erman
for
ler >,
Y Of
for
1 for
sical
ETH
ering
| and
Dr.H.
ffen-
for
data,
gital
entes
alti-
-ein
ozes-
12:2,
enha-
ation
econd
ce of
1993,
ı the
esent
ophy-
3 rin
pl.9,
und
der
sen).
iss.,
r-
ften,
USING BACKSCATTER FROM RADAR IMAGES FOR CLASSIFYING AND DETERMINING THE BULK DENSITY OF
THE URBAN ENVIRONMENT
Catherine Ticehurst (Ph.D student), Bruce Forster (Professor) & Yunhan Dong (Research Assistant)
School of Geomatic Engineering
University of New South Wales
Sydney NSW 2052 AUSTRALIA
Ph: 612385 4182
Fax: 612 313 7493
Commission VII, Working Group 9
KEY WORDS: Land Use, Urban, Classification, Radar
ABSTRACT
The urban environment is a mixture of buildings (of varying size, shape, and density), and vegetated and non-vegetated
open spaces. The proportion of each of these characteristics in an area is generally related to its land use. Radar is
showing potential in assisting classification of the built environment due to its close correlation with the bulk density of
buildings. However one of the problems with using radar is that the backscatter is particularly sensitive to the radar look
direction with respect to street orientation. This paper examines this property, and how different aspects of radars
interaction with the built environment can assist in urban classification.
INTRODUCTION
There is a need to classify the environment into its
various land uses to keep records of a cities size and
layout. Some cities in developing countries are growing
rapidly, making it difficult for governments to maintain
these records by conventional surveying and mapping
methods. Satellite and airborne radar remote sensing
provides a relatively cheap and fast method of acquiring
up-to-date information about the environment, especially
in regions where cloud and rain may affect visible and
infrared sensors. Radar images give detail about the
shape and physical properties of the earth's surface, and
are showing potential for determining the vertical bulk
density (as compared to planimetric density from
visible/infrared sensors) and for the classification of the
urban environment into its various land uses.
Although there has been a substantial amount of analysis
done on radar and the environment, very little has
concentrated on the urban environment alone. Bryan
(1982) and Hardaway et al (1982) have analysed radar
backscatter with respect to urban street orientation and
showed a strong correlation. Henderson (1985) carried
out a detailed study using dual-polarised Synthetic
Aperture Radar (SAR) to classify the urban environment,
but found the radar had some confusion between classes.
Deguchi et al (1995) examined the relationship between
building coverage ratio, from visible/near infrared images,
and bulk ratio of urban environments, using radar data.
Forster et al (1996) have also investigated the possibility
of combining visible/near infrared data of a city, with radar
information.
Single polarised data provides limited information
compared to multipolarised images. Some work has been
undertaken on quad-polarised radar and polarisation
Signatures, using them to distinguish between urban,
park, forest, geology, and ocean. (See for example,
Fvans et al (1988), Van Zyl et al (1987), and Zebker et al
(1987)). These studies looked at the urban environment
I^ a broad sense rather than classification of the built
environment into all its land uses. A detailed study of the
types of scattering in the urban environment from
709
multipolarised radar by Dong et al (1996) has shown
some promising results.
The present study involves examining radars interaction
with the urban environment so that information about land
use and the bulk density of the urban area can be derived
from radar images. The bulk density is related to the
volume of buildings in a particular region. For example a
central business district has a higher bulk density than a
residential area. An initial investigation looks at the
relationship between radar backscatter and urban land
use classes (obtained from a SIR-B image over Sydney,
Australia). A model has been developed to give the
expected radar backscatter from a block of buildings, and
its results are also presented.
RADAR BACKSCATTER IN THE URBAN
ENVIRONMENT
Radar transmits a wave of known length, polarisation,
azimuth direction, and incidence angle. The built
environment has properties, such as building size, shape,
orientation with respect to radar, and material (surface
roughness and dielectric properties including moisture
content), with which the radiation interacts, and
determines the properties of the backscattered radar
response. Radar is particularly sensitive to building "bulk"
and the orientation of buildings with respect to the radars
look direction.
One of the main contributions to radar backscatter in the
urban environment is due to corner reflections. Double
bounce corner reflection, or dihedral reflection, occurs
when the radar look direction is perpendicular to a
building wall. The radar wave will bounce from the ground
to the wall and back to the sensor, or vice versa.
RELATIONSHIP OF LAND USE AND RADAR
BACKSCATTER
A preliminary study has been undertaken which involved
examining SIR-B data (L-band, HH) over a region in the
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
