Hyperspectral imaging (HSI) sensors are able to capture both spatial and spectral information of a given scene, providing images in which each pixel is represented by a vector array, named spectral response, with hundreds of values coming from different wavelengths across the electromagnetic spectrum. HSI normally covers the visible and shortwave infrared regions (400-2500nm), where each one of the values in the spectral response represents the reflected light intensity for a particular wavelength within that range.

Combining this spatial and spectral information has great potential, where each material in nature has a potentially unique spectral response, leading to excellent capabilities for pixel-wise classification. However, in remote sensing, it is often the case that more than one material is captured in the same pixel of the image, leading to a mixed spectral response at different proportions depending on the quantity (abundance) of each material (endmember) in the pixel.

This PhD will develop novel artificial intelligence-based techniques for spectral unmixing in the context of Earth/Space observations for different scenarios, including the potential use of data from several sensors (e.g., HSI & LiDAR data fusion) to boost the unmixing capabilities. The methodology planned for developing the proposed research aims to include already existing datasets publicly available for performance benchmarking, and also data captured at lab-based conditions for further research in more challenging scenarios.

Further information

An exciting opportunity to develop fundamental research on hyperspectral imaging and multimodal data fusion for Space applications.