Scholarlink

<p>In the post-pandemic era, the demand for contactless and secure identity authentication has surged. Traditional biometric features such as facial and palmprint recognition, though widely adopted, are susceptible to forgery via high-resolution photos and exhibit varying accuracy over time. To address these limitations, we propose an integrated biometric authentication system leveraging both palm print and palm vein images. Our approach employs RGB lenses for palm print capture and NIR lenses for unforgeable palm vein extraction. The palm print images are preprocessed using Real-ESRGAN to sharpen features, while the palm vein images undergo Gamma correction for enhanced contrast. An optimal fusion ratio of 20:80 is identified to merge these features, and an improved ROI capture method ensures image consistency and batch processing efficiency. Experimental evaluations using the YOLOv10 to YOLOv12 models reveal that YOLOv12 demonstrates the best overall performance, achieving a high mAP@50 of 0.964. Compared to traditional CNN and VGG16 models, YOLOv12 demonstrates relatively higher performance in terms of accuracy, stability, and anti-counterfeiting capabilities. This study suggests the potential contribution of feature fusion and rigorous data preprocessing in advancing biometric authentication systems. The project of this work is made publicly available at <a href="https://github.com/Mariiiiiio/FusionPalmID/tree/master">https://github.com/Mariiiiiio/FusionPalmID/tree/master</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced biometric authentication through integrated palm print and palm vein images

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Chi Hung Wang,
  • Wei Ren Chen,
  • Jun Jie Yen,
  • Xiang Shun Yang,
  • Yu Siang Siang

摘要

In the post-pandemic era, the demand for contactless and secure identity authentication has surged. Traditional biometric features such as facial and palmprint recognition, though widely adopted, are susceptible to forgery via high-resolution photos and exhibit varying accuracy over time. To address these limitations, we propose an integrated biometric authentication system leveraging both palm print and palm vein images. Our approach employs RGB lenses for palm print capture and NIR lenses for unforgeable palm vein extraction. The palm print images are preprocessed using Real-ESRGAN to sharpen features, while the palm vein images undergo Gamma correction for enhanced contrast. An optimal fusion ratio of 20:80 is identified to merge these features, and an improved ROI capture method ensures image consistency and batch processing efficiency. Experimental evaluations using the YOLOv10 to YOLOv12 models reveal that YOLOv12 demonstrates the best overall performance, achieving a high mAP@50 of 0.964. Compared to traditional CNN and VGG16 models, YOLOv12 demonstrates relatively higher performance in terms of accuracy, stability, and anti-counterfeiting capabilities. This study suggests the potential contribution of feature fusion and rigorous data preprocessing in advancing biometric authentication systems. The project of this work is made publicly available at https://github.com/Mariiiiiio/FusionPalmID/tree/master.