refore ten 
¢ structure 
irons. The 
"difficult" 
the "easy" 
m squared 
a correct 
of epochs 
1eeded 27 
racy. 
rks were 
g network 
id and 10 
) compare 
after the 
following 
umber of 
after the 
void this 
nent was 
lefined by 
racy, the 
roducer's 
d 96.8 96, 
‘he class 
s in the 
y of the 
raditional 
network 
nbination 
ry. In the 
nined for 
  
land cover map in the neighborhood of Budapest. The 
training of the individually designed networks was 
executed by the application of the Levenberg-Marquard 
optimization. This algorithm could accelerate the learning 
so the preclassification time was decreased, but the 
memory need has been enlarged. The independently 
trained networks can bypass this problem. In order to 
accelerate the network simulation an equivalent 
transformation is executable. Using this procedure the 
speed is drastic raised. 
  
  
il 
   
  
nd Ed 
  
Jb. 221 Do Pa 
  
Figure 4 
User's and producer's accuracy for the thematic classes 
Taking the output of the neural network as fuzzy class 
memberships concerning of the thematic categories the 
fuzzy decision and its tools are applicable. Thanks this 
achievement we can take into consideration not only 
intensity but other image and not image like terrain 
information. 
The quality of the result map is equivalent to the 
traditionally made land cover maps. It has an enormous 
advantage: it's flexibly extensible when further terrain 
data are ready so the method can be made suitable for 
land use mapping. 
References 
Barsi, Á., 1996. Thematic Classification of a Landsat 
Image Using Neural Networks. International Archive of 
Photogrammetry and Remote Sensing, Vol. XXXI. Part 
B3, pp.48-52 
Barsi, Á., 1997. Landsat-felvétel tematikus osztálvozása 
neurális hálózattal. Geodézia és Kartográfia, Vol. XLIX, 
No.4, pp.21-28 
Brause, R., 1995. Neuronale Netze. B.G.Teubner. 
Stuttgart 
Cox, E., 1994. The Fuzzy Systems Handbook. A 
Practitioner's Guide to Building, Using and Maintaining 
Fuzzy Systems. AP Professional, Boston 
Grace, A., 1995. Optimization Toolbox for Use with 
MATLAB. MathWorks Inc., Natick 
Nauck, D. - Klawonn, F. - Kruse, R., 1994. Neuronale 
Netze und Fuzzy Systeme. Grundlagen des 
Konnektionismus, Neuronaler Fuzzy-Systeme und der 
Kopplung mit wissensbasierten Methoden. Vieweg, 
Braunschweig 
Rojas, R., 1993. Theorie der neuronalen Netze, Eine 
systematische Einführung. Springer Verlag, Berlin 
Rumelhart, D.E. - Hinton, G.E. - Williams, R.J., 1986. 
Learning Representations by Backpropagating Errors. 
Nature, No.323, pp.533-536 
Zadeh, L.A., 1988. Fuzzy Logic. Computer, Vol.21, 
No.4, pp.83-93 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 327