Unsupervised Multispectral Classification

Unsupervised classification techniques share a common intent to uncover the major land cover classes that exist in the image without prior knowledge of what they might be. Generically, such procedures fall into the realm of cluster analysis since they search for clusters of pixels with similar reflectance characteristics in a multi–band image. All of them are also generalizations of land cover occurrence since they are concerned with uncovering the major land cover classes and thus tend to ignore those that have very low frequencies of occurrence. However, given these broad commonalities, there is little else that they share in common. There are almost as many approaches to clustering as there are Image Processing systems on the market. Idrisi Selva is no exception. The primary unsupervised procedure it offers is unique (CLUSTER). However, it also offers a variant on one of the most common procedures to be found (lSOCLUST). As implemented here, this procedure is really an iterative combination of unsupervised and supervised procedures, as is also the case with the third procedure offered, MAXSFT.