LPIDB v1.0 - Latent palmprint identification database

Abstract

This paper presents a new public available database for latent palmprint identification. Latent palmprint identification is an important research area which includes scientific challenges as well as social interest. Latent palmprints appear frequently in criminal investigations so developing accurate identification systems is critical in solving these investigations. Latent palmprint identification includes several pattern recognition challenges such as matching, feature extraction, and image processing. The lack of public latent palmprint databases has limited advances in scientific state-of-the-art researches. The database presented in this paper comprises 380 latent palmprints from 100 palms acquired under realistic conditions. The database includes the minutiae (position and orientation) taken manually and automatically. Additionally new research opportunities based on this database are presented as well as the benchmarks obtained with different publicly available minutiae extractors and matchers. As an example of the possibilities of the database a comparison between automatic and manual minutiae extraction is included.