ared to a
land use
t. Forest,
al signa-
ne class.
> "sealed
nd "resi-
signation
an enor-
it comes
! unsharp
not exist
e "build-
only the
parcel is
o matter
t.
forms of
tribution
into this
1 inexact
"polluted
pulated"
‚here are
ature do
and are
e of ele-
vhen the
as many
| param-
the cor-
TE
t limited
erlapped
graphic
Analyti-
although
n photo-
leverthe-
without
t surpris-
el reality
nd lines"
are arti-
iseful for
vorld.
lling still
acteristic
Sealed
Sürlace
Road
Parking
Site
Square
Indust.
Area
Resid.
Area
Water
Park
Swamp
Grassland
Forest
Forest
Grassland
Swamp
Park
X| X| X| X
Water
X|X|X
Resid.
Area
Indust.
Area.
XX
Square
XX|X
Parking
Site
X| X| X| X
Road
XX XXX
XX XXXX
Sealed
Surface
Figure 3: Confusion matrix for land use semantics
of the analogue and analytical world (Kuhn, 1993) are the
concepts " generalisation, resolution and scale". The concept
"scale" loses meaning when we describe the real world by
means of analytical methods, i.e. with coordinates. Coordi-
nates conform to reality by principle in scale 1:1. The concept
of "scale" as we use it in every day science, is therefore always
related to the visualisation of the real world.
This does not apply to generalisation, at least not to an im-
portant part of it. The so-called "generalisation during data
capture" is basically scale-independent. Scale onlys acquire
importance when considering a posterior visualisation of the
captured information. The concept "generalisation" is very
helpful for semantic modelling, which is not the case of the
concept "scale". It was shown in Figure 1 that transforma-
tions of the real world over the iconic to the symbolic level
always impliy generalisation of the " generalisation during data
capture” type.
"Resolution", finally, is a concept that has many aspects.
Originally it described the quality of sensors and their com-
ponents. In this sense, the modulation transfer function is a
measure of the resolution of sensors which describes mathe-
matically in a very elegant way the effects of each component.
For semantic modelling this is of little use. The concept of
resolution in this context must be understood as the capacity
to recognize and label an object. If this is to be accomplished
today with modern tools and the assistance of a computer,
it should be stated here that even with analogue processes,
resolution was defined in relationship with " recognizability of
objects".
The discussion over the terms "point", "line", "generalisa-
tion", "resolution" and "scale" shows that they stand for
images that have each of them be developed in an individual
technological context and that they are only wholly applicable
there. In a different technical context the direct transference
can only lead to confusion which is basicalle of linguistic na-
ture in a similar way as unprecise terms.
11
8 CONCLUSION: OLD PATTERNS OF THINKING
BLOCK SOLUTIONS
New materials and new tools, consistently used, must lead to
new products. This requires a wide theoretical analysis, even
new patterns of thinking.
In oposition to the coordinates of the analytical photogram-
metry, the new material, "digital images" carries not only
geometrical but also semantic information. The value of the
new products is characterised by attribute data, less by ge-
ometrical data. The methods of adequately describing such
data are today in full evolution. The human operator shall
in the future be substituted in the highest possible degree by
computer assisted procedures
It is our main aim to show that new patterns of thinking
are required or, in other words: without the disposition to
radically question apparently logical model assumptions (in
the real world there do not exist such things as "points" !)
false assumptions are created, blocking scientific progesss. It
should then not be surprising when complicated theories and
unsatisfactory results are obtained.
An example shall suffice: The (false) assumption that the
Earth is the centre of the Universe, lead to complicated theo-
ries to explain the plante epicycle, precisely an unsatisfactory
result.
That is why image semantics cannot be treated, for example,
with the methods of Analytical Geometry. It is apparently
feasible, but it complicates the theory and unsatisfactory re-
sults are obtained.
REFERENCES
Ackermann, Friedrich, 1995. Digitale Photogrammetrie - ein
Paradigmasprung. Zeitschrift für Photogrammetrie und
Fernerkundung, pp. 106-115.
Bahr, Hans-Peter, Franz Quint and Uwe Stilla, 1995. Mo-
dellbasierte Verfahren der Luftbildanalyse zur Karten-
fortführung. Zeitschrift für Photogrammetrie und Fern-
erkundung. pp. 224-234.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B6. Vienna 1996