Image Processing Reference
In-Depth Information
1993
; Heikkonen and Varfis
1998
). Remote sensing by definition involves sensors
that are not in contact with their objects and as such remotely sensed data are capable
of only measuring energy deflected off, or emitted from, physical substances from a
distance. Environmental applications, such as those focusing on the monitoring of
deforestation, changes in meteorological and hydrological cycles are closely amena-
ble to such types of measurements. However, urban areas, although definable by
tangible substances such as brick, concrete, metal, vegetation and so on, which are in
themselves capable of being detected from a distance by remote sensors, are also
collections of human habitation and social interaction, which
are not directly tangible and therefore impossible to mea-
sure remotely. Some purists would like to limit the range of
urban thematic categorization from remote sensing to just
land cover, and indeed as illustrated in Chapter 6, the Ridd
(
1995
) Vegetation-Impervious-Soil three-way model com-
fortably fits within the technical capabilities of remote sensor
instrumentation (Fisher
1997
). However, and as alluded
earlier, urban, unlike most environmental, remote sensing classifications underpin a
variety of pragmatic planning and policy-making applications which demand far
more meaningful labels (Donnay
1999
). They demand semantics that describe how
physical and socio-economic characteristics are not only used but how they also func-
tion. Fortunately, there are conceptual and mechanical rationales between urban land
cover and land use, where one is dependent on the other (Dobson
1993
). The process
of inferring anthropogenic land use from the physical configuration of land cover is
an extricable link between form and function (Geoghegan et al.
1998
). For example,
changes in the physical shape, size and density of urban built
morphologies are an indication of urban growth, which in
turn is a proxy for inferring processes such as a suburban-
ization of the population and a decentralization of commerce
(see relative discussions in Chapters 2 and 3). Similarly,
measures of ratios between built and biomass may be the
outcome of quality of life indicators that can be validated
by reference to census parameters (inter alia Brugioni
1983
;
Ogrosky
1975
; Donnay
1999
; Chen
2002
; Harvey
2002
).
Nevertheless, all inferences between land cover and land
use should be treated as tenuous and should be attempted
always with respect to application objectives and the scale of
investigation (Woodcock and Strahler
1987
).
land cover refers
to physical
properties, readily
measurable by
remote sensor
data
land use, which
is not directly
measurable by
remote sensor
data, is an
inference of
anthropogenic
activities from
land cover
composition and
configuration
Hard Versus Soft Classification: Is Refining Scale the Solution?
Whether spectral or spatial, the inherent geometrical variability of the urban
landscape dictates that the scale of measurement is a major factor in both
the quality and detail of the resulting classification (Welch
1982
). Recent
