An optimised latent fingerprint matching system using cuckoo search

Abstract

Biometrics is a scientific automation based on behavioural and physiological human traits to identify a person uniquely. Fingerprints are one of the eminent human psychological cues with an admiral potential for person identification. It has been widely adopted by many law enforcement agencies, forensics for criminal investigation. Amongst the three categories of fingerprints, latent fingerprints are the most intricate prints found in crime scenes and functions as dominant evidence in forensic investigations. The development of an accurate and reliable automatic latent fingerprint system is one of the pivotal issues for fingerprint researchers. This paper seeks the development of an optimal system for accurate latent fingerprint matching. The proposed system includes preprocessing of latent fingerprint image to remove noisy or unwanted information followed by minutiae extraction. The final step of proposed system includes the employment of swarm intelligence based cuckoo search algorithm to acquire optimal matching results. The performance of the system is evaluated on NIST Special Database 27 in terms of similarity score and identification rate. For the three categories of NIST dataset good, bad & ugly, the proposed system has achieved similarity score value of 94.50%, 86.22% & 84.78% and identification rates of 97.72%, 87.05% & 83.52% respectively. The proposed system outperformed in comparison with other considered concept of latent fingerprint matching techniques.