ul 2004 
a thorough 
needed as a 
really a tool 
r hand it is 
ived DTMs) 
ie usefulness 
sults on the 
)S should be 
ies obtained 
be used as a 
he cm-range. 
| pursued by 
e from the 
lon purposes 
) types yield 
'ED level 2 
in (H:60- 
DTED 
1v2 
-0.32 
4.28 
2.07 
heck points 
(H = terrain 
a step that is 
al bottleneck 
se for which 
ry automated 
canning, the 
tion is still 
> DTM as a 
spot possible 
visualization 
t techniques, 
best quality 
jective, quite 
1 of figure 1 
ives a good 
1e model and 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
  
Figure 1. extract of DTED level2 as a shaded image 
  
Figure 2. extract of DTM10.000type4 (based on digitized contour lines) 
as a shaded image 
During the processing of the data, regular checks of consistency 
within the data itself or consistency with other data are needed 
to spot problems as early as possible. (e.g. contour line 
topological consistency, statistics on point clouds to eliminate 
outliers or spot systematic errors, ...) Repeated checks and 
cross-checks are reliable ingredients for a good quality 
assurance policy. 
2.8.2. Heterogeneity. Heterogeneity issues can be divided into 
two levels. Within one DEDS of the same type there may be 
quite some variation of quality. E.g. when data for a DTM are 
measured using photogrammetry, wooded areas will yield 
poorer results if any. The quality of manual measurements 
inherently varies from operator to operator, not only because of 
each individual appreciation of stereoscopic depth but also 
because of each individual interpretation. of the elements 
making up the geomorphology. This kind of heterogeneity is 
inevitable but should be carefully considered when performing a 
quality analysis. 
Between DEDSs of a different type there can be differences in 
geometrical accuracy and in data content. Stereoplotted contour 
lines will for example reflect a slightly generalized, interpreted 
“natural” landscape, «excluding big artificial structures. 
Automatically measured DEDSs on the other hand will show no 
bias towards a “natural” or an “artificial” landscape. Reckoning 
with the differences in geometrical accuracy between DEDSs is 
relatively easy using classical edge-matching procedures. 
Taking into account the differences in data-content is however 
not possible. Consequently, an adjusting of the data content of 
the different DEDSs needs to be pursued as far as practically 
feasible. 
Un 
UA 
3. THE COMBINATION OF DIFFERENT DATA SETS 
3.1. Overview 
When different DEDSs of the same area with complementary 
properties are available, it is possible to exploit this in order to 
spot problems and sometimes to remedy them. Studies have 
been made on the possibilities of combining a laser derived 
DSM with edgelines derived from image matching (McIntosh 
and Krupnik, 2002) and combining a laser derived DSM with 
spectral information from color aerial photographs (Haala, 
1999). Laser derived DEDSs are in practice supplemented with 
photogrammetrically captured structure lines and field survey 
results (Reiss, 2002). Linear and polygonal elements are added 
to correct the height of points derived from image correlation 
(Dupéret,1999) and DTMs based on digitized contour lines are 
used for the early detection of gross errors in laser data (Artuso 
et al., 2003). 
3.2. Test: goal and used material 
We performed a test in two small areas in the Belgian Ardennes. 
The goal of the test was to investigate the possibilities to exploit 
the complementarity of different DEDS, not only in order to 
spot problems and guide the interactive work but also to 
introduce a degree of automation in the combination of the 
DEDSs. We chose to use DEDSs that were readily available for 
a large part of the country (digitized contour lines and 3D CAD 
files containing elements for the topographical map on 1:10 000 
) or that could be easily and quickly produced (DEM from 
image correlation). We also chose to use softwares which we 
already use. We have established a thorough experience with 
MATCH-AT (INPHO), ISSD (Z/I) * MicroStation (Bentley), 
automatic DEM derivation with VirtuoZo (Supresoft), TIN 
functionalities for DEM production with Terrain Analyst 
(Intergraph), DSM filtering and general DEM management and 
editing with SCOP and GVE (INPHO). For the test it was 
necessary to combine these softwares in one production chain. 
Test area 1 includes a moderately wide valley flanked by 
relatively steep, wooded slopes. Area 2 has moderate relief in a 
more or less open landscape. In both areas the first DEDS 
consists of stereoplotted contour lines with a Sm interval based 
on aerial b/w photographs on a scale of 1:20.500. The overall 
quality of these contour lines is good but in the valley they are 
frequently missing. Along the river valley we find a number of 
quarries where contour lines are also missing. A second DEDS 
consists of a DSM generated automatically on a different set of 
aerial b/w photographs on a scale 1:20.500, with a posting of 
20m. During the calculation of the DSM a partial filtering is 
performed by the software, eliminating to a large degree the 
small areas of higher elevation (trees, individual buildings) but 
not filtering larger areas of higher elevation (large structures, 
forest canopy). 
3.3. Combining two data sets 
The DSM is quite usable as a DTM in open areas but gives no 
indication of terrain height in wooded areas and it has the 
general disadvantages of an automatically generated DSM for 
which there are no remedies except interactive editing 
(Ackermann, 1996). On the other hand, the contour lines are