e Dossier Level 
At this stage the final classification of the whole 
declaration (dossier) is being made to “accepted”, 
“rejected” and “doubtful”. Table 6 indicates the rules 
for classified the dossier. 
The identification of the land use in a first stage for each 
plot was made carrying out supervised classification on 
multitemporal images set. The results of the classification 
were checked against ground samples different than the 
ones used in the classification algorithm. The obtained 
accuracy of the classification was better than 85% for any 
classified land use, which were cereals, cotton, corn, 
clover and vegetables. 
For the identification of a specific use in a plot, the 
accuracy of the classification for the specific use must be 
greater than 75%, it must appear in an area bigger than 
7096 of the plot's total area and the second found land- 
use must be less than 20% of the plot's total area. This 
was done by computing summary tables, resulted from 
the overlay of the vectors on the results of the 
classification. 
Those plots that can not be controlled by the supervised 
classification or have been declared for set-aside, should 
be controlled by visual interpretation of the satellite 
images on the computer screen on plot level. However, 
all the plots were finally photo-interpreted to check for the 
adjustment of their boundaries and their area, and the 
automatic classification was finally used only to check 
land use. 
In AITO zone, for 300 applications which were randomly 
selected, control was additionally carried out to check 
whether the declared for subsidy plots were arable during 
period of 1990 -1991, using multitemporal LANDSAT TM 
images of those years. 
CAPI (Computer Aided Photo Interpretation) was carried 
out in a specially designed environment of the ERDAS 
Imagine 8.2 software. The environment allows for the 
simultaneous use of several windows, geographically 
linked among them, to display the multispectral, 
panchromatic and classified images with the vector 
database which contains all necessary information and 
provides entries for the user to add control results such 
as control code (see Table 4) and measured area. 
CAPI results were then processed by our diagnosis 
software to perform control at group and dossier level. 
Finally all the “rejected” and 20% of the “doubtful” 
applications were to be checked by the Ministry of 
Agriculture by ground truth survey and the results were 
returned for statistical analysis. 
The diagnosis software was also used to print three types 
of lists, on plot, group and dossier level, that were 
handled to the administration. These were accompanied 
by printed maps in A3 size and 1: 10,000 scale, for 
helping the on-the-spot controls. The maps contained the 
SPOT-P image, the parcel boundaries (changes in 
boundaries introduced by CAPI are indicated in a different 
colour) and cartographic reference codes (parcels to be 
field inspected are again indicated by code numbers of 
different colour). All documents and maps were prepared 
in a community level. 
740 
5. RESULTS 
In dossier level, the results of Remote Sensing were 5094 
accepted, 35% rejected and 15% doubtful. 
By comparing the results of Remote Sensing and those of 
the on-the-spot checks for 3,400 plots, differences in 
land-use and area (more than 0.1 Ha) were occurred for 
less than 5% and 3% respectively. Differences in land- 
use are mainly caused from the delayed sawing or re- 
sawing of some plots with maize (corn) cultivation caused 
by abnormal weather conditions in the control zones, 
while differences in area were founded in some plots 
which were cartographicaly referred on ortho-photomaps 
and the delineation of their boundaries were inaccurate. 
6. REFERENCES 
JENSEN J., 1986. 
Processing Prentice Hall, New Jersey, USA, pp 379. 
Introductory Image 
GAF/GEOMET, 1991. by 
Remote Sensing of some Agricultural Production, 
Region GREECE, Final Report, Munich, GERMANY 
Integrated Control 
GEOMET, 1992b. Remote 
Temporary Set-Aside and Oilseeds Vn Greece, Final 
Report, Athens, GREECE 
Sensing Control of 
GEOMET, 1993b. Remote Sensing Control of Surface 
Subsidised Arable Land and Forage Areas, Final 
Report, Athens, GREECE 
CHIESAC., TYLER W.,1994. Cubic 
Convolution: ERIM Restoration for Remotely-Sensed 
Beyond 
Imagery, Earth Observation, Volume 3, Number 2, 
USA 
EAGGF,1995. RECOMMENDATIONS FOR 
THE 1995 REMOTE SENSING CONTROLS, Brussels, 
BELGIUM 
ERDAS Inc., 1995. Erdas Field Guide, Third 
Edition, Version 8.2, Atlanta, USA, pp. 628 
GEOMET, 1995b.Remote Sensing Control of Surface 
Subsidised Arable Land and Forage Areas, Final 
Report, Athens, GREECE 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996