Contactless palm print recognition: Novel design and palm openness classification

Abstract: Biometric technologies, such as facial and fingerprint recognition, have become widely adopted for identity verification in various applications. Palm biometrics, utilizing the unique patterns of the human palm, has gained significant attention for its accuracy and security. This thesis aims to investigate and propose a comprehensive design for certain as- pects of a contactless palm print recognition system, taking into account the issue of usability and palm openness. Contactless palm print recog- nition offers several advantages over other commonly adopted biometric systems. Firstly, it can operate effectively with low-resolution images and inexpensive cameras, making it more cost-efficient. However, the most significant advantage, particularly in the present circumstances, is its hygienic nature. But the technology also comes with new challenges: distance to the camera and the changes in palm print due to palm open- ness to name a few. Through the employment of transfer learning and landmark-based methods, the classification of palm openness from in- put images achieved an average accuracy of 0.90. In conclusion, despite their perceived visual quality, openness creates slight variations in width, depth, and crease distances on the palm print, which in turn affects the matching between images. Our results indicate that palm images are best matched to ones of the same openness. These findings validate the importance of palm openness and its impact on system performance, as well as the ability of the classifier to correctly classify the openness. By addressing these challenges, we can improve the accuracy and applicab- ility of contactless palm print recognition technology.