Istanbul 2004 
: the proposed 
hm provides 
mes compared 
algorithms. At 
Casing together 
curacy of the 
e accuracy of 
it should be 
thm of density 
necessity of 
variant to the 
retation 1s the 
features of the 
etation. In the 
osed to define 
or ANN ways 
es of classes. 
is with use of 
"he purpose of 
ys of forming 
about classes. 
go 
L2 
as 
04 
H5 
e2b 
ical and ANN 
model images 
s of accuracy 
€ space. 
| — traditional 
' RP algorithm 
vased upon (3) 
gorithm using 
— the proposed 
characteristics 
wing types of 
d: typela and 
with three and 
he second type 
nd six bands 
irate results of 
orming feature 
| multispectral 
ase of ANN 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
classification with type2a image when spectral information 
almost is absent (the one band only), and classes textures are 
strongly marked. 
At the same time Figure 4 presents sufficient advantages of 
proposed complex nonparametric algorithm, which is based 
upon first-order Haralick texture characteristics and 
nonparametric optimization of feature space. 
Besides, the investigation on the basis of time series 
mutlispectral RS images for the Uymon steppe area are 
conducted. The spatial resolution of RS data is 30 meters. The 
major classes of the RS images are the agricultural land, 
bushes, water, and vegetation. 
  
  
  
a) b) c) 
Figure 5. Time series RS images from RESURS-OI 
  
Figure 6. The fragment of forecast map designed for 2001 by 
various techniques 
Figure 5 shows the time series RS mages for 1998, 1999 and 
2000. Figure 6 shows the fragments of forecast maps for 2001 
Obtained by three various techniques: a) by the traditional 
maximum likelihood classification without CA; b) by the 
traditional maximum likelihood classification and ordinary CA; 
€) by the proposed algorithms of advanced interpretation and 
enhanced TS analysis. 
Figure 6 shows that the best forecast map is in c) fragment. It is 
a less noisiness and it has more legible classes that could say 
497 
about availability of the proposed algorithms for designing 
forecast maps. 
CONCLUSIONS 
Permanent improving of comprehensive satellite systems allow 
to obtain more qualitative RS images of high resolution, which 
might be effectively used for designing of thematic and forecast 
maps. Increasing amount of information requires the 
specialized processing, RS and mapping software, forecasting 
land use/cover change models. 
The approach to the designing algorithms for advanced 
interpretation of RS images and on their basement including 
GIS designing forecast maps is developed in the paper. The 
proposed approach is based upon nonparametric statistical and 
ANN classification using spectral and spatial features. To make 
the forecasting more adequate it is proposed the approach with 
use of high-order Markov chains and CA with optimal 
neighborhood size. 
The preliminary investigation results conducted with use of 
model RS images and TS RS images for Uymon steppe area, 
show high efficiency of the proposed approach and availability 
of further research in that field. 
References 
A.V. Zamyatin, N.G. Markov, 2004. Multi-stage Processing of 
Time Series Aerospace Images for Obtaining Enhanced 
Forecast Land Cover Maps. In: Proceedings of the 7th AGILE 
Conference on Geographic Information Science, Heraklion, 
Greece, AGILE, in press. 
Baker W. L., 1989. A review of models of landscape change. 
Landscape Ecology 2, pp. 111-133. 
Haralick R.M., Joo H. A, 1986. Context Classifier, [EEE 
Transactions on Geoscience and Remote Sensing, N24, pp. 
997-1007. 
L CR. Eastman, 2001; Idrisi32" “yelease 2 
http://www. clarklabs.org, (accessed 1 Nov. 2003). 
tutorial, 
N.G. Markov, A.A. Napryushkin, A.V. Zamyatin, 2003a. 
Advanced Thematic Mapping Approach for Forecasting 
Landscape Change Using GIS. In: Proceedings of the 6th 
AGILE Conference on Geographic Information Science, Lyon, 
France, AGILE, pp. 687-693. 
N.G. Markov, A.A. Napryushkin, A.V. Zamyatin, E. V. 
Vertinskaya, 2003b. Adaptive Procedure of RS Images 
Classification with Use of Extended Feature Space. In: 
Proceedings of SPIE, Vol. 4885, pp. 489-500. 
Park, S. and Wagner, 1997. D. F., Incorporating Cellular 
Automata simulators as analytical engines in GIS, Transaction 
in GIS, vol.2, no. 3, pp. 213-231. 
Verburg P.H. et al., 2003. A method to analyse neighbourhood 
characteristics of land use patterns, Computers, Environment 
and Urban Systems, in press. 
Acknowledgements 
The authors are grateful for the support this research bv grant 
N903-07-90124 from Russian Foundation for Basic Research.