Hyperspectral Imaging and Artificial Intelligence: Methods, Challenges, and Applications

Hyperspectral imaging provides rich spectral information by capturing data across hundreds of narrow and contiguous wavelength bands. This technology is widely used in fields such as remote sensing, agriculture, environmental monitoring, mineral exploration, and medical imaging. However, hyperspectral data analysis remains challenging due to several inherent limitations. The main difficulties include high dimensionality, noise, spectral mixing, limited annotated data, and variability in spectral signatures caused by environmental, sensor, or material-related factors. These challenges can reduce classification accuracy, complicate feature extraction, and limit the generalization of traditional analytical methods. Artificial intelligence, particularly machine learning and deep learning, offers powerful tools to address these issues by automatically learning relevant spectral-spatial features, reducing dimensionality, improving noise robustness, handling complex spectral mixtures, and enhancing performance even when labeled data are scarce.

This Research Topic aims to address the major challenges that limit the effective analysis and interpretation of hyperspectral data. Hyperspectral images contain hundreds of highly correlated spectral bands, which often leads to high computational complexity, redundancy, and overfitting. Artificial intelligence can help overcome these limitations through dimensionality reduction, band selection, and automatic feature extraction, thereby improving model performance and robustness. Another key issue is the limited availability of annotated samples. Recent advances such as few-shot learning, transfer learning, semi-supervised learning, and deep learning with small training sets make it possible to exploit available data more efficiently and improve generalization. AI also contributes to noise reduction, spectral variability management, and mixed-pixel analysis through spectral-spatial models, autoencoders, CNNs, Transformers, and unmixing approaches. A further objective is to encourage the development of models specifically adapted to hyperspectral data, capable of jointly exploiting spectral, spatial, and physical information rather than relying on generic architectures. These advances may support applications in remote sensing, agriculture, environmental monitoring, and medical diagnosis, while remaining complementary to essential preprocessing steps such as radiometric calibration and atmospheric correction.

This Research Topic welcomes contributions on the role of artificial intelligence in advancing hyperspectral data analysis. We particularly invite manuscripts addressing the following themes: High-dimensional data analysis, including dimensionality reduction, band selection, and feature extraction. Learning with limited annotations, including few-shot learning, transfer learning, semi-supervised learning, self-supervised learning, and domain adaptation. Noise, spectral variability, and mixed pixels, including denoising, spectral unmixing, robust classification, and sub-pixel analysis. AI models adapted to hyperspectral data, including spectral-spatial deep learning, CNNs, Transformers, graph neural networks, explainable AI, and physics-informed approaches. Real-world applications, including remote sensing, precision agriculture, environmental monitoring, mineral exploration, food quality assessment, and biomedical imaging. We welcome original research articles, reviews, mini-reviews, methodological papers, perspectives, and case studies presenting theoretical advances, new algorithms, benchmark evaluations, or practical applications.