3. The number of linear features ;
4. The tonality of the linear features.
The headlines of the study were carried out as follows : photo-
graphic prints were made on different scales : 1/250.000, 1/100.000 and
1/50.000 in order to find a suitable scale to draw up an inventory. The inven-
tory capacity for scales between 1/250.000 and 1/50.000 proved, by experiment
to be the same. As for larger scales, lines were detected purely caused by
speckle effects. Further enlargement causes a break-up of the image.
As 1/250.000 and 1/100.000 seemed rather unpractical to draw the lines on an
overlay, photographic prints at a scale of 1/50.000 were used to draw up the
inventory.
The image was "line-scanned" visually along azimuthal, range and
diagonal directions. The human eye doesn't detect pixelwise but rather has a
simultaneous overview on a part of the photograph or the whole photograph.
Confirmation of this statement is found in the "Gestalt"-psychology. So the
visual detection of linear features on photographic prints is mainly based on
the contrast of the pixels on line with their neighbouring pixels. Consequent-
ly a logical distinction was made between pale lines neighboured by darker
pixels and dark lines neighboured by pale pixels. The use of a more detailed
division was out of request, due to the lack of continuity in grey value
of neighbouring pixels, caused by speckle.
The overlays on which all linear features detectable, show that
most of the lines are not longer than 2 cm. This means that only important
features as highways and shipcanals are detectable as continuous lines.
Comparison was made with the groundtruth, based on terrain observation,
maps and air photographs. It is impossible to reconstruct the existing network
of all kinds of roads except highways, without further additional information.
This is shown by comparison of overlays with image information with overlays
based on map information concerning roads canals and motorwaycomplexes.
Distinction among the motorwaycomplexes is based on :
1. the importance and width: of each way, according to the
map legend.
2. accompanying features as tree-rows and ditches. This infor-
mation is found on airphotographs from approximately the same
date as the SEASAT imagery.
In order to quantify the interpretability of the linear features
according to their nature and their relative orientation, several tests have
been made in which variables were compared two by two.
Comparison was made as follows : we draw up an inventory of
white lines on a transparent overlay on the image. Independent of the first
overlay we draw up a second one with the black lines. These two overlays
are superposed and adjacent black and white lines are counted.
The intention was to know whether linear features are neigh-
boured by other contrasting linear features or whether they are neighboured
by contrasting patches. We also wanted to know if the way of detection by
human eyes does work bi-directional or not, fore as this bi-directional
detection might show an example of optical illusion. It was found that not
five percent of the lines (as well in number as length) were overlapping
on another. The coincidence of white and black lines are mostly caused by
motorwaycomplexes with accompanying trees and/or canals. On the other hand,if
we lookat the map information ‘first the greater part of motorway complexes were
250
not de:
their 1
genera.
roughnge
that it
furthe:
image,
Or rows
ration
the fie
on the
on the
of line
because
axis. Z
result
quantit
lation
This di
both ty
in real
directi
not mor
image c