A METHODOLOGY FOR A LARGE SCALE OLIVE TREES INVENTORY
G. FRAYSSE, P. MONTELLANICO
European Communities
JRC, ISPRA, Italy
Abstract : In order to establish a methodology for the
determination of olive trees area and number in cadaster
parcels, several methods were compared: black-and-white
aerophotography, color infra-red aerophotography, multispectral scanner imagery. Human photointerpretation,
automatic processing were applied to the main categories
of olive trees fields. The results of these experiments
are presented and the sources of errors are analyzed.
1. INTRODUCTION
In 1975, the Council of Ministers of the European
Communities adopted a regulation on the establishment of
a register of olive cultivation in the Community.
In accordance with an article of the Council Regulation,
the register must furnish among other information for
each olive cultivation holding in the Community the
total olive growing area together with the cadastral
reference numbers of the parcels comprising it and the
total number of olive trees. Because the area to be
inspected is large ( ~--- 80.000 km 2 ), conventional ground
surveys cannot be considered. The Services of the
Commission decided to test the more advanced and rapid
methods. Under the responsibility of the DirectorateGeneral of Agriculture, a pilot experiment was organized
in 1976 in Southern Italy. The experiment was coordinated
by the J.R.C. and several national Institutes were in
charge of the different methods to be tested.
485
The results were submitted by the Directorate-General
of Aqriculture to a group of independent experts and
a common methodology was derived.
The Commission adopted this methodology (ref. 1).
2. METHODS TESTED
2.1. General Criteria:
Olive cultivation can take ve~y different aspects:
- density of the trees
- type of distribution (square, delta, random,
lines and hedges)
- hight of the trees
- diameter and shape of the crown (due to age or
cutting mode)
- density of the leaves
- association with other trees (almond, orange, etc)
or with meadow or with crops (vegetables) or with
natural vegetation
- nature of the soil
- presence of irrigation (mainly for production of
table olives)
- lie of the land (flat, slope, terrace)
- state of cultivation (neglected or well maintained)
Due to the large variability of the aspects of olive trees
cultivation, it was necessary to organize a complex experiment with several methods, several scientists and several
test-sites in order to obtain significant results and
conclusions.
2.2. Test-sites:
Three test-sites were chosen in one of the most
productive area of Southern Italy (Apulia) near
Bitonto, Fasano and Palagiano where most of the
aspects indicated above could be met.
486
2.3. Reference method and ground-truth survey:
In order to compare the results obtained by each
method to a reference, 52 p.arcels of about 1 hA
were observed by a well-trained team of technicians.
Cadastral maps showing the location of 12.197 trees
(from this total 7.027 olive trees) were established
using reference grids.
The time necessary to make this survey was measured,
in order to determine the comparative cost of a
ground inventory.
A part of the results was communicated to the scientists
testing the various methods in order to provide them
with interpretation keys (ground truth), the other
part being used as a check of the quality of their
determination.
2.4. Flight survey:
A flight covering 2 strips of about 70 x 2 Km, N.W- s.E
and N.E-S.W, was done in July 1976 by a company with
an aircraft equipped with a metric camera and an
11 channel BENDIX Mul t~ectral Scanner deli verinq
digital data.
The scale of the aerophotographies in black and
white and in color-infra-red was 1/10.000 and 1/18.000.
at the corresponding flight altitudes of 1500 and
2500 m, the.instantaneous field-of-view of the M.s.s.
was 3,75 m. and 6,5 m.
2.5. Photointerpretation methods:
2 Groups operated in parallel using conventional
(human) photointerpretation and automatic photointerpretation (Fig. 1).
487
The criteria which were then used to define the
interpretation keys were mainly tonality, morphology,
structure and texture.
Automatic photointerpretation started from the
decomposition of the false colour image into 3
(blue, green and red) components (Fig. 2) and its
digi taliza.tion.
In addition, an experiment of photointepretation
using directly the restitutor was done, allowing to
produce a cadastral map with the location of trees
in each parcel.
Photointerpretation experiments were also done
using the video diplay of an interacting system
visualizing the MSS tapes (3 best channels or 3 first
principal components).
2.6. Diqital methods:
Experiments were made on ratioing (for example
Channel 9 - Channel 5
. .
Channel 9 +Channel 5 ), on level Sl1C+ng, on
clustering and classification algorithms.
With a resolution at ground of only 3,75 m, experiments were only partially successful.
An interesting method for the counting of the trees
started from a digitized aerophotography. Three
operators, the Laplacian (operator which measures
the individuality of a pixel by contrast with
neighbouring pixels), the local minimum (operator
characterising all the points having a value lower
than that of the eight neighbouring points) and the
gradient (operator giving the direction towards which
the values of the neighbouring points increase) were
combined in order to produce a detection method by
local texture (Fig. 3).
488
3. RESULTS:
Detailed tables were produced for each method and each
category of olive cultivation showing the accuracy and
the time necessary for the determination (cost E=lement):
3 geographic situations multiplied by 11 types of olive
trees gave 33 conditions.
The following conclusions were common to all experts:
- Scale between 1/8000 and 1/10000
Orthophotography is indispensable even in flat zones
- B & w aerophotography is indispensable to complement
coulour-infra-red
- All methods produce good results when the conditions
are favourable (flat ground, evenly sp~ced big olive
trees) and give accuracies of the order of a few percent.
Where conditions are very difficult it is necessary to
use several supplementary methods (slope, terraces, high
tree density, association, etc.)
