belonging to a point cloud (in our case, the two sets of errors estimated with the 2.5D accuracy tests may be due to 
measurements). The generated DEMs have been compared to modelling errors or to the planimetric errors. Again, the larger 
the reference ones. The results are reported in Table 7. It can be errors have been found in mountainous areas (DEM 5-2), while 
seen a large reduction of tlie mean distance and standard in flat terrains the accuracy of the generated DEMs is very 
deviation in all datasets. This demonstrates that part of the —— good. The two orientation procedures give very similar results. 
Table 5. Accuracy results using 2.5D comparisons. All results are in meters. 
DEM | Number of points All points All points Without trees Without trees 
Procedure 1 Procedure 2 Procedure 1 Procedure 2 
  
| 
'anbul 2004 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
| 
  
IGP  |Reference 
DEM DEM 
Min Diff. 
Mean 
RMSE 
Max Diff. 
Min Diff. 
Min Diff. 
Max Diff. 
Mean 
RMSE 
Max Diff 
Mean 
RMSE 
Max Diff. 
Min 
Diff 
Mean 
RMSE 
  
1 35448| 1000000} 22.1|-26.1| -4.0| 6.2} 25.1|-32.9] -2.6| 5.7} 13.8| -23.6| -3.0| 5.4} 15.4|-23.7| -1.7| 46 
2 32932| 1000000[ 37.7] -37.1| -3.0] 5.5} 29.1) -37,1| -1-2) 5.0) 32.3/-31.2/ -1.8]. 3.9| 29.1] 31.7| 02| 3.6 
3 33450. 1000000] 19.7] -17.8| -1.9. 4.0] 20.7] -17.2| -0.5j 3.2] 19.6|-14.1| -2.3| 3.4| 20.7]-13.6| 0.1| 2.9 
4 32067| 1000000] 11.5|-21.7] -3.8| 5.2} 13.6|-23.1} -25| 471 9.5/-17.1| -2.8) 3.9] 10.5]-18.4| -1.2;7 32 
5-] 10327 21200] 271|-36.8| -6.7| 9.4] 1932] -33.58| -5.8 83] 16.11-23.3| -34|. 53] 19.1] -13.3] -1.7| 4.9 
5-2 71795| 139200}130.3| -86.1| -5.7| 11.2{136.8| -89.3| -4.3| 9.5}104.7| -55.5| -4.0| 7.4] 49.8/-66.8] -1.3] 6.7 
6 130558/  600000|-22.7|-19.0} 11] 3.81 26.8[-27.1| 1.51 4.01 21.9/-14.6| -0.7| 3.9} 26.8/-25.9| 2.1| 44 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
called 5-1 
ith bias up 
VI 6, all the 
DSMs are 
tained by 
that the 
= 1; The 
are in the 
Table 6. Accuracy analysis based on terrain heights. First column: name of reference dataset. Second column: 2D distribution of the 
height differences. Third column: Frequency distribution of the height differences of all points. Forth column: frequency 
distribution of the height differences without the tree areas. 
  
First results Without trees 
DEM 
  
the height 
of 
differences 
e other one 
ncy values 
n fact trees 
while the 
It can be 
ised by the 
y using the 
1ctrate the 
have been 
tests have 
25,26, 17. 
spectively. 
own in the 
je bias was 
the new 
f Table 5. 
at in steep 
re positive 
ed by the 
constraints 
mooth out 
| under the 
ched linear 
    
  
  
  
  
  
  
(ween two 
ommercial 
s software 
ngle of a 
osest point