Hyperspectral imaging, processing, and new challenges: a comprehensive review

Abstract

Hyperspectral imaging (HSI) captures rich spectral information across hundreds of contiguous narrow wavelength bands, enabling fine-grained material characterization beyond conventional imaging modalities. With the rapid development of sensor technologies and data-driven algorithms, HSI has been widely adopted in diverse scientific and engineering fields while also facing fundamental challenges arising from high-dimensional nonlinear spectral mixing, severe scarcity of labeled data, cross-sensor and cross-domain variability, and growing demands for real-time processing and scalable multimodal integration. We present a comprehensive review of hyperspectral imaging from the perspectives of imaging principles, system calibration and preprocessing, and image processing methodologies. In particular, we systematically review core hyperspectral image processing tasks, including spectral mixture analysis and unmixing, dimensionality reduction, classification, and segmentation, as well as detection-related tasks, and summarize representative methods and their technical evolution across different analytical objectives. Finally, open challenges and future research directions are discussed, with a focus on data annotation scarcity, multimodal integration, real-time processing, and the emerging impact of large models and generative artificial intelligence on hyperspectral image analysis. We aim to provide a structured reference and a holistic understanding of current progress and future directions in hyperspectral imaging and processing research.