Adaptive latent fingerprint indexing

Based on biometric system, identification of individual is complex and very interesting task. In law and forensic applications, fingerprint plays an important role in person recognition. Earlier focus was on full fingerprint but today latent fingerprint matching which is poor in quality, having non-linear distortion and small overlapping area is still a challenge for researchers. Latent prints contains noise in it, therefore to deal with latent matching having larger database, indexing techniques are required. Using indexing technique, matching process is speeded up because it places most possible match at the top from large database therefore reduces the time required for matching. In this paper, we have proposed an efficient latent indexing technique that reduces manual intervention and maintain latent matching accuracy. In this technique we have used manually extracted features as well as automatically extracted features and applied three approaches i.e. MCC, triplets, hybrid orientation field on both manually marked and derived minutiae and orientation field for indexing. We have carried out our implementation on NIST SD27 Database, to determine the performance of proposed approach by using large background database of approximately 200K of fingerprints. The results of this proposed algorithm has 96.7% hit rate at penetration rate of 39%, which is better than Paulino et al. and the latent matching accuracy was also maintained.