4.1 Segmentation
As a first step, image objects are extracted by segmentation in different scales and different
composition of homogeneity criteria, which represents the image information in different
spatial resolutions. The multi-scale analysis of data was conducted at three levels. The
segmentation parameters are given in Table 3 and the segmentation output of the same region
is shown in figure 5.
The level-3 segmentation was applied to create main objects for a general
discrimination as land and water body areas. However, in level-1, very small image objects
are formed to conduct more detailed discrimination of water body features. As an
intermediate segmentation, in level 2 which is main output level, the classification process
was interfered by using both sub and super object properties from all levels to get a better
classification result.