plour
anal
ith
dct.
2 ac
luen
T.
ico
78
a e
agini
il
gra
ar
975
stem
elp
EXTRACTION OF TEXTURAL SIGNATURES FROM
FOREST AERIAL PHOTOGRAPHS USING A SPECIAL-PURPOSE
DIGITAL COMPUTER
W. Schneider
Institut für Vermessungswesen und Fernerkundung
Universität für Bodenkultur
Peter Jordan Straße 82, A-1190 Wien, Austria
Abstract
The reliable discrimination and identification of forest stands from
aerial photographs by automatic methods requires a classification
process based on both textural and spectral parameters. These
parameters, besides from being relevant for the classification process,
should satisfy certain requirements: a) The extraction of the para-
meters from photographs should be possible at a reasonable
b) The parameters should show little variation with
expenditure.
or they should at least be related to the angle
the angle of view,
of view in a simple manner.
In this contribution, a method for the extraction of textural signatures
satisfying the above-mentioned requirements is investigated. The
method is based on the use of the Quantimet 720 image analyzer.
With the aid of this small special-purpose digital computer, the input
image is decomposed into a series of binary images by hardclipping
at varying hardclipping thresholds. In these binary images, number,
area and perimeter of isolated objects are determined. These para-
seen as a function of the hardclipping threshold, represent
meters,
se relationship to properties of the
textural signatures showing a clo
individual tree.
The usefulness of these textural signatures for automatic tree species
recognition from aerial photographs is discussed on the basis of
experimental data.