A and of DEM 
tem (scanner as 
>uId be made by 
>lution DEM. 
3,1 village near 
-cause a recent 
iuse it is quite 
area: they are 
rences from top 
nts range from 
km. 
ing the hill side 
lined downhill; 
:e a main front 
arge blocks are 
’ a pattern of 
several areas, 
active. The hills 
shes and trees, 
it, resulting in a 
ige but (though 
the quarry, in 
'7, consists of a 
9 ^ 52) at the 
ampilation of a 
stained in the 
; flown around 
rom the banks, 
dfy the exact 
>f the project is 
the same area, 
ie span and to 
a topographic 
id in order to 
I generation by 
Three aspects 
?ht and camera 
les along and 
versus terrain 
,t height means 
areas prone to 
A longer focal 
and also help 
r perspective 
ant, this would 
worsening the 
•s, compare to 
s adopted in 
undermine the 
as only for the 
ised on this 
il lenght have 
ling flight. 
:ct in advance, 
and terrain 
>EM generated 
digitized with a DWS 100 scanner of the IRIS lab at 
JRC Ispra at a resolution of 12.5 urn (about 7 cm on 
the ground). Thought one of the goal of the 
investigation was to find out the best resolution value 
to be adopted, no test have been performed at the 
original resolution: the amount of data (more than 300 
MB per Image) proved too hard to handle, so images 
have been resampled to 25 pm. Further test will be 
executed by using a DTP scanner to check dependency 
on scanner performance. 
3. TEST ON DEM GENERATION 
Two digital photogrammetric systems have been used 
in the test: a high end system (hereafter called system 
A) and a low cost one (system B). An area of about 
320000 square meters has been selected for the test. 
5' to provide more than 80000 points on a PII 350 
machine. 
Fig. 2 - Aerial view of the Botticino quarry 
which encloses most of the quarry and an area mostly 
covered by trees and bushes (see Fig. 2). After 
performing a semi-automatic image orientation 
(automatic relative orientation with operator-assisted 
check for possible mismatches) System A was set on 
DEM generation with adaptive strategy: this means the 
system is capable of adjusting the matching parameters 
to the terrain and texture characteristics. 
Points were determined on a mesh 2 m wide, without 
supplying any a priori information on the terrain in the 
form of breakline location and so on. It took less than 
5045700. OC-C-i. 
i ì H 
5045600.00- 
5045500.00- 
5045400.00-^ 
5045300.00—^. 
The values have been compared to the interpolated 
reference data, resulting in a 2m RMS of the 
discrepancies, with about 50% of the values in the 
range (-0.5: 0.5m) m and a bias of about 0.7 m. 
Figure 3 shows a plot of the DEM errors larger than 1 
RMS (10% of the data) with the DEM contour lines in 
the background. As it could be foreseen, the largest 
errors occur on terrain breaklines. 
5045200.00-^ cy 
5045100.00- . 
5045000.004 
5044000.004 
5044600.00- 
5044700.00 
1602300.00 
1602500.00 
1602700.00 1602900.00 1603100.00 1603300.0- 
-1-CX3 is> -2.00 
-2,00 to 2.00 
2.00 » 30.7« 
Fie. 3 - DEM error plot for system A (2 m grid)