Page 327 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

segmented with a constant value of scale parameter and variable color values of 0.9-0.5-0.1. (Fig 2) This segmentation
image shows that the lower color, leads to losing details. As a result, the color parameter, and basically spectral factor

plays an important role in making an image object. Finally, by visually comparing the different levels of segmentation, in
the size of the image object and preserving the details, the best-segmented level is selected for applying classification.

5-2 Image classification based on Object-oriented

After selecting the best segmentation level, in order to describe the classes, the sample objects were declared and spectral
separability done. The best feature selection combination has applied for classification. The results of both methods of
classification on the IRS and IKONOS images are given. (Fig.3, 4)

Object oriented classification
Pixel base classification

Fig.3. classification maps of two methods in IRS_LISS3

6