Introduction
Aerial photography is a widely used technique for inventories and
observations of forests (Hildebrandt, 1977). The aerial photographs
are usually evaluated and interpreted visually by specially trained,
experienced photointerpreters.
Many attempts have been reported to support or even to substitute
the manual, subjective methods by automatic, computerized, objective
interpretation techniques (Kalensky and Wightman, 1976). For a
reliable identification and classification of forest stands, the
classification process has to be based on both spectral and textural
parameters. These parameters, besides from being relevant for
the classification process, should satisfy certain requirements:
a) The extraction of the parameters from photographs should be
possible at a reasonable expenditure. b) The parameters should
show little variation with the angle of view, or they should at least
be related to the angle of view in a simple manner.
In this contribution, the use of the image analyzer Quantimet 720
of IMANCO for the purpose of extracting textural signatures from
forest aerial photographs is discussed. It is shown that some of the
textural parameters that can be obtained with the Quantimet satisfy
the above-mentioned requirements. The usefulness of these para-
meters for automatic tree species recognition is demonstrated.
Test data
The photographs used for this investigation were taken in 1975 in
an area approximately 60 km south of Vienna. This forest area,
which is very intensively managed, is characterized by an inhomo-
geneous distribution of species. It also has rather complex relief
features with an average gradient of slope of 15° to 20°. It was,
nevertheless, selected for this study for two reasons: This forest
area is used by the forestry institutes of the Universität für Boden-
kultur for instructional purposes. Many ground investigations are
being performed there continuously, so that, from the point of view
of available ground data, it is an ideal test site. The second reason
for choosing this test area was that, in addition to aerial photo-
graphs, multispectral scanner data had been acquired in this area.
Comparative studies of the performance of these two remote sensing
systems can therefore be performed.
The photographs used for this study were taken on Aerochrome
Infrared film 2443 with a Wild RC-10 camera equipped with an
Aviotar f-30 cm lens, from a height of approximately