Image Processing Reference
also offer, for the first time, the possibility of analyzing the spatial configuration
and density of individual addresses within neighborhoods, which are typically hidden
by zonal representations. The geographic co-ordinates of each point are positioned
to correspond with the center of the building, allowing measures relating to the
distribution of points indicating density (compactness or sparseness) and arrange-
ment (linearity or randomness). The creation of ADDRESS-POINT TM and COMPAS TM
represents an important step forward in the pursuit of 'framework' data that encap-
sulate the desire for higher quality urban information not just for image pattern
recognition and classification improvement but also for all potential urban-based
spatial data analyses.
In addendum, it is important to remember that all
classification methodologies classify the remotely sensed
image not reality. Before over-focusing on algorithm sophis-
tication an appreciation and awareness of what exactly the
image represents in reality is of paramount importance.
At the same time, effective communication of how reality
is represented is critically dependent on the scope and detail
of thematic categorization. Finally, it is worth noting that
despite the many caveats, compared to conventional mapping, image classification is
still a relatively rapid, comprehensive, multidimensional and inexpensive alternative.
it is important to
image not reality
This chapter emphasized the main issues prevalent in urban classification; the
distinctions between land cover and land use, the difference between hard and
soft classification, and the incorporation of ancillary data in spectral and spatial
modifications. In all, real world examples were given to demonstrate not only
these contrasts but also the emerging importance of large-scale urban remote
sensing, particularly, the advent of very high spatial resolution satellite sensor
data and the pressing issues of global urbanization and urban sustainability.
In addition, differences between spectral based and spatial based classification
was introduced, both in relation to the role of ancillary data.
List of Software Packages Capable of Spectral Classification
ENVI (Research Systems Institute). Free evaluation available.