object space, the transformation was carried out iterati- 
vely. 
4.2. Priorat 
4.2.1. Project description. The second set of test data 
was provided by the Institut Cartografic de Catalunya, 
where SPOT panchromatic scenes are being used since 
1988 for the production of orthophoto maps at 1 : 50 000 
scale. For that purpose a stereopair around Priorat, 
South Catalonia was available. The images were recor- 
ded on September 3 and 9, 1988 with 13.5 and 17.3 
degrees sensor inclination respectively and a base-to- 
height ratio of 0.6. The image coordinates of nine conju- 
gate points were measured interactively in both digital 
images. Additionally about 30 GCP are- given, which 
were identified and measured in the images; however, 
the measurements were performed in either the left or 
the right image, i.c. no conjugate GCP are available. 
Image matching was performed in the same way as 
described for the project "Heidelberg". Using the nine 
conjugate points as starting points and a value of 10 for 
STEP more than 250.000 conjugate points were obtai- 
ned. 
For the bundle adjustment about 1300 equally distri- 
buted points were selected such that the correlation 
coefficient of each point is a local maximum. These 
points were processed together with seven GCP in each 
of the two images and the XYZ coordinates of the 
projection centres. The following standard deviations 
were assumed: 
- Sum for the image coordinates of the points, 
- 1mfor the XYZ coordinates of the GCP, 
- 1m and 1000 m for the coordinates of the projection 
centres (for the relative and absolute accuracy re- 
spectively). 
4.2.2. Results. In this project no independent check 
points were available. However, the derived heights of 
the 1300 points were checked against the countrywide 
DTM data base of Catalonia, which is available for the 
complete territory with an accuracy of approximately 
1-2 m. The RMS error of the empirical standard devia- 
tions of the object points results in uz — 8.8 m. 
470 
S. CONCLUSIONS 
It has been demonstrated that automatic photogramme- 
tric processing of SPOT imagery for point determination 
and DTM generation using very few control points is 
feasible. Independent checks revealed an accuracy of 
about 15 m in planimetry and 7 m in height. 
An important issue, which has not been discussed in this 
paper, is the low reliability of point determination using 
SPOT imagery. Due to the local redundancy of 1 for each 
object point (4 observations, 3 unknowns), the ability to 
detect observation errors in the adjustment is restricted. 
A significant improvement in terms of reliability can be 
achieved , if e.g. 3-line scanner systems are used. 
In future an interactive procedure for quality control on 
all levels of processing will be incorporated into the 
system. In this way also image areas showing low texture 
can be reliably matched. For a comfortable measure- 
ment of the image coordinates of the GCP and for 
further processing (e.g. visualization of the results, gene- 
ration of follow-up products) the software should be 
integrated into a digital photogrammetric workstation. 
6. ACKNOWLEDGEMENT 
The reported results of the "Heidelberg" project have 
been obtained within an IABG study 15128, contract no. 
E/F41F/10365. The help of Zhang Shu Ping, Werner 
Mayr, Albert Wiedemann and Tobias Kunst is greatfully 
acknowledged. 
7. REFERENCES 
Dowmanl., 1992: OEEPE Test of triangulation of SPOT 
data, OEEPE publication (in print). 
Ebner H.; Kornus W.; Ohlhof T., 1992: A simulation 
study on point determination for the MOMS-02/D2 
space project using an extended functional model, 
IntArchPhRS (29) 1. 
Ebner H., Reinhardt W., Hófler R., 1988: Generation, 
management, and utilization of high fidelity digital 
terrain models, IntArchPhRS (27) B11, III-556-566. 
Fórstner W., 1982: On the geometric precision of digital 
correlation, IntArchPhRS (24) 3, 176-189. 
Heipke C., 1990: Digitale Satellitenbildauswertung, Stu- 
die der Industrieanlagen-Betriebsgesellschaft,