Incremental self-organised map represents the distribution of the input feature vectors easily by modifying the topology of the network. A simple learning algorithm is utilized for the analysis of
the remote-sensing images. Since this algorithm does not increase the computational load, the user
can analyse different segmentation results of the same remote-sensing image by changing the
threshold value in a descending manner. This analysis enables the user to observe the relation between the segmentation result and the number of classes and to observe the distribution of the novel classes formed on the segmented image.