Unsupervised classification of LANDSAT imagery can be accomplished very efficiently by using a four-dimensional histogram in table form. The clustering scheme isolates the peaks in the distribution which are used as cluster centers to classify the remaining vectors in the histogram. The general outlines of this scheme, detailed description and programming implementation are given together with flow diagrams. The program has been implemented on both DEC PDP-10 and DEC PDP-11/40 computers. A discussion of the extension of the algorithm to other than LANDSAT imagery is also given.
Date of this Version