164
Nevertheless the satellite images are available only back
to the 70s. Therefore a time-wise analysis can be carried
out only for the last twenty, twenty five years.
This paper presents an original solution that uses aerial
images to automatically detect forest areas and their
evolution.
2. MATERIALS AND METHODS
2.1 Purposes
The purpose of this work is the automatic forest area
evolution. It has been accomplished in three steps:
1. Orthorectification of the aerial images;
2. forest area recognition on each image;
3. comparison of the forest areas on the different
images.
AH these task have been carried out in automatic with a
minimum operator work.
The forest area evolution has been used as an input for a
study of the social, cultural and economic study to explain
regional transformations.
2.2 Available images
Aerial images of the year 1954 and 1983 are available for
the chosen area, while the image taken in the year 1994
is already in digital format as orthophoto.
The first image has been taken from U.S. ARMY in the
year 1954 with a flight height of 9114 m above sea level,
corresponding to an height of 8800 m over the ground.
The focal length is 153 mm with a mean scale factor of
1:57500. The image has been digitised using a A4
scanner with a resolution of 1200 dpi.
Fig. 1 Aerial image of the year 1954
The second image has been realised by Provincia
Autonoma di Trento in the year 1983 for the technical
cartography. The flight height is 6636 m above the sea
level, corresponding to an height of about 6320 m over
the ground. The focal length is 153,11 mm with a mean
scale factor of 1:41300. The image has been digitised
using a A4 scanner with a resolution of 800 dpi.
Fig. 2 Aerial image of the year 1983
The orthophoto images has been realised by Provincia
Autonoma di Trento in the year 1994 using the “Volo
Italia' images, it is available in digital format with a ground
resolution (pixel size) of 1 m.
Fig. 3 Ortho-photo of the year 1994
2.3 image analysis
The aerial images have been ortho-rectified using the
Grass GIS software. The forest areas have been
recognized on the images by a supervised classification of
Fig. 4 Orthophoto of the year 1954
