SESSION 8:
AREA BASED MATCHING TECHNIQUES
CHAIRMAN: W. Fórstner (FRG)
AUTHORS: A. Grün and E.P. Baltsavias (Switzerland)
TITLE: HIGH PRECISION IMAGE MATCHING FOR DIGITAL TERRAIN
MODEL GENERATION
DISCUSSION:
Ehlers (USA):
Baltsavias:
I have two questions. As I understood, this is an
interactive process. You select certain areas
around points of interest in one image and try to
match them with the same area in the other image.
Have you done anything to find those "points of
interest" by automatic techniques? And the second
question is: You used interpolation techniques to
create a regular grid based digital terrain model?
And what are your further plans for DTM generation?
I have already mentioned that this was a project in
cooperation with the firm Wild. So, we had some
specifications to follow. The specifications were
that we should determinate the height at given X
and Y ground positions. So, to answer your second
question, in this case you don't have to interp-
Olate. After you determinate the heights, you have
a regular grid on the ground. Now, of course, what
you mentioned is correct. I mean that it is an open
subject, whether we should follow this approach or
follow another approach which chooses the best
points for height determination and then using
these, which would not probably be a regular grid,
interpolate to a regular grid. That is correct, I
agree with you. As far as your first question is
concerned the points we chose were not really points
of interest, in the sense that they had good geo-
metric and radiometric properties for precise height
determination. We just chose some points, ranging
from good to bad, so that we could measure them at
the AC1, and were able to compare the heights from
the manual measurements to the heights from the
matching algorithm. The choice of points of inter-
est was out of question for this project, since
we were constrained to determine the heights at
given X,Y positions.
Grün (Switzerland): I just would like to add to your first ques-
tion. It was not really an interactive process.
That was exactly the reason why we could not really
74