city of Sydney, Australia. The chosen area had a mixture
of residential, commercial, and industrial land uses. The
study involved obtaining pixel values over a number of
residential, commercial, and industrial areas of varying
street orientation with respect to the radar azimuth angle.
The average pixel values and standard deviations were
then calculated for the subject areas. Although the
sample size was small, there was a definite trend
showing a correlation between radar backscatter and
building size.
The results show that the average pixel value and
standard deviation for the residential classes were lower
than those of the commercial classes while the industrial
areas had the highest values (for their relative orientation
angles). It is expected that residential land use would
have the lowest average pixel value and standard
deviation since residential buildings are generally smaller
in size (therefore less area to backscatter the radar
wave). The residential building materials were mainly
brick or fibrous cement walls with tiled roofs. These are
primarily dry with a low dielectric constant. A low dielectric
constant means a large amount of the radiation will
penetrate the surface of the building reducing the
backscatter measured at the radar receiver.
Areas of commercial buildings are usually more dense
and variable than residential. A local shopping area often
has metal clad roofs and terrace type buildings. A central
business district (CBD) contains many large buildings,
both in floor area and height, with a high density. There
are more metal structures (having a high dielectric
constant and therefore a strong reflection) acting as
support in large buildings. Hence areas of commercial
buildings show higher average pixel values and standard
deviation, than residential classes.
Industrial buildings gave the largest average pixel value
and standard deviation. Industrial regions consist of large
buildings mostly clad in metallic materials. Metals are
conductors with a high dielectric constant. They can give
a very strong backscatter at particular orientation angles.
The residential class, with the largest sample size of the
three, was examined further to show a relationship
between the backscatter response (being directly related
to the pixel value, or digital number, on a radar image)
and building orientation with respect to radar direction
(Figure 1). The standard deviation was also determined,
When the angle between the normal to the street or
building and the radar look direction (phi) is equal to zero,
the backscatter response is at its highest. As phi
increases to around 220. the backscatter decreases to a
minimum. For values of phi above 20 degrees the
backscatter varies little with phi. This result is similar to
that obtained by Hardaway et al (1982).
A MODEL FOR PREDICTING URBAN
CHARACTERISTICS
A model has been developed to give the expected
backscatter from a group of buildings of user defined size
(single or multiple storey), shape, material (including
surrounding ground surface), and radar parameters.
Existing formula, giving the backscatter for a corner
reflector and a rectangular facet, have been adapted into
the model. It is presently designed for a simple building of
rectangular shape with either a flat or sloping roof. When
the model is used to find the expected backscatter from a
block of buildings, of either small residential size or large
commercial size, the roof facing the radar, and the corner
reflector effect from the front wall, are the dominant
contributors to the backscatter (Figures 2 & 3). Figures 2
and 3 show an example of the simulated backscatter with
respect to phi (for residential and industrial areas
respectively) for each component of a building, as well as
the total backscatter. The oscillations of the backscatter
for both buildings are due to difference in phase as the
distance the wave travels between the extremes of the
object leads to either constructive or destructive
interference. These oscillations become more frequent as
building size increases.
As expected, Figures 2 and 3 show the backscatter is
greatest when the radar approaches the building normal
to the wall (when phi equals zero). The backscatter drops
as phi increases to 459 (with predominantly diffuse
backscatter). Around this angle the backscatter response
from trees (essentially volume scattering) need to be
considered.
Figure 1. Average pixel values of SIR-B vs Orientation
angle (Residential class)
P
9
2
E
ê e ee + + +. e € pixel value
= Std Dev
oD
3
€ am S n c £3
0 10 20 40 50 60
phi (degrees)
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996