Fingerprint subclassification using rotation-invariant features

A new method for fingerprint subclassification is proposed. First, a rotation-invariant feature is generated by rotation-invariant distance between orientation features. Then, fuzzy methods are adopted in both clustering step and classification step. With an introduced parameter, the balance between accuracy and average comparison times is made. The results show that the clustered subclasses are reasonable and the classification accuracy outperforms the combinations of crisp methods and fuzzy methods.