Computational geometry methods in fingerprint identification

The thesis introduces an automatic fingerprint identification system and proposes new topology-based algorithms for fingerprint matching. Fingerprint identification is a complex pattern recognition problem. Matching two fingerprint images is equivalent to matching the feature representations of them.; The major contribution of the thesis is in the area of the feature extraction and fingerprint pattern matching. The thesis improves the minutiae extraction by proposing new methods for fingerprint image preprocessing, enhancement and binarization. A novel multi-resolution approach to singular points detection is introduced in the thesis. This new method is able to detect singular points in a very precise manner.; The novelty of the thesis is in the topology-based algorithms for fingerprint matching. We concentrate on three major aspects of fingerprint matching: local matching, tolerance to deformation and global matching. Efforts are made to improve the accuracy and speed of fingerprint identification. Computational geometry methods are utilized for this purpose, which include Delaunay triangulation, spatial interpolation and distance transform technique. The proposed topology-based methods are able to efficiently deal with the distortions of fingerprint.; Experimental results and theoretical analysis confirm that the algorithms presented are effective and more efficient compared to other fingerprint matching algorithms.