Introduction.
The aim of this work is to characterize some individuals from
different arboreal species starting from data obtained by a digi
tized false colour image (Fig. 1).
The proposed thesis is that the characterization of an indivi
dual which belongs a certain species does not depend only on cha
racteristics discernable through pseudospectral behaviour but ca
n also take place on the basis of morphological and morphometric
characteristics of both the individual and the scene to which it
belongs.
This work illustrates the first results of an exploratory ana
lysis aimed at defining some features which characterize the ind
ividuals of some species of interest and the line which will be
followed for the use of the morphological and morphometric chara
cteristics of scene.
These characteristics have a precise algorithmic form, and
thus constitute the necessary basis for a possible automatizatio
n of the system.
l. Exploratory analysis.
By exploratory analysis of data we mean a mathematical handli
ng of data with the aim of finding out indication that may serve
as hints in approaching further data.
The indication obtained allows the user (in this case the re
mote sensing user) to decide whether they are significative with
respect to the system observed and the experiment under way or
whether, to the contrary, they are to be judged as pure formal
accidents.
Our analysis aims at gathering indications relative to charac
teristics or to combinations of characteristics of a structural
and morphological nature able to help characterize an individual
as belonging to a particular arboreal species.
1.1 Pseudospectral analysis.
By pseudospectral analysis we mean that sort of analysis whi
ch is carried out with the classic criteria of multispectral ana
lysis, but starting out with data obtained with a spectral conve
rsion.
The typical case is that -adopted here- in which an analysis
is made of the chromatic components of an object as they emerge
from a false colour image, these being handled as though they we
re true and proper spectral bands, despite the fact that a trans
fer function has been introduced into the image recording system.
The type of spectral analysis of the data undertaken involved
the following phases:
a) identifying of the co-ordinates of the areas of precise grou
nd truth (training areas)
b) acquiring three hystograms (blue, green, red) relative to the
training area data