Multi-spectral remotely-sensed data such as satellite imagery can yield excellent insights into complex phenomena such as weather systems. Analysing the multi-channel space to separate out different features still presents a challenge, which may increase with the availability of hyper-spectral satellites. We use component labelling and population thresholding techniques to separate out clusters in hyper-dimensional channel space and use this information to identify different cloud types in geostationary satellite imagery. Three dimensional visualisation techniques are used to study the hyper-dimensional channel population data.
Keywords: remote sensing; multi-spectral imaging; visualization; component labelling.
Full Document Text: PDF version.