Abstract
Hyperspectral imaging is a powerful technique that captures detailed spectral information across several narrow spectral bands, which enables a diverse range of applicationsin environmental monitoring, precision agriculture, and industrial inspection. However, spectral variability, caused by several environmental factors, poses a significant challenge when performing hyperspectral image analysis. This research project proposed to develop innovative and robust hyperspectral image processing methods that address the adverse effects of hyperspectral variability. The focus will first be on developing moderately complex models, which leads to only mildly non-convex optimization problems, as opposed to the highly non-convex problems encountered today. Second, techniques from Sparse Polynomial Optimization and the theory of moments will be combined to reformulate non-convex optimization problems as convex ones. Lastly, deep learning approaches will be explored. The proposed methods will be applied to several key hyperspectral image processing tasks, including spectral unmixing, denoising, and clustering. The project's outcomes are expected to advance the state-of-the-art in hyperspectral image processing and contribute to the development of more accurate and reliable hyperspectral imaging applications.
Researcher(s)
Research team(s)
Project type(s)