PalmMatchDB: An On-Device Contactless Palmprint Recognition Corpus

A palmprint recognition system identifies a person by using palmprint qualities that may or may not be apparent to the naked eye. It has proven to be appropriate for a variety of purposes and applications such as access control, law enforcement and forensic analysis. Although there are several existing palmprint recognition systems in the literature, they are mostly developed based on localized database, thus querying optimality in other locations. Due to the sparsity of publicly accessible palmprint image databases for black people, we built a non-contact palmprint image database containing 12,000 grayscale images captured from 200 black subjects using three different mobile phone cameras with 5-, 8-, and 12-megapixel resolutions respectively. The palmprints images were preprocessed using mean filtering technique, segmentation and alignment, and this study presents the preliminary results with respect to the dataset for the proposed palmprint recognition system.