mm
VI Advantages of multitemporal data sets
What now is the advantage of a multitemporal data set compared with a classi-
fication of a single Landsat frame ?
We will try to answer this question quantitatively by direct comparison of
the different classification results in Table 1. The numbers listed there
are the percentages of correctly classified pixels within the corresponding
training sets. If we use the Landsat frames separately, we get the classi-
fication results given in columns 4 to 8. We see e.g. that in May, cereals
and root crop are quite well separable and in June some classes of cereals
become distinguishable.
But analysis of the 20-channel data set does yield significantly better
results for all categories.
VIII Conclusions
We have tried to show how and how much a relatively simple data enhance-
ment method will improve the classification result to a significant extent.
Even if the effort of combining a number of Landsat frames is not inconsi-
derable it does appear to be rewarding from the point of view of certainty
in classification and reliability in delineation of the boundaries.
We have still problems with the finely structures agricultural fields in
our country; those can probably be mastered when satellite data with better
geometrical resolution will become available.
IX Literature
(1) LICHTENEGGER J., SEIDEL K. + KUEBLER 0.: Methoden zur Ueberlagerung von
Landsat-Bildern für multitemporale Landnutzungskartierung. in: Bild-
messung und Luftbildwesen, No 46, 1978.
(2) EVANS W.E.: Marking ERTS images with a small mirror reflector. in:
Photogrammetric Engineering, No 40, 1974.
(3) DIXON W.J.: BMD Biomedical Computer Programs. Berkeley, Los Angeles,
London, Univ. of California Press, 1973.
Summary
The almos
Famenne,
Four land
abies dom
cultivate
17/5/1976
Homogenec
located b
were stat
shown thz
selective
Radiance
EXIO pros
and a dis
obtained
The super
four spec
duced by
The appli
economy :
Besides t!
sentatior
xels on í
can be aj
and then
the over:
