better quality for the point determination. There haze produces the
effect of an increasing number of rejected points and a higher value of
mqr7, what originally had to be expected. The same has to be stated for
the forest areas too. In contrast to Deggendorf, where haze probably
reduces the errors in the projection, the surface topography for Vils-
biburg is uncomplicate, what makes it easy to approximate the object
correctly. That results in a higher accuracy and in more sensivity for
exteriour influences like haze.
6. FINAL REMARKS
Comprising the results of chapter 5 the accuracy of the correlation
procedure seems to be acceptable (ef. tab. 5).
object type mean square residuals [m]
field (FD) 22 - 28
forest (FO) 23 - 44
build-up area (BA) 21 - 31
Table 5 : Generally attained accuracy
Problems arised in areas with steep slopes, what especially for forest
resulted in a distinot decrease of marge Haze reduced accuracy for about
10-20%, except for steep slopes. But in general the object surface could
be approached very close. Using an a posteriori median filter process
the heights showed a very homgenous character with very good accuracy
(15 n < mary < 31 m).
The satisfying results might be due to the small scale of the model,
because
1. the frequent variations in the object characteristic
(fields- forest- urban area) gave enough structural information
for the correlation,
2. although the surface slope came up to more than 100% the
height variations in respect to the exposure distance were
relative small.
Therefore this example seems to be promising for further application of
the correlation technique onto small scale images.
Finally it has to be considered whether the attained results are good
enough for mapping purposes. As the correlation produces a DTM the
quality of the calculated heights has to be compared with the require-
ments for contour maps.
The accuracy of the heights determines the quality of the height repre-
sentation in the mapping scale, what affects the smallest contour inter-
val. Therefore it would be recommendable to start with the height accu-
racy and then to calculate the contour distance /Schoppmeyer, 1983/. But
due to the lack of experience for small scales, the direct calculation
of the contour interval failes. The consideration therefore has to be
oriented to the official german sample maps, for which the values of the
contour interval are listed in tab. 6.
From the cartographic point of view it is indipensable to have an accu-
racy for the determined height which is at least half that value of the
height distances in the map. Sometimes one third of the contour distance
is recommended as accuracy.
- 90 -