>. Istanbul 2004 
imi 
DIM. 
of curvature 
DATA 
ly measures as 
' laserscanning 
irements — but 
much is left to 
terrain points. 
analysis of the 
e. 
ALS project in 
od risk areas of 
s about 1000m. 
ition. 
  
C DATA 
urements have 
ith SCOP. The 
image scale of 
ius, 2004) this 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
leads to an accuracy of €17cm for the measurement of terrain 
points. 
The photogrammetric DTM was analyzed in the same way as 
the ALS-DTM. The point density is 0.0016 points/m? in the 
areas where primarily a raster of points has been measured. 
Additionally, breaklines were digitized manually, cf. Figure 10. 
The maximum RMS (cf. section 2.2) of the original elevation 
values is 70cm; the quadratic mean of the RMS values equals 
15cm. The curvature values l/r at the grid points of the DTM 
are roughly 20% lower than the curvature values of the ALS- 
DTM (reason: the lower point density of photogrammetry leads 
to a smoother DTM compared to laserscanning). The accuracy 
measure O,prm IS shown in Figure 10. The values range from 
2.1cm to 1.9m; their median is 14cm. These values are 
significantly poorer than the values derived for the ALS-DTM 
in Section 2.4. 
  
  
0 005 01 016 -025 [m 
Figure 10: Accuracy o, pr of a photogrammetric DTM 
including digitized breaklines, white: areas without original 
data 
In the case, that the DTM is determined without the manually 
digitized breaklines and the accuracy is derived, the results 
shown in Figure 11 are obtained. The deterioration along the 
breaklines can be recognized clearly. The values vary between 
2.5cm and 4.5m; their median is 18cm. Not only the accuracy 
measure o; pru gets worse, but already the RMS worsens from 
70cm to 1.7m (maximum value), and from 15cm to 21cm 
(quadratic mean), respectively. The curvature values for 
deriving the areas A were taken from the DTM computed with 
the breaklines. 
H7 
  
HEE EN 
0 005 01 015 2025 [mi 
Figure 11: Accuracy o, pr of a photogrammetric DTM without 
breaklines, white: areas without original data 
6. CONCLUDING REMARKS AND OUTLOOK 
The aim of this publication was to present a method for the 
derivation of quality measures — especially accuracy measures - 
for DTMs. Applying the developed tools, the qualities of DTMs 
acquired by different methods can be specified and compared. It 
has to be noted critically, that the obtained quality measures 
specify in the first instance relative DTM accuracy. 
Furthermore, detail information not contained in the DTM used 
for deriving the curvature, does not show up in the presented 
accuracy measures. 
The comparison of ALS-DTMs with photogrammetric DTMs 
started in Section 5 will be developed further in the coming 
months and presented in a separate publication (Kraus et al., 
2004). 
The intention is to provide in future in addition to the DTM 
data used so far, also the presented quality measures. 
Concerning this, an implementation in the software SCOP is 
planned. 
The following questions have to be answered based on 
empirical tests: 
e Appropriate size for the analyzing unit. It has to be 
estimated from the density of the original set of data 
(Section 2.1). 
e Estimation. of o,,,,! for different methods of data 
capturing (Section 2.2). 
e Appropriate choice of the accuracy of approximation, 
which influences the areas A strongly (Section 2.4). 
e Using the geometrically based approach, the areas 
exceeding the curvature radius were classified unusable 
(white areas in Figure 8). Presumably, the value for the 
critical distance d, has to be set to a value significantly 
smaller than the radius r of the curvature, e.g. 0.5r. 
The results obtained so far already demonstrate the capabilities 
of the method for the determination of quality measures for 
digital terrain models.