A novel indexing algorithm for latent palmprints leveraging minutiae and orientation field

Abstract

Latent palmprints represent crucial forensic evidence in criminal investigations, necessitating their storage in governmental databases. The identification of corresponding palmprints within large-scale databases using an automated palmprint identification system (APIS) is time-consuming and computationally intensive. To address this challenge, this paper introduces an innovative approach: delineating the region of interest (ROI) for palmprint segmentation and presenting a novel indexing algorithm founded on minutiae and the orientation field (OF). Additionally, a novel feature vector is proposed, leveraging minutiae triplets and ellipse properties, marking the pioneering algorithm to consider minutiae importance in palmprint indexing. Significantly, an improved version of an existing palmprint indexing algorithm tailored for latent palmprints is introduced. The study demonstrates the indexing and retrieval of both our feature vectors and those obtained by the improved palmprint indexing algorithm, using two clustering algorithms and locality-sensitive hashing (LSH). The method's robustness is evaluated across three diverse databases with extensive palmprint records. The experimental results underscore the superior performance of our approach compared to current state-of-the-art algorithms.