Paper Details
Abstract
Biometric identification systems, such as face, fingerprint, and palm print recognition, have become increasingly popular, with contactless modalities like face recognition gaining prominence due to the sensitivity of finger surfaces. However, face recognition is often hindered by high costs and performance degradation from factors like aging or physical exertion. Palm print recognition, on the other hand, offers greater stability over time and distinctiveness across individuals, making it a promising alternative. A key challenge in palm print recognition is achieving reliable performance in uncontrolled environments with dynamic backgrounds, varying lighting, and crowd interference, which complicate region of interest (ROI) extraction. In this paper, we propose a novel, real-time, contactless palm print identification system that is efficient, low-cost, and robust under such conditions. Our system achieves an AUC of 0.9286 and identity accuracy of 95.53\% on the Birjand University Mobile Palmprint Database (BMPD), and an AUC of 0.9981 with 99.93\% identity accuracy on a small, realistic dataset we collected. Source code and demo are available at: \url{https://github.com/VKev/Real-Time-Contactless-Palm-Print-Identification-System}.