Scholarlink

<p>Palmprint recognition’s outstanding sanitary, non-invasive, and user-friendliness properties have sparked a lot of research attention. The majority of palmprint recognition techniques used today are deep learning approaches, which typically use palmprint images to learn discriminative features. In most cases, considerable labelled samples are needed to perform well enough for identification. Here, we propose employing a deep learning method for palmprint recognition to get around the problems. First, the original picture should be (a) converted to greyscale, (b) cropped and resized, and (c) contrast-enhanced using the contrast-limited adaptive histogram equalization method. Next, use the ConvNeXt approach to extract the most essential characteristics from contrast-enhanced palm pictures. After removing redundant collected attributes, the improved spotted hyena optimizer algorithm selects the most important features. Finally, the deep fuzzy neural network (DFNN) technique determines if the palmprint image matches. Use the sooty tern optimization algorithm for increased identification and categorization accuracy to increase categorization accuracy. The proposed approach effectively reduces the number of attributes and computation time while increasing the accuracy of the optimized DFNN—an approximation of the proposed methods using Tongji, IITD, and CASIA open palmprint datasets. The approach has better accuracy with 99.62%, 99.72%, and 99.67% on IITD, Tongji, and CASIA palmprint databases. Experiments show that our approach achieves a high identification rate while using a substantially fewer number of features.</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

An effective feature selection and classification technique for palmprint biometric identification systems

  • 期刊论文

  • 发表日期：

    2025年5月31日

查看全文
  • Aravind Nalamothu,
  • Eswaraiah Rayachoti

摘要

Palmprint recognition’s outstanding sanitary, non-invasive, and user-friendliness properties have sparked a lot of research attention. The majority of palmprint recognition techniques used today are deep learning approaches, which typically use palmprint images to learn discriminative features. In most cases, considerable labelled samples are needed to perform well enough for identification. Here, we propose employing a deep learning method for palmprint recognition to get around the problems. First, the original picture should be (a) converted to greyscale, (b) cropped and resized, and (c) contrast-enhanced using the contrast-limited adaptive histogram equalization method. Next, use the ConvNeXt approach to extract the most essential characteristics from contrast-enhanced palm pictures. After removing redundant collected attributes, the improved spotted hyena optimizer algorithm selects the most important features. Finally, the deep fuzzy neural network (DFNN) technique determines if the palmprint image matches. Use the sooty tern optimization algorithm for increased identification and categorization accuracy to increase categorization accuracy. The proposed approach effectively reduces the number of attributes and computation time while increasing the accuracy of the optimized DFNN—an approximation of the proposed methods using Tongji, IITD, and CASIA open palmprint datasets. The approach has better accuracy with 99.62%, 99.72%, and 99.67% on IITD, Tongji, and CASIA palmprint databases. Experiments show that our approach achieves a high identification rate while using a substantially fewer number of features.