Non-parametric classification algorithm with an unknown class

In the classification of pixels of a multispectral image by methods of supervised classification, a problem can arise in case when an unknown class is present. In this paper, we suggest a method that gives good results in such a case. The method provides an estimation for a posteriori probability vectors (and consequently, classification), and, besides, estimates the prior probability of classes, including the unknown one, and thus, the areas occupied by every class