Summary: | 碩士 === 國立中央大學 === 資訊工程學系在職專班 === 103 === Singularity points are key features in fingerprint images, representing the flow patterns of ridges and valleys. Therefore, the position and quantity of singularities are crucial in fingerprint recognition systems. In conventional singularity detection methods, the Poincaré index is prone to influence by local directional noises. This often results in excessive false reference points or the loss of true reference points. This study proposed an improved singularity detection method through applying a series of processing techniques comprising discrete wavelet transform image preprocessing for reducing image noise, Gabor filtering for strengthening the fingerprint patterns and enhancing fingerprint images, and Poincaré indexing for detecting possible singularities. In the detection process, local binary patterns surrounding every possible singularity were calculated to filter false reference points and reduce error judgments. The experimental results show that, in contrast to conventional methods, this method effectively reduced the number of false singularities caused by directional noise, thereby enhancing the detection rate of the fingerprint singularities.
|