= 491 =
WHAT IS A REMOTELY SENSED TREE?
A PROPOSAL FOR STRUCTURAL AND
PSEUDO-SPECTRAL RECONNAISSANCE.
O
P. Brambilla
G.M. Lechi°°
P. Mussio??
The aim of this work is to characterize the typical chroma-
tic, structural and morphometric performances of the individuals
of particular vegetal species starting from false colour aerial
transparencies.
Neverthless most important aspect of this work is the proposal
of an automatic reconnaissance methodology of the objects. à
For this work it was employed a false colour aerial transparen-
cy 9x9 inches size, 1:10.000 scale: this image was scanned with
an Optronic Photomation 1700 in the blue, green, red band with
a scan spot dimension of 50 micron, corresponding on ground at
50cmx5Ocm pixel size; the three analog electrical signals were
converted into digital form on the basis of 256 levels.
Two steps there are in the employed methodology: first a pseudo-
spectral (rough reconnaissance performed in order to exclude from
the scene most of the undesired objects, then a characterization
of the objects from structural point of view. We described now
the second part, which is performed in three steps:
a) the description of each individual in the scene and some sim-
ple relation among them derived from the Fresta Mussio (FM)
code. The descritpion is performed in terms of elementary to-
kens (a set of primitive figures) and multivalued operators
b) the use of the elementary tokens allows the precise defini-
tion of several convexity and or concavity classes
c) from the previous criteria there was derived some structural
coefficient as: boundary complexity, area/perimeter ratio,
vertices number/perimeter ratio etc.
Such a combined methodology fournishes a very good results as
it is described in the paper.
O no"
Università di Milano - Facoltà di Fisica
00
Istituto per la Geofisica della Litosfera-C.N.R.
Via Mario Bianco 9 - Milano
00 ; : ;
Gruppo Ricerca Scientifica Informatica-Regione Lombardia
Corso Como 19 - Milano