Scholarlink

<p>Hyperspectral image classification significantly contributes to remote sensing, providing extensive information for diverse applications such as environmental monitoring, agriculture, and urban planning. The spectral richness within hyperspectral data enables detailed characterization of numerous target objects, ranging from distinguishing between various features on Earth’s surface in remote sensing to discriminating between healthy and unhealthy tissues in medical diagnosis. The ongoing advancements in hyperspectral image classification have introduced numerous new techniques not included in prior reviews. Given the substantial volume of articles in this domain, a comprehensive understanding of current developments can only be achieved by collectively incorporating newly proposed methods. Moreover, most existing reviews concentrate on deep learning methods, which dominate research trends. However, the evolving landscape of hyperspectral image classification demands a comprehensive exploration of both traditional methodologies and contemporary machine learning trends to realize the full potential of this technology. This paper examines various approaches to classifying hyperspectral images (HSI), covering traditional and machine learning-based methods. The initial discussion introduces standard tools such as Spectral Angle Mapper, Minimum Distance, Maximum Likelihood, and Spectral Feature Fitting. Subsequently, we analyze the challenges these conventional tools find difficult to address effectively and explore the efficacy of machine learning techniques in overcoming these issues. A framework is built to systematically review the recent advancements in machine learning-based HSI classification, categorizing the corresponding works into supervised and unsupervised learning techniques. Furthermore, a comprehensive analysis of popular and promising state-of-the-art methods in both categories is conducted, discussed, and summarized to provide in-depth insights. Following the discussed content, the paper explores emerging trends and future directions within this domain.</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

Employing machine learning for hyperspectral image classification: traditional approaches and modern trends

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Amit Kumar Singh,
  • Sneha Mishra,
  • Krovvidi Sai Pranav,
  • Abhishek Joshi,
  • Chandra Prakash Singh,
  • Harjas Partap Singh Romana

摘要

Hyperspectral image classification significantly contributes to remote sensing, providing extensive information for diverse applications such as environmental monitoring, agriculture, and urban planning. The spectral richness within hyperspectral data enables detailed characterization of numerous target objects, ranging from distinguishing between various features on Earth’s surface in remote sensing to discriminating between healthy and unhealthy tissues in medical diagnosis. The ongoing advancements in hyperspectral image classification have introduced numerous new techniques not included in prior reviews. Given the substantial volume of articles in this domain, a comprehensive understanding of current developments can only be achieved by collectively incorporating newly proposed methods. Moreover, most existing reviews concentrate on deep learning methods, which dominate research trends. However, the evolving landscape of hyperspectral image classification demands a comprehensive exploration of both traditional methodologies and contemporary machine learning trends to realize the full potential of this technology. This paper examines various approaches to classifying hyperspectral images (HSI), covering traditional and machine learning-based methods. The initial discussion introduces standard tools such as Spectral Angle Mapper, Minimum Distance, Maximum Likelihood, and Spectral Feature Fitting. Subsequently, we analyze the challenges these conventional tools find difficult to address effectively and explore the efficacy of machine learning techniques in overcoming these issues. A framework is built to systematically review the recent advancements in machine learning-based HSI classification, categorizing the corresponding works into supervised and unsupervised learning techniques. Furthermore, a comprehensive analysis of popular and promising state-of-the-art methods in both categories is conducted, discussed, and summarized to provide in-depth insights. Following the discussed content, the paper explores emerging trends and future directions within this domain.