not been 
:Juded in 
. statistic 
ic DEM 
difficult 
can be 
natically 
hreshold 
and with 
at, but is 
'actically 
esent the 
ng in 
yrted that 
ating 3D 
ble for 
igly high 
ration is 
ventional 
s almost 
urements 
s can be 
considered representative remains to be seen. There are 
certainly more investigations to come in which the 
performance of the MATCH-T system with rugged terrain, and 
in relation with break lines, will be studied. 
3. Time Performance 
The test models implied the execution of the complete digital 
processing related to the automatic DEM generation, starting 
from reading the digital image data. The models were 
processed in batch mode on the Silicon Graphics 4D35 
workstation which is a 33 MIPS, 6 MFLOPS machine. The 
time performance is quoted here with regard to this machine, 
because all models were run on it, although the program is 
operating on other workstations as well. The quoted 
computation times include all digital operations, except the 
absolute orientation and the measurement of border lines which 
were taken over from the analytical plotter. 
Project # size : 15 size : 30 remark 
3h 1.6h 1991 
“2,1h 1) 1) 1991 
6h 1.9h 1991 
    
3.2h 1.1h 2) 1992 
3.2h 1.1h 2) 1992 
D part of the model estimated 
Table 2: Total computation times for automatic DEM 
generation (Silicon Graphics 4D35) 
The table shows that the program version of 1991 took roughly 
six hours of computing time per model, with 15 pm pixel size. 
With 30 pm pixel size the computing time was reduced to less 
989 
than two hours. In the meantime the computing performance 
has been increased by a factor 2. At present one model takes 
about three hours at 15 pm pixel resolution, and about one hour 
at 30 pm pixel resolution. The further break-down of 
computing time gives the interesting result, for the time being, 
that the DEM generation module takes up 43 % of the total 
computing time. 67 % are used by the pre-processing 
modules, concerning the image normalization, the image 
pyramids, and the extraction of the interest points. The total 
system has certainly reached a high level of economic 
processing performance. 
4 References 
Ackermann, F., Krzystek, P. (1991): MATCH-T: Automatic 
Mensuration of Digital Elevation Models. Proceedings of 
Technical Seminar of the Sociedad Espanola de 
Cartografia Fotogrametria y Teledeteccion. Barcelona, 
12th April, pp. 67 - 73. 
Hahn, M. , Forstner, W. (1988): The Applicability of a Feature 
Based Matching and a’ Least Squares Matching Algorithm 
for DTM Acquisition. International Archives of 
Photogrammetry and Remote Sensing, Kyoto , Vol. 27, 
Part B9, pp. III 137-150. 
Hahn, M. (1989): Automatic Measurement of Digital Terrain 
Models by Means of Image Matching Techniques. 
Proceedings 42nd Photogrammetric Week, Stuttgart , pp. 
141-151. 
Krzystek, P. (1991): Fully automatic measurement of digital 
elevation models. Proceedings of the 43rd 
Photogrammetric Week, Stuttgart, pp. 203 - 214. 
Krzystek, P., Wild, D. (1992): Experimental accuracy 
analysis of automatically measured digital terrain models. 
Robust Computer Vision: quality of vision algorithms. 
Forstner, Ruwiedel (ed.). Wichman Verlag. Karlsruhe, 
1992, pp. 372 - 390.