can be instructed to perform them. These tasks belong primarily in the
realms of analytical photogrammetry and cartography. At the other
extreme, we may find tasks easy when performed by people, but at the
same time experience difficulties in developing computer solutions. This
is the case when not a detailed enough theory is available that allows
deriving the necessary computer instructions. Rather than trying to find a
solution by trial and error, it is more constructive to gain enough insight
into the problem, such that the proper methods (tools) for its solution can
be chosen. Problems people solve with reasoning, such as the example
mentioned with the building in the middle of street intersection, probably
defy an algorithmic approach. Artificial intelligence, together with other
disciplines, may help us to understand a problem and to formulate a theory.
By the same token, artificial intelligence may also offer tools to
efficiently solve the problem once it is well enough understood.
We are only at the beginning of tackling the more difficult tasks. We need
to pay attention to these tasks even though the photogrammetric and
cartographic research community may feel that some the problems appear
to be too simple. The results of this type of investigation are needed as
part of the foundation for concept and theory development.
References
Marr, D., 1982. Vision.
W.H. Freeman and Company.
Marr, D., T. Poggio, 1979. A computational theory of human stereo vision.
Proc. R. Soc. Lond. B. 204, pp. 301 - 328.
Mayhew, J.W.W., J.P. Frisby, 1981. Psychophysical and Computational
Studies towards a Theory of Human Stereopsis.
Artificial Intell. 17, 1981, pp. 349 - 385.
Rich, E.,1983. Artificial Intelligence,
New York: McGraw Hill.
Schenk, A, 1986. Stereo matching using line segments of zero crossings.
Proc. ISP Symp. Comm. III, Helsinki.
476