The main class groups are the land, river bed and flooded areas. Flooded areas are
grouped in two sub-classes. One for the regions that are under water on the day that flooding
occurred, and the other for the regions those remain under water in the image taken 10 days
after. The discrimination is achieved by operations through the hierarchical network, like
subtracting the “land under water” classes from the general land class and assigning them to
flood group, and subtracting the “flood” and “river bed” classes from the water class as subclasses.
The classification results in three levels for a zoomed area are given in figure 7.
Consequently, resultant classification is not as general as level 3, and not noisy as level 1
classification. The classification result for the whole image is shown in figure 8.