An efficient palmprint recognition method based on block dominat orientation code

Abstract

A practical palmprint recognition system should have high recognition accuracy, fast matching speed, and high-security capabilities. In the past decades, most previous palmprint recognition systems that were designed on the basis of natural light, have achieved more preferable results in terms of recognition accuracy, matching speed, and security. However, they have certain limitations. It is necessary to improve the security capability of the system and assure recognition accuracy. The research proposes a novel hierarchical multispectral palmprint recognition method based on block dominant orientation code (BDOC). As the multispectral palmprint image is captured under red, green, blue, and near-infrared (NIR) illumination sources, more discriminative information which possibly contributes to improved recognition accuracy and security can be acquired. The experimental results showed that the recognition accuracies were all high in the each spectral band. By analysing the recognition results, the different bands are fused using a feature level fusion scheme to improve recognition accuracy. Meanwhile, as the fusion scheme was based on block selection from different bands, the system was secure because the combination of blocks differed between bands, making stronger discrimination possible. Therefore, the proposed method was provided with higher recognition accuracy, and a high-security capability. It offered a novel means of developing on-line multispectral palmprint systems.