Freiburg. The author also stated his familiarity with the area helped minimize confusion
errors in this instance. This area of southwest Germany was also analyzed by Schmidt
(1986) using VV-polarized X-band and L-band airborne SAR imagery. Depending upon the
kind and groups of buildings present, the author found a more-or-less regular pattern that
was easily visible on the X-band imagery. He also stated that isolated buildings were
visible on the L-band scene. Unfortunately, no elaboration of these points were given.
Sieber and Hartl (1986) classified land cover as one part of a SIR-B data observation
experiment of the Freiburg, Germany, area. Using three paths or data takes they
concluded that the steep incidence angle path (18 degrees) most significantly
discriminated among classes that included settlements as a category. Incidence angle was
considered to influence discrimination of land use classes more than the aspect angle.
Aside from their mention of three cities no measure of accuracy or pixel/sample size was
provided.
Trevett (1986) reported on a project to survey land use and vegetation in Nigeria
using a real aperture X-band airborne system and two look directions. In the northern arid
regions of the country, villages could not be distinguished as the mud huts and grass roofs
gave a signal response similar to the surrounding grassland. In the southern, more humid
forested areas of the country villages were detected as the corrugated iron roofs of village
dwelling gave a bright response that contrasted with the surrounding cover.
Imhoff, et.al. (1987) used SIR-B imagery as part of a project to survey flood
boundaries in Bangladesh. Using imagery with a 46 degree incidence angle a classification
accuracy of 83 per cent for a combined settlement-dike category was attained. Data on
omission errors (12 of 75 villages) and commission errors (13 agricultural areas were
mis-identified as villages) were also included in the report. The authors stated that the
raised elevation of the dikes and corner reflection from buildings against a background of
nonturbulent water in the surrounding rice fields abetted visibility. Of note is the fact
that these dwellings were also made of thatch. This contrast to the above mentioned
results of Trevett in Nigeria indicates that the change in the terrain background may be a
significant factor in settlement detection of the same building type.
It is apparent that few investigations have given attention specifically to settlement
detection and reasons for variation in the results. Fewer still provide quantitative data.
Many include only a passing, general reference to settlement detection as a side light to
broader or other-topic focussed research. In an attempt to broaden the understanding of
these radar/terrain relationships, this study examines effects of environment
(target/background contrast), radar incidence angle, and to some degree, look direction on
settlement detection.
STUDY AREA AND IMAGERY
L-band, like-polarized (HH) SIR-B imagery of southern and southwestern Germany
and northeastern France was recorded on three data takes in October, 1984. Relevant
parameters of these data are listed in Table I. As can be
205