Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Comprehensive Survey for Hyperspectral Image Classification: The Evolution from Conventional to Transformers and Mamba Models

Muhammad Ahmad, Salvatore Distifano, Adil Mehmood Khan, Manuel Mazzara, Chenyu Li, Hao Li, Jagannath Aryal, Yao Ding, Gemine Vivone, Danfeng Hong

TL;DR

This survey addresses the challenges of Hyperspectral Image Classification (HSC) by tracing the field from conventional machine learning to Deep Learning (DL), Transformers, and the emerging Mamba state-space models. It systematically reviews data representations, learning paradigms, conventional methods, DL architectures (CNNs, DBNs, RNNs, AEs), and Transformer- and Mamba-based approaches, highlighting key trends such as spectral–spatial fusion, domain adaptation, diffusion denoising, and explainable AI. Experimental results on three benchmark HS datasets corroborate the performance of Transformer-based models and graph/CNN hybrids, while identifying limitations including data scarcity, computational demands, and interpretability. The paper also outlines open challenges, proposes future research directions, and discusses practical implications for real-world applications in precision agriculture, environmental monitoring, and resource management. Overall, it provides a comprehensive baseline and roadmap for advancing accurate, efficient, and interoperable HSC systems.

Abstract

Hyperspectral Image Classification (HSC) presents significant challenges owing to the high dimensionality and intricate nature of Hyperspectral (HS) data. While traditional Machine Learning (TML) approaches have demonstrated effectiveness, they often encounter substantial obstacles in real-world applications, including the variability of optimal feature sets, subjectivity in human-driven design, inherent biases, and methodological limitations. Specifically, TML suffers from the curse of dimensionality, difficulties in feature selection and extraction, insufficient consideration of spatial information, limited robustness against noise, scalability issues, and inadequate adaptability to complex data distributions. In recent years, Deep Learning (DL) techniques have emerged as robust solutions to address these challenges. This survey offers a comprehensive overview of current trends and future prospects in HSC, emphasizing advancements from DL models to the increasing adoption of Transformer and Mamba Model architectures. We systematically review key concepts, methodologies, and state-of-the-art approaches in DL for HSC. Furthermore, we investigate the potential of Transformer-based models and the Mamba Model in HSC, detailing their advantages and challenges. Emerging trends in HSC are explored, including in-depth discussions on Explainable AI and Interoperability concepts, alongside Diffusion Models for image denoising, feature extraction, and image fusion. Comprehensive experimental results were conducted on three HS datasets to substantiate the efficacy of various conventional DL models and Transformers. Additionally, we identify several open challenges and pertinent research questions in the field of HSC. Finally, we outline future research directions and potential applications aimed at enhancing the accuracy and efficiency of HSC.

A Comprehensive Survey for Hyperspectral Image Classification: The Evolution from Conventional to Transformers and Mamba Models

TL;DR

This survey addresses the challenges of Hyperspectral Image Classification (HSC) by tracing the field from conventional machine learning to Deep Learning (DL), Transformers, and the emerging Mamba state-space models. It systematically reviews data representations, learning paradigms, conventional methods, DL architectures (CNNs, DBNs, RNNs, AEs), and Transformer- and Mamba-based approaches, highlighting key trends such as spectral–spatial fusion, domain adaptation, diffusion denoising, and explainable AI. Experimental results on three benchmark HS datasets corroborate the performance of Transformer-based models and graph/CNN hybrids, while identifying limitations including data scarcity, computational demands, and interpretability. The paper also outlines open challenges, proposes future research directions, and discusses practical implications for real-world applications in precision agriculture, environmental monitoring, and resource management. Overall, it provides a comprehensive baseline and roadmap for advancing accurate, efficient, and interoperable HSC systems.

Abstract

Hyperspectral Image Classification (HSC) presents significant challenges owing to the high dimensionality and intricate nature of Hyperspectral (HS) data. While traditional Machine Learning (TML) approaches have demonstrated effectiveness, they often encounter substantial obstacles in real-world applications, including the variability of optimal feature sets, subjectivity in human-driven design, inherent biases, and methodological limitations. Specifically, TML suffers from the curse of dimensionality, difficulties in feature selection and extraction, insufficient consideration of spatial information, limited robustness against noise, scalability issues, and inadequate adaptability to complex data distributions. In recent years, Deep Learning (DL) techniques have emerged as robust solutions to address these challenges. This survey offers a comprehensive overview of current trends and future prospects in HSC, emphasizing advancements from DL models to the increasing adoption of Transformer and Mamba Model architectures. We systematically review key concepts, methodologies, and state-of-the-art approaches in DL for HSC. Furthermore, we investigate the potential of Transformer-based models and the Mamba Model in HSC, detailing their advantages and challenges. Emerging trends in HSC are explored, including in-depth discussions on Explainable AI and Interoperability concepts, alongside Diffusion Models for image denoising, feature extraction, and image fusion. Comprehensive experimental results were conducted on three HS datasets to substantiate the efficacy of various conventional DL models and Transformers. Additionally, we identify several open challenges and pertinent research questions in the field of HSC. Finally, we outline future research directions and potential applications aimed at enhancing the accuracy and efficiency of HSC.
Paper Structure (45 sections, 3 equations, 18 figures, 4 tables)

This paper contains 45 sections, 3 equations, 18 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Limitations of Existing Surveys
  3. Significance of Accurate Classification
  4. Brief Introduction to Deep Learning
  5. Structure of the Survey
  6. Hyperspectral Data Representation
  7. Learning Categories
  8. Supervised Learning
  9. Unsupervised Learning
  10. Semi-Supervised Learning
  11. Reinforcement Learning
  12. Transfer Learning
  13. Online Learning
  14. Self-learning
  15. Brief Overview of Conventional Methods
  16. ...and 30 more sections

Figures (18)

  • Figure 1: Various Real-World Applications of Hyperspectral Imaging (HSI):Agriculture: HSI is utilized to monitor crop health through stress detection and nutrient assessment. It can identify diseases at an early stage by analyzing spectral changes in plant reflectance. Additionally, HSI helps assess soil properties, including moisture content and organic matter levels, which are crucial for optimizing crop yields. Environmental Monitoring: In this domain, HSI plays a vital role in tracking pollution levels by analyzing spectral signatures of pollutants in air and water. It facilitates ecosystem mapping by providing detailed information about vegetation types and health. HSI also effectively assesses water quality by detecting harmful algal blooms and monitoring sediment levels. Mineralogy: HSI is employed in mineral identification by differentiating between mineral types based on spectral characteristics. This capability is invaluable in mining exploration, as it aids in locating valuable mineral deposits and optimizing extraction processes. Healthcare: In the medical field, HSI assists in disease diagnosis by identifying abnormal tissue characteristics. It is also utilized in monitoring tissue health, particularly in assessing the efficacy of treatments through non-invasive spectral analysis. Surveillance and Security: HSI enhances surveillance capabilities by enabling target detection through the analysis of specific material signatures. It is also used in material identification, allowing for better security measures by distinguishing between benign and hazardous materials.
  • Figure 2: University of Houston: Hyperspectral Image Cube representation.
  • Figure 3: Major Learning Categories.
  • Figure 4: Example of Deep Artificial Neural Network.
  • Figure 5: Basic example of Spectral Convolutional Neural Network.
  • ...and 13 more figures