Qualitative and spatial comparative study of satellite images classified by supervised and fuzzy logic based classification algorithms: a case study in kilkis prefecture, central macedonia, greece

摘要

In a first stage a SPOT image of the survey area was classified for land use/land cover classification, using the Maximum Likelihood Algorithm (MLC). Howeverk, due to the spatial uncertainty which exist mainly between the borders of the spectral categories, as they defined by MLC, in a second stage a supervised classification based on fuzzy classifiers was applied. A sigmoid function defines the degree that every pixel belongs in each category and differentiates the results of the classification in comparison with those of the classical Boolean logic. The results of the fuzzy classification leads to the construction of another land use/land cover map. For reasons of comparison between the two methods, the results of eadch classified category in both methods was converted to an integer binary image. As qualitative index of agreement between the two methods, teh Kappa index of agreement and for each category was used. The results are evaluated with field work.