The feature space represents in various dimensions all the features that can be used for classification (e.g. image bands, band math parameters, derived texture properties). A point in that space is also called a vector with values for each feature (or dimension). Polyhedralization is a form of vector space quantization where a vector is assigned to the closest centre point of one polyhedron.
Explain the advantage of polyhedralization when adding new classes to an existing image classification system
New