1 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BI. Istanbul 2004 
  
this paper. This is why we will focus on the 2 sites of Bavaria 
and Barcelona for which we have got several results (5 for 
Bavaria and 3 for Barcelona). 
This paper will not describe the production processing of 
each investigator because each of them wrote his own paper 
to be presented during the ISPRS congress in Istanbul (July 
2004). 
The general method for our assessment was the following: 
each DTM is compared to references and also to other 
produced DTM. Where the results seem to be strange or 
unusual, we try to find an explanation in the author’s paper. 
We made our own analysis and statistics to be sure that we 
compare exactly the same area in all DTM. 
The comparison is not straightforward because each 
investigator decided to produce DTM with a specific 
sampling. The sampling of received DTM ranges between 5 
and 50m and it's more of the same for references. So for each 
site, we start by describing the references and then the 
produced DTM. After that we give all the statistical results 
and their analysis, followed by a qualitative evaluation to 
assess the significance of the altimetric restitution of those 
DTM by drawing some profiles. 
In conclusion, we study the results in comparison of the 
results we had during the in-flight commissioning phase of 
the SPOTS satellite. 
In this document the term DTM or DEM will be used 
indifferently, knowing that products computed by space or 
aerial processing are mostly something in between. 
2. TS8: BARCELONA (SPAIN) 
2.1 Input Data 
Location: 0?00'-3?30'E / 40?30'-43?00'N 
Elevation range: 0— 1700 m 
SPOT 5 data are either HRS or HRG images, described in 
Table 1: 
  
  
Lhe HR " banane —A a dans 
Fig2: Barcelona with each DTM areas 
  
Principal Investigator (blue): 
Dr.-Ing. Wolfgang Kornus - Institut Cartografic de Catalunya 
(ICC) 
Col-2 (yellow): 
Dr. Peter Reinartz — DLR —Germany 
Col-4 (red): 
Hannes Raggam - Joanneum Research, Institute of Digital 
Image Processing, Austria 
H. Raggam also produced DTM from HRG/HRS data on 
areas 1, 4, 6. 
   
  
PI (blue) CoL (yellow) Col4 (red) 
W. Kornus P. Reinartz H. Raggam 
  
Areal 961 x 141 1pix 
Spacing: S m 
  
Area2 | 826 x 929 pix. 
Spacing: 10 m 
951 x 141 1pix 
Spacing: 5 m 
  
Aread | 796 x 799 pix. 
Spacing: 10 m 
961 x 141 1pix 
Spacing: Sm 
  
  
  
Area5 | 797 x 909 pix. 
Spacing: 10 m 
  
  
  
Area6 | 722 x 833 pix. 95] x 1411pix 
Spacing: 10 m Spacing: 5m 
  
  
  
  
Global 3975 x 4795 pix | 6001 x 9201pix 
Spacing: 15 m Spacing: 10 m 
  
  
  
  
  
  
  
  
  
  
HRS scene Year | Month | Day 
50462650210151051481S 02 10 15 
504626502101510532028 02 10 15 
504626602101510515618 02 10 15 
50462660210151053282S 02 10 15 
HRG scene Year | Month | Day 
50452650210151052341A 02 10 15 
50452650210151052341B 02 10 15 
50452660210151052421A 02 10 15 
50452660210151052421B 02 10 15 
  
  
  
  
  
Table 1: HRS and HRG data on the Barcelona area 
Reference data: 
Reference data has been provided by the Principal 
Investigator, Dr.-Ing. Wolfgang Kornus of the Institut 
Cartogräfic de Catalunya (ICC) and is composed of : 
DEM "Laser" Size: 4669 x 6601 pixel, spacing: 15 m, 
accuracy: :1,1 m rms 
Digital color orthoimages (1:5000) Nb :32 (4 x 8), Size : 
4800 x 6601, Pixelsize : 0,5m 
2.2 output DTM: 
Table 3 DEM provided by investigators 
2.3 Statistical results 
We started by calculating the differences (meter values) 
between each produced DTM and the reference. All the 
differences have been computed in the same way : 
received DEM — Reference. 
For each DTM, the difference image has exactly the size of 
the produced DTM and the smallest sampling between 
reference and produced DEM. Some DTM have been 
provided with altitudes in decimeter, for those ones the 
difference has been calculated in meter as well. 
Before calculating statistics we took off (mask) all the invalid 
values in each DTM (external values ...). 
The calculated figures for statistics are the following ones : 
negative and positive maximal values, mean and standard 
deviation for 100%, 98%, 90% and 68% of “valid dots”. 
As it’s not feasible to put in this paper all the statistics we’ve 
computed, we just indicate the 98% values because most of 
the time we've got some non significant edge effects with 
     
Intei 
— 
sign 
the 1 
All: 
  
All 4 
Stant 
The 
refer 
that 
grou 
We 
DEN 
Diff 
all r 
rece 
oppc 
that 
chan 
beca