Deep Learning Meets Hyperspectral Image Analysis: A Multidisciplinary Review.

Area of Science:

• Multidisciplinary science
• Computer vision
• Data science

Background:

• Hyperspectral imaging systems generate massive datasets, presenting significant analysis and interpretation challenges.
• Deep learning (DL) provides advanced opportunities for classical imaging tasks and novel spatial-spectral domain problems.
• Hyperspectral technology, originating in Remote Sensing, now impacts diverse and evolving application sectors.

Purpose of the Study:

• To provide domain professionals with an updated overview of hyperspectral imaging techniques combined with deep learning architectures.
• To offer machine learning and computer vision experts insights into deep learning applications for hyperspectral data from a multidisciplinary viewpoint.
• To highlight the potential and critical issues in the development trends of deep learning for hyperspectral data analysis.

Main Methods:

• Review of current literature on deep learning architectures applied to hyperspectral data.
• Analysis of hyperspectral imaging techniques and their integration with machine learning algorithms.
• Exploration of diverse application fields, including but not limited to Remote Sensing.

Main Results:

• Deep learning approaches are effective for solving complex imaging tasks in the spatial-spectral domain.
• The review synthesizes knowledge for both hyperspectral domain experts and machine learning specialists.
• Identified are key potentialities and critical issues in the application of deep learning to hyperspectral data.

Conclusions:

• Deep learning is a transformative tool for unlocking the information potential of hyperspectral imaging datasets.
• The review emphasizes the multidisciplinary nature of hyperspectral data analysis and its expanding applications.
• Further research is needed to address the challenges and fully realize the capabilities of deep learning in this field.