Fingerprint pattern classification

摘要

This paper presents a new method for fingerprint classification. In this method, fingerprint images are divided into 32 × 32 subregions to obtain direction pattern. Next, the relaxation smoothing process with singularity detection and convergency checking is performed. Starting from the singular regions found, feature parameters of the fingerprint are obtained by extracting major flow-line∗ traces. The result of the experiments shows that this approach is capable of classifying fingerprint patterns into more than ten categories.