35 
ause, by using the 
> (represented by 
ing data to identify 
Df global accuracy, 
was performed, in 
with the correction 
il context (digital 
ial context is the 
>nly one land cover 
id from the digital 
ie GIS environment 
ition in the image 
assification filtering 
to allow speckle 
r 
from the randomly 
by-pixel basis with 
ellite imagery. This 
5 the results of both 
latrices (Tab. 1), k 
inclusion of the 
! IKONOS image 
f the results (from 
ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
Class 
Description 
1 Ml SOWN GROUND 
2 ARTERIAL ROAD 
3 HI MEDITERRANEAN BUSH 
4 Hi HIGH DENSITY BUILDINGS 
5 UNCULTIVATED GROUND 
6 OLIVE-GROVE 
7 fgm COUNTRY ROAD 
8 T: ; ^3 ASPHALT ROAD 
9 |i| LOW DENSITY BUILDINGS 
Fig. 3 “Per-field classified study area" 
The per-pixel method had produced misclassifications, not only 
as a consequence of internal variability within fields, but also due 
to the utilisation of land use as opposed to land cover classes. 
Per-field classification was able to overcome these problems in 
some instances, but in others, such as inexact geometric 
registration and errors in the original vector data, they meant in 
the selection of uncorrected classes. 
5. CONCLUSIONS 
Flexibility of the integration process in the present software 
packages and the high spatial resolution of the new generation 
satellite imagery may globally lead to an increase in geometric 
detail and accuracy with which land cover can be mapped over 
images of coarser spatial resolution in which, as many recent 
researches attest, the presence of mixed pixel is the dominant 
problem to resolve. 
At present, an increasing amount of geographical data are 
stored in geographical information systems. This data could 
prove useful in the processing of remote sensing images. In 
addition, remote sensing images may be considerably applied to 
store and update data in a GIS. 
The per-field classification developed in this paper should be 
considered as a test to validate on a local scale well established 
methodologies of classification applied at on over-regional scale 
and future researches are advisable to reduce sources of 
resulted misclassifications. Moreover, such types of studies 
allow opening of new unexplored techniques applicable on large 
scales by adding meaningful inputs to those multidisciplinary 
studies connected to decision-making and planning activities. 
6. REFERENCES 
Aplin, P., Atkinson, P. M. and Curran, P. J. (1997). Fine spatial 
resolution satellite sensors for the next decade. International 
Journal of Remote Sensing, 18, pp. 3873-3881. 
Carbone, G. J., Narumalani, S. and King, M. (1996). Application 
of remote sensing and GIS technologies with physiological crop 
models. Photogrammetic Engineering and Remote Sensing, 62, 
pp. 171-179. 
Cowen, D. J., Jensen, J. R., Brensnahan, P. J., Ehler, G. B., 
Graves, , D., Huang, X., Wiesner, C. and Mackey, Jr, H. E. 
(1995). The design and the implementation of an integrated 
geographical information system for environmental applications.