Urban areas have complex components, such as buildings,
roads, green areas, water bodies and bare lands. Although
automatic methods have been developed to map urban
land use for individual cities for a long time [18,40],
there are four difficulties in mapping urban expansions in
large areas. First, the inconsistency of acquisition time in
the remotely sensed data, which makes it difficult to apply a
uniform automatic method to extract urban areas. Second,
the uncertainty of spectral characteristics in urban areas due
to the use of various kinds of building materials, which
makes it hard to describe urban areas in a way that can be
easily understood by image analysts. This always makes it
difficult when mapping is done by multiple interpreters.