er
Irs
ion of
on
trees
st
, Van
o the
ial
nal,
ee top
ional
ational,
eeing
in
SRI
ly.
nage
sing.
)0
United
order of
MULTISCALE IMAGES IN FORESTRY
Király, Géza
Departmental Engineer, PhD Student
University of Sopron, Department of Surveying and Remote Sensing, H-9400 Sopron, Hungary
email: kige@efe.hu
Commission VII, Working Group 4
KEY WORDS: Classification, Texture, Ortophoto, DTM, Accuracy, Forest Inventory, GIS.
ABSTRACT:
The paper presents forest classifications based on different sources. The complementary and comparing forests classification of
Landsat TM, Spot Pan and scanned aerial photographic images are presented. It shows a new way of classification which the large scale
texture and some other, non-imaging data (e.g. digital terrain model (DTM)) are involved in. The results of the different-scale
classifications are combined, put to geographical information system (GIS), and analysed from the forestry point of view. The
classification procedures are presented on a test area, in Hungary with a comparison with traditional methods. The extracted stands’
parameters are compared with the traditional forest inventory data. The paper concludes with an overall evaluation of the method, and
gives some ideas on using it in the forestry practice.
1. INTRODUCTION
Recently the Geographical Information Systems in forestry have
become more and more important. The fast changes of the
circumstances have put a significant emphasis on remote sensing
and GIS. The social expectations have been changed from the
timber production to the multipurpose forest functions. The main
purposes of Remote Sensing in forestry are determining the
stands’ parameters for inventories and detecting the changes for
better forest plans for the community. In this paper I'm only
engaged in the forest stands’ parameters.
There are two ways of getting better results. One is using better
and better sources, such as hyper-spectral cameras, the other is
developing more sophisticated methods, based on the same old
sources. The first one involves new methods too, of course. Due
to limited possibilities, I applied the latter way.
2. THE TEST AREA
The representative test areas are situated beside Sopron, at the
western board of Hungary (Figure 1), called the Mountainous
Region of Sopron.
Figure 1: The situation of the test area.
These area is a medium-height mountains, the altitude of the area
is between 200 and 560 meter. The annual mean temperature is
9.0-9.5 °C, the annual precipitation is about 700 mm. The main
soil type is brown forest soils, especially acidic, non-podzolic
brown forest soils. The natural vegetation of the area is
submontane oak-hornbeam forests and submontane beech
forests. The present forests are cultivated mixed forests, with a
lot of coniferous species, not naturally, but very diverse, a good
area for testing classification methods. Nowadays there are
serious woodborer disasters in spruce stands. Some select cutting
stands are represented too. The overall impression of the region
is very diverse, because our ancestors thought that a wide variety
would mean better forests.
3. THE APPLIED DATA
I used a Landsat TM, a SPOT Pan and a scanned aerial images in
my investigations. The Landsat TM - O ESA (1991), distributed
by EURIMAGE, FOMI RSC - was from 5® of October, 1991,
has a 25 m ground resolution. The SPOT image - © CNES
(1995), distributed by SPOT IMAGE, FOMI RSC - was made
on 11” of august, 1995, with 10 m ground resolution. There were
two series of aerial photographs. The first series was a false
colour infrared contact copy at scale 1:31°580 approximately,
made on 29" of September 1991, and was scanned with 300 dpi-
true colour RGB (approx. 2.67 m ground resolution)(P1) and the
other series was a false colour infrared contact copy at a scale
1:8°300, scanned with 300 dpi-true colour RGB, too (approx.
0.7 m ground resolution)(P2). I used a digital topographic map at
a scale 1:50'000, named DTA-50 - € MH TÉHI, a digital forest
base map (DFBM) at a scale 1:10°000 - © AESZ - DigiTerra
Eng. Office, the National Forest Database (NFDB) - © AESZ,
and a 10m-10m grid base digital terrain model, named DDM-10 -
© MH TÉHI.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 365