(2.5 m) gives a modest improvement, while a further reduction
of the output pixel size to 20 um (1 m) improves the result on
vineyards only, while it remains unchanged for other features or
even getting worse. This indicates, that the limit due to the
resolution of the observers eyes is between the 100 um and the
50 um pixel size, as suggested by Doyle (1982)
The orthophotos from the photography 1:30,000 should be
compared with the one from the 1:60,000 images of the same
input and output resolution on the ground, thus the third case in
table 5. The clearly inferior result can be explained from the
clearly lower contrast in this photography, which seems to
hamper mono interpretation but not the stereo interpretation.
Moreover we had no control over the photographic processing,
and there were indications, that there were problems in this
particular case. New contact prints gave almost the same result
as from the photography 1:30,000, but this test was not done
under the same circumstances and it was not completed.
5. CONCLUSIONS AND RECOMMENDATIONS
Scanning wide angle aerial photography of standard resolution
with a pixel size of 15 um gave a significant reduction in the
interpretability, especially for isolated small houses. Pixel sizes
of less than 15 um should be tested too. For high resolution
photography, which has a 3 to 4 times better overall resolution,
separate tests with much smaller pixel sizes will have to be
made.
The test presented here shows the dependency of the interpreta-
bility of the tested features on the ground pixel size rather well.
Although a generalization to completely different environments
can not be made can be used is input to find a good trade off
for image scale, pixel size and field completion. Other
parameters, like costs of fieldwork and additional costs for
larger photoscales and additional time/costs for smaller pixel
sizes depend on many circumstances and are not treated here.
The results on interpretations of orthophotos 1:50,000 are
obtained by a well trained operator. Interpretation is highly
subjective, depends strongly on the experience of the interpreter
and can thus not be generalized for the average map user, who
depends even more on the resolution of the images. For optimal
interpretability the ground pixel size should be smaller for the
input image than for the output image. For orthophotomaps
1:50,000 the input images should therefor have a pixel size of
less than 3.6 m on the ground, and the orthophotos less than
100 um at presentation scale. Going to pixel sizes below 1.8 m
for the input or 50 um for the output does not increase the
interpretability for observation with unaided eyes.
Orthophoto maps at scale 1:50,000 will need annotation mainly
for important point features. For line- and area features this is
not so much required if good images and appropriate pixel sizes
are used in the orthophoto production.
Extrapolating from our results we would recommend for future
tests photoscales 1:80,000 to 1:100,000, scanning with 15 um to
30 um and printing with 50 um pixels. For high resolution
photography even smaller scales can be tried, but the scanning
pixel size should be less than 2.5 m on the ground.
6. REFERENCES
Leberl, F.W., 1992. Design alternatives for digital
photogrammetric systems. In: International Archives of
Photogrammetry and Remote Sensing, Washington, DC-USA,
Vol. XXIX, Part B2.
Leberl, F.W., & al. 1992. Photogrammetric scanning with a
square array CCD camera. In: International Archives of
Photogrammetry and Remote Sensing, Washington, DC-USA,
Vol. XXIX, Part B2.
Schiewe, F.W. and Siebe, E., 1994. Revision of cartographical
databases using digital orthoimages. In: Proceedings of the
Symposium of Com.III of ISPRS, Munich, Germany, Vol. 30,
part 3/2.
310
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
Key
1.In
Thre
very
the
the
and
and
Furt
avai
visu
for :
the
autc
date
Spe
imp:
mod
expe
in tr
only
tool:
exai
aspe
the
appe
and
data
proc
to c:
3D «
and
man