rain
on
are
-he
wi
-8))/
"fects
mi-
yias.
ing
lied
ilts
ised
74
All of the photo pairs used were supplied photographically by
an equal scheme of dots. The centrum of the scheme was placed
according to the uniform coordinates with the help of topographic
maps. The inventory area was first divided into squares of 8km
x 8km on topographic maps. Half of the squares were taken into
the sample systematically. Those photos which had the nadirs
fall closest to the centrums of the squares of the sample,
were those supplied by a scheme.
It could be seen that the sample obtained was very close to that
which would have been resulted if it had been defined on the
basis of topographic maps, alone. This is because the differences
in elevation are fairly small in Finland and the photos were
rectified. On more mountaineous regions or with a photography
with larger angle, the distortions would have been, of course,
larger.
Conclusions
The application of the principles of cluster sampling is suggested
when remote imagery is used to inventory natural resources. The
remote images, air photos for example, are regarded as clusters.
In the first stage, the clusters must be taken in statistically
sound way on orthoprojection, with help of topographic maps, for
example. In the second stage, the clusters are supplied with a
scheme which defines the cluster elements directly on the scale
of the image.
It was shown that the vertical photos serve as an unbias basis
for sampling in the conditions of cluster sampling described
above. This conclusion seems to be valid also to tilted air
photos and images based on scanning. The presupposition to be
linked to the conclusion is that the terrains in the part of
image applied should not cover each other in the imaging
projection.
In the remote imagery taken with large angle over a rough terrain
it may easily appear that some low parts of the field are
shadowed by nearby high terrains. This is true especially at
the borders of the image. Because of this handicap it may be
advisable to use only the central parts of images for sampling.