ISPRS Commission III, Vol.34, Part 3A ,,Photogrammetric Computer Vision“, Graz, 2002
history does not exist for the first epoch. Section 6.1 describes
the semantic nets used.
Resulting from interpretation of the first epoch is a scene
description with interpreted segments of moorland. For
interpretation of the next epoch these segments are used as
input for a prediction of state transitions. This prediction uses
prior information concerning possible changes. The possibilities
are represented in a state transition diagram (see section 5). The
output of the prediction are “predicted new states” for every
segment. The segment borders may change during
interpretation intervals. Therefore it is necessary to perform a
resegmentation of the already interpreted segments by
additionally using the images of the new epoch. The
resegmentation results in segments with possibly new borders,
which are used for interpretation of the new epoch together with
the semantic nets for multitemporal interpretation and the
possible state transitions. Thus, the iteration continues for all
epochs to be interpreted. For more information see (Pakzad,
2001).
7. RESULTS
Pas
1998
Figure 4. Aerial images from test area
Test area is the moor “Totes Moor” northwest of Hannover in
Lower Saxony. The methods developed were tested for two
parts of the test area. Aerial images with a resolution of
0.5m/pel from five epochs were used. For the part in figure 4
we used the epochs 1975, 1981, 1986, 1989 and 1998. The
input sources of the last two epochs are CIR-images, the rest are
greyscale. The reason of using greyscale images is that most
available aerial images are greyscale images, despite of the
better information contents of colour images. Figure 5 shows
the obtained results of one part of the test area. The size of the
area is 3288m x 2964m. The initial segmentation for 1975 led
to 35 segments. Most of them were assigned to the class “area
of degeneration”. The resegmentations during the interpretation
for the five epochs led to a number of 96 segments for 1998.
The amount of segments, which were classified as “area of
degeneration”, decreased during the epochs, while other classes
increased. The reason is the fact, that at the beginning of
interpretation the uncertainty of the interpreted classes was
high, especially as the images of the first epochs were greyscale
images, and the probability for the class “area of degeneration”
is higher than for the other classes.
1978 1981
199€
Forest Grassland Area of Milled Peat Milled Peat — inactive Area of Area of Area of Re-
Degeneration Area Strip Area of Regeneration Regeneration generation
Extraction Extracton Peat Extr. Birch State Heather State Wet
Figure 5. Results of moorland interpretation
A - 238
