spectral responses from bright soil are similar to bright impervious
surfaces and some building rooftops. The user's accuracy of the
swimming pool category was also exceptionally low (50.77%). There
were many different sample points identified as swimming pools
actually belonging to all other classes. We did not anticipate this
signature confusion, since spectral responses from some categories do
not seem similar to swimming pools. Another low user's accuracy was
produced by the grass category. There were many sample points that
were identified as grass which were trees and shrubs. From the
preceding discussion, it can be noted that the traditional per-pixel
approach is not effective in urban mapping, since the approach is
unable to handle spectrally similar classes. The accuracy of the
classical approach is low because there are many spectrally similar
land-cover classes within an urban environment.