has the produced by different DEM densities of the same data while 
the same orientation parameters are used. 3 er "E 
The test was carried out on a part of the image limited by sind dan : 5 
e four maximum pixel No. = 2216 and maximum line No. = 5065. SN s 
for the The corresponding DEM data is generated by digitizing the 3 
ot give contour map and the DEM was produced with an original * E 
tage (a regular grid of 100 by 100 m; covering about 140 square oil 
kilometers of flat, moderate and rocky mountains areas. Sid i 
The study area is divided into smaller areas of 2 by 2 km Sabi o be 
IS and these areas are classified according to terrain roughness Soci 
(Equation 3). Toe Oo l L 1 L L 1 1 
Side In the first stage the orthoimage is resampled using the $200, 300 400, 600 ; ern a in 
; eli x Gri i m 
1986) original DEM — data (of 100 by 100 m regular grid), as a nid er omi 
; “i 1 ; 1 Figure 3 . Moderate Terrain (Anchorpoints Technique) 
EN a position il as DEM pom is i on nen 6 Affine span. 1.2% Eight parameters fran. 
e original image. e second stage orthoimages are 
resampled using a courser DEMs (grid size varying from 200 Sos 
condi by 200 m. to 1000 by 1000 m.), were some points where X [[eseries 1 ? 
omitted from the original DEM data. In order to be able to S {Series 2 
fi judge the relative geometric quality of the orthoimages which E 8 Series 3 
Bificant are resampled using the courser DEMS (second stage), the 8 Eseries 4 
position on the original image of points forming regular 4 E 
oe ground grids of 100 by 100 m as given by these orthoimages z ire 
8 is computed and compared with the position of the same 5 4 M 
points as given by the orthoimage generated in the first stage. 5 bt 
eeds , n . : deett S c0L Ee 
Then the discrepancies in units of pixels of the original image > m 
it between the position of these points is used to compute RMS = 
its two : 3 ; : 13 c; 0 1 L l 1 1 1 l 
d and is compared for each terrain type with different grid size 66 5060009400 Tae Date: BT Bored 
and different resampling algorithm. The following Figures Grid Size {m} 
Er 
m del ( Figure 1 to Figure 6) show the plot of RMS of vector Figure 4 . Moderate Terrain (Pixel by Pexel Technique) 
; mode displacement (i.e. in line and pixel directions) against the Series 1=Nearest Neighbour  2= Inverse Distance 
unately grid size for different resampling method. 3=Inverse Square Dist. 4= Bilinear Polynomial 
S study. ——ÀQ - 
> terrain a PM © 3. 
% 2 A € Series 1 
X [|e Series 1 T a A 
D, = PISeries 2 ni = 
E a 
5 3 
>rrain. a a 
a 1 = 
n dar a ; Les 
o 0.5 Eee et INNER s 
7 $ 
5 g 9 
oblique 0 = 200 300 400 500 600 700 800 900 1000 
| mirror 4 200 300 400 500 600 700 800 900 1000 Grid ize im) 
Li thi Grid Size (m) 
in this | 
286 and Figure 1 . Flat Terrain (Anchorpoints Technique) Figure 5. Rough Terrain (Anchorpoints Technique) 
m Series 1= Affine tran.;  2= Fight parameters tran. Series 1= Affine tran.;  2- Eight parameters tran. 
iced by à 
[^] c- 
)00 and 3 ? Eu 
control a Series 1 = a Series 
[s were 2 us S [Series 
. These s : © [Series 
-m. The à a Flseries 
À e 3 siepe 
the 18 $ s 
he lack 8 $ 
control $ s 
MS) of 2 0. 2 
uals are 5 = 
] points 2 © 
rS map 2 oo 300 400 500 600 700 800 900 1000 — 
sults of Grid Size {m} 2 %0 800 400 500 600 700 800 900 1000 
i i i Pexel Technique Grid Size (m) 
)jmmages Figure Elie Torre exe Senn e distance Figure 6 . Rough Terrain (Pixel by Pexel Technique) 
= ist. 4= Bilinear Polynomial Series 1=Nearest Neighbour 2= Inverse Distance 
geinverse Square Dis 3=Inverse Square Dist. 4= Bilinear Polynomial 
  
  
  
  
  
249 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996