10 groups according to the size of the missing part in each image. Then we randomly selected random images from each group to compose a new challenge dataset to be tested in two different approaches which are global, and Local feature extraction approaches. The experimental results show that global approach has about 87% while the local approach has 17% of identification and verification effectiveness. This means global approach improves the performance of the fingerprint identification and verification system on partial (incomplete) fingerprint images by 70% more than the classical approach.