Altered fingerprint matching by using patch based classification in deep learning

Automated Fingerprint Identification Systems are widely used in border control and law enforcement applications, thus their importance should not be overlooked. Several challenges have been handled in the fingerprint system’s security, including the use of fake fingerprints to conceal one’s identity, as well as fingerprint change and other security concerns. The fundamental contribution of the study is a method for matching altered fingerprints to rolled fingerprints that avoids concealing identity and alteration by utilizing a technique called fingerprint rolling. (a) Representations of patches are learned for similarity and (b) minutiae on the linked patches. The optimal representations of picture patches are investigated using a deep learning network. A distance measure is used to find commonalities between patches from the changed fingerprint and reference fingerprints.