WAN WW W VI J af) WWW R(T WW CT J uu WW 
^ Re WW WW Ve TN VE AU X 5 Vi ps 
AUTHOR: 
TITLE: 
DISCUSSION: 
Ebner: 
Rauhala (USA): 
  
G. Strunz (FRG) 
PROCESSING OF NON-PHOTOGRAMMETRIC DATA IN COMBINED 
BLOCK ADJUSTMENT 
Thank you, Mr. Strunz, for your general remarks on 
combined point determination and summary of compu- 
tational aspects and problems which still have to 
be solved in this field. May I ask for questions, 
please. 
I would like to come back to one point mentioned 
by Mr. Strunz: the consideration of.a DTM in 
conjunction with point determination. This is 
a rather new possibility which we recently have 
investigated in Munich. That means you consider 
an existing digital terrain model, for instance, 
in raster form which basically gives control 
information not only in height but in case of 
hilly terrain also in x and y. If the DTM is good 
enough you have the special case that all your 
points existing in. the total block can be used as 
control points. That means that you can expect 
rather good accuracy. This accuracy is particu- 
larly good so far as height is concerned but 
even the accuracy in x and y is not bad. We have 
done some simulations first and then a practical 
test and have found that at least for ordinary 
mapping purposes, in case of hilly terrain you 
really can work without control points, just using 
a DTM. This possibility is interesting from another 
point of view, block adjustment can be interpreted 
as a matching procedure which is very similar to 
least squares matching techniques which are used 
in the mean time in digital image correlation. I 
wouldn't say you should use DTM as the only one 
control information but at least as an additional 
one. This might become more and more important, 
because DTMs will be available to a high extent 
in future. 
I think the use of DTM is an excellent idea and one 
can also expand the concept by introduction of every 
single pixel as your transfer point. That is 
specially effective if you have multi-ray stereo 
models and this means that you actually are trans- 
ferring all DTM-points between the images and 
throwing the resulting least square correlation 
results into a simultaneous block adjustment. You 
actually introduce an independent unknown for every 
single pixel. But of course the resulting system 
is very large and you need efficient computation 
techniques for solving them. That was the topics 
for my paper. 
47