Performance evaluation of an automatic fingerprint classification algorithm adapted to a Vucetich based classification system

Abstract

We study and evaluate an automatic fingerprint classification algorithm that we apply over the fully manual identification system being used by the Dirección Nacional de Identificación Civil (DNIC). To be compatible with the existing system and provide a gradual transition into a fully automatic procedure we mimic the classification scheme being used by DNIC technicians, which is based on a four-class Vucetich system. The classification algorithm we use is based on the method by Karu and Jain [4]. Some modifications to the original algorithm are proposed and evaluated over images extracted from a 4 million fingerprint card archive maintained by DNIC. The algorithm was also tested on fingerprints from the same individuals taken at two different points in time (separated several years) to further evaluate its performance and consistency.