You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

International cooperation and technology transfer
Mussio, Luigi

Fig. 10 Detail of the 1954 Ortho-photo
3.1 Histogram interpretation
The first step of the image classification is the
construction of the gray level distribution histogram. On
each image two test areas which represent a forested
area and a non forested area respectively have been
Forested-non forested area in 1983
Gray level histogram
Fig. 8 Gray level histogram for test areas in 1983
A first classification scheme simply uses a threshold
value to assign a pixel to a forest area rather than to a
non-forest area according to its gray level value.
Binary maps for forested-non forested area have been
obtained in this way.
The threshold values have been chosen for each image
by making minimum the error of attributing a pixel to the
Forested-non forested area 1994
0.14 . - —
0.12 i —
a forested
© 0.1 4 .
' »non forested
2 0.0S 1 “
§ 0.06 .
0.04 .
51 o.o2 ;
Q ] »1 J j |.
1 21 41 61 81 101 121 141 161 181 201 221 241
Gray level Histogram
Fig. 9 Gray level histogram for test areas in 1994
wrong class.
Some problems arise using this classification technique:
high gray level values due to noise in forest areas result in
a leopard-skin pattern binary map and areas with dark
shadows are classified as forested regardless to their real
This can be easily verified by superimposing the obtained
binary forest maps on the ortho-photos (Fig. 11).
the gray level values using the image analysis capability
of Grass G!S. To get rid of gray level noise peaks, a low
band pass filter has been applied to the original images.
Using this method we are able to realize maps of the
Fig. 5 Orthophoto of the year 1983
extension of the vegetation in each photograph and to
compare them through the years.
Fig. 6 Orthophoto of the year 1994
The main step of this procedure is the classification of the
images. This is a well known problem in the case of multi-
spectral image analysis (Jensen, 1986) but no algorithm
can be found in the literature for panchromatic images
(images with only one band). Therefore the classification
that will be presented in the next paragraph is original.
Forested-non forested area in 1954
Fig. 7 Gray level histogram for test areas in 1954
0.02 -
0.018 .
0.016 .
O 0.014 .
2 0.012 .
c Ml .
Z 0 008
5 0.006 .
0.004 .
0.0C2 ...
0 ....
> B forested ar«v>
101 121 141 161 18i 201
Gray level histogram
002 .
f 0.01 s
8 001
0005 -