d using the.
After the
/ha) can be
nted blobs.
rmining ‘the
own can be
on the angle
entre of the
f the trees
of the tree
thod, and it
ree species,
> number of
, divided by
separated in
rees. In the
st concavity
n light), the
eparated. In
pine), a less
d. I tried to
s method is
/ inside the
GIS
to GIS. The
case of all
anaged and
d put to the
the vectors
TM - SPOT
P image, from 1:50000 - 1:15000 is the P1, and from 1:15000 -
1:500 is the P2. The coniferous-deciduous classification of the
Pl was compared to the classified Landsat TM. I joined the
seven classes to these two groups. The crosstabulation matrix are
shoed in Table 14.
Landsat TM |P1 | coniferous | deciduous
coniferous 0.98 0.01
deciduous 0.02 0.96
Table 14: The crosstabulation matrix of Landsat TM and P1.
Then I made a new attribute to the thematic coverage, called age-
group, and if it was necessary some polygons were divided, on
the strength of the P1 classes. The major roads out from P1 was
vectorised and conversed to an arc coverage experimentally, but
far wasn't proper.
The next step was the division of the stands (from the Landsat)
to part stands on the basis of P2 classification, and some other
attributes, i.e. unit trunk number/ha, average crown diameter,
health status, the blend rate were added to the polygon attribute
table.
The result is an ArcView project, in which a coverage with the all
extracted data are presented, with the appropriate raster
background, and containing the originally available data, i.e.
NFDB, DFBM, DTM for the analysis’s.
8. EVALUATIONS OF THE ACHIEVEMENTS
The results of the Landsat TM’s classification with the
maximum likelihood algorithm using empirical probability
density is very promising for forests. Although the determination
of the classes isn’t an easy task. The general distributions of
forest stands should be specified (e.g. for Hungary or for
Europe). A rough DTM determination is possible from the band
Iv.
The reclassification of the SPOT P image is a very simple, but
accurate tool for forest - non-forest separation. The SPOT P
hardly usable for coniferous - deciduous classification.
The Brovey-transformed Landsat TM and SPOT P is very
spectacular, and provides a good background for the vector maps
in GIS, but (mostly at different date of sources) almost useless
for classification purposes
The texture analysis of a 3 m ground resolution images (P1)
demonstrated that is suitable for age estimation.
The new way of classification of an 0.5 m ground resolution
ortho-images (P2) showed, that a lot of forest stands’ parameters
can be obtained (semi)automatically from images at this scale.
These items of information are unfortunately still too expensive
for the forestry practice, but hopefully it would change before
long.
9. ACKNOWLEDGEMENT
I acknowledge the data, the software and the support from the
DigiTerra Engineering Office, the data from National Forestry
Service (AESZ), and from the Mapping Office of the Hungarian
Army (MH TEHI).
10. REFERENCES
Baskent, E. Wardayo, W., 1997. The Integrated Use of
Geographic Information Systems and Remote Sensing in Forest
Ecosystems Management. In: International Workshop,
Application of Remote Sensing in European Forest Monitoring;
Proceedings Vienna, Austria, pp. 115-121.
Czimber, K. - Markus, 1., 1997. Mapping and Monitoring the
Forested Area Applying Remote Sensing and GIS. In: Archives
of Earthphoto Seminar, Budapest, 1997.03.12-13. pp. 104-106.
Eckstein, W., 1996. Segmentation and Texture Analysis. In:
International Archives of Photogrammetry and Remote Sensing,
Vienna, Austria, Vol. XXXI., Part B3, pp. 165-175.
Hocevar M., Kovac N., Kobler K. (1997): Application of Remote
Sensing and GIS from Landscape to Forest Stand Level. In:
International Workshop, Application of Remote Sensing in
European Forest Monitoring; Proceedings Vienna, Austria, pp.
33-45.
Hyyppä J., Inkinen M., Hyyppa H. (1997): Applying Multi-
Source Remote Sensing Data for Forest Inventory. In:
International Workshop, Application of Remote Sensing in
European Forest Monitoring; Proceedings Vienna, Austria, pp.
69-77.
Király, G., 1997. Large-scale Forest Inventory by Means of
Digital Images. Thesis work, University of Sopron.
Koch B., Fritz R, Kremmers Th. (1997): Possibilities to
Integrate High Resolution Satellite Data in Forest Inventories. In:
International Workshop, Application of Remote Sensing in
European Forest Monitoring; Proceedings Vienna, Austria, pp.
27-32.
Laws, K. 1980. Texture Image Segmentation. PhD. Dissertation,
Dept. of Engineering, University of Southern California.
Lohmann, G., 1994. Co-occurance-based analysis and synthesis
of textures. In: 12 International Conference on Pattern
Recognition, pp. 449-453.
Markus, 1, (Ed) 1995. PHARE MERA Forest Ecosystems
Mapping, Final Report, Sopron.
Narendra, P.M., Goldberg, M., 1980. Image segmentation with
directed trees. IEEE Transaction on Pattern Analysis and
Machine Intelligence, vol. PAMI-2, no. 2, pp. 185-191.
Shao, J., Förstner, W., 1994. Gabor Wavelets for Texture Edge
Extraction. In: International Archives of Photogrammetry and
Remote Sensing, Vol. XXX., Part 3/2, pp. 745-752.
Turner, M. G., 1989. Landscape Ecology: The Effect of Pattern
on Process. Annu. Rev. Ecol. Syst., 20, pp. 171-197.
Zhu, C., 1996. Remote Sensing Image Texture Analysis and
Classification with Wavelet Transform. In: International
Archives of Photogrammetry and Remote Sensing, Vienna,
Austria, Volume XXXI., Part B3, pp. 1036-1038.
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 369
