Hyperspectral (HS) imaging has been used extensively in remote sensing applications\nlike agriculture, forestry, geology and marine science. HS pixel classification is an important task to\nhelp identify different classes of materials within a scene, such as different types of crops on a farm.\nHowever, this task is significantly hindered by the fact that HS pixels typically form high-dimensional\nclusters of arbitrary sizes and shapes in the feature space spanned by all spectral channels. This is\neven more of a challenge when ground truth data is difficult to obtain and when there is no reliable\nprior information about these clusters (e.g., number, typical shape, intrinsic dimensionality). In this\nletter, we present a new graph-based clustering approach for hyperspectral data mining that does not\nrequire ground truth data nor parameter tuning. It is based on the minimax distance, a measure of\nsimilarity between vertices on a graph. Using the silhouette index, we demonstrate that the minimax\ndistance is more suitable to identify clusters in raw hyperspectral data than two other graph-based\nsimilarity measures: mutual proximity and shared nearest neighbours. We then introduce the\nminimax bridgeness-based clustering approach, and we demonstrate that it can discover clusters of\ninterest in hyperspectral data better than comparable approaches.
Loading....