- The M.S.S. was not considered as a competitor in its actual
performance
Any method has to be followed by a final human control
(staff member highly trained)
- It was not possible to determine if there was an
advantage to fly in July (full vegetation) or in October
(advantage~ olive trees have permanent leaves and some
associated trees loose their leaves in autumn, but
disadvantage; larger shadows).
The proposed method is represented in fig. 4.
4. ACKNOWLEDGEMENTS :
The documents listed in the references 2 to 9 have been
used to write this paper; we express our gratitude to all
scientists who have made. this study possible, and to our
colleagues of the Directorate-General of Agriculture of
the Commission of European Communities, in Brussels,
Dr. Schiratti, Dr. Lucheti, Dr. Cassotta for their support
and understanding.
489
REFERENCES
1.
commission Regulation (EEC) No 2276/79 of 16 October 1979
L:,ving down detailed rules for the drawing-up of a register
of olive cultivation in the Member States producinq olive
oil.
Official Journal of the European Communities 1262, 18 Oct. 197Q.
2. Detection et Comptage des oliviers par teledection numerique.
CEGERNA (Centre d'etudes pour la Gestion des Ressources
Naturelles) Paris.
J.M. MONGET, M. POISSON, G. VILA
3. Application of Analog Processing Techniques and Visual
Interpretntion on Photographic pictures.
- Application of Digital Processing Techniques on Scanner
Images and Photographic Pictures,
Z.G.F. (Zentralstelle fuer Geo-Photogrammetrie und
Fernerkundung (DFG.)
Prof. Dr. J. BODECHTEL
Dr. R. HAYDN
Dipl. Geol. JASKOLLA
Dipl. Aer. FERNANDEZ
Dipl. Math. SEIDERER
4. Interpretczione di fotogrammi pancromatici Be N, verticali.
CNR. Laboratorio di Geologia Applicata alla pianificazione
viaria e Qll'uso del sotto-suolo. Padova.
Dott. Bruno MARCOLONGO
5. Analisi Aerofotogeologica di particelle coltivate
prevalentemente ad ulivi su .fotogrammi pancromatici
Servizio Geologico d'Italia - Roma
Dott. Ing. Gian Lupo del BONO
Prof. C. AVENA
Dott. S. BARLETTA.
490
R eN.
6. Rapporto sulla possib~ilita di uso degli epidioscopi
stereosrrmici nel ?rogramma per il Catasto olivicolo.
Dott. In0. Gian Lupo Del BONO
7. f<apporto sull 'analisi delle immagini MSS a falsi colori
~l fine dPl C~tasto Olivicolo
Dott. In q. Gir:m Lupo Del BONO
g. 1\na lisi dei fotograJ'ltmi i11 falso colore nella zona
Metaponto-Fas~no.
CNR. Tsti.tuto per la Goefisica della litosfera. Milano
Dr. I~0· G.M. L8CHI
9. R:Lcerc(J Sperimenta1e per Ja formazione di un Catasto
CHi vi colo. OperCJzioni 1'opografiche e fotogrammetriche
A. DeterminCJzione dei punti di appoggio Sll.l terreno per
l 1 r1erotrianqolazione nelle zone A.B. e c. Grafici di
copertu:rn.
Direzione Generale del Catasto et dei Servizi Tecnici
Er<triali. Roma.
Dr. Inq. VITELLI
Dr. Inry. MARAFFI
B. Triangolazione Aerea e sua compensazione.
Istituto di Topografia, Fotogrammetria e Geofisica del
Politecnico di Milano.
c.
Prof. L. SOLAINI.
Produzione delle ortofoto e riporto dei limiti catastali.
I.G.M. (Istituto Geografico Militare). Firenze.
Colonnello M. CARLA'.
491
DATA
.....1--
fALSE COLoUR(i.R)
BLAC.I<
WHiTE
MAGNETJ·c TA PES(Mss) I--
e
r- GROUND
~
ClASSiCJ\t.
~ ?lio1oiNTERPR£TATio~
1-1--
A1
k-
fHoToiNTJORl'RETAiiON
CLASSiCAL
TRUTH
CARTOGRA?HY
I--
A2
~
~
,.
.
.
.
..
AUTOMATIC
:DiG-iTAL
CLASSiFiCATiON
'
AUToHAT(C
~
.
fl.lcTontfERfRETATioN
::S 1
.
..
~
~
..
.
.
T.V. System
Slow-Scan S.!Jskm. .
AUiOMATic
B2
,_ ft-~oToiNT£RrR~TA1iON
~
: /
/
/
/
,.....,
/
/
/
.;;/.
7. V. System
Sloi.JJ"'- Scan S);.skm
/
~
.
COMPtli~R-ASSiST!.D C
r- fHoroiNit:RmTAiiON
.
/•
.
IJ
C.OM PARi SON
foE:-
.BEIW~E.N A1 B,
""
~ D
c
METHO.:OS
'~..-...--.------l ....
'RE..SULrs
location
olive.
(rus, .counb'n~ ~
of
stab.Ls of trees)
_9~er [fa hr r~s
L~92
AUTOHA.TI'c
:DiGiTAL
PHOTOGRAPHIC DIGITAL RECOGNITION PROCEDURE
1
BLUE
(rulgreen)
l'alsa Colour
lm~g~
Ti.n::
Cost (5):
I
Analog to d!Qital
~
clln~erllon
~
Chroma
Soparatlon
GREEN
(rulred}
LL_!JE.
Histogram of
lranlng areas
Hl&logram ol
tranlng areas
.1--
Hlstogron1 of
trEning aroaa
J--
Oeclalon on
propor width
1--
Oeclalon on
proper width
l_ffi\_.
Compoaltlon
r-
Photographic
r- Reproduction
Computing
Algorithms
,...__
Time:
C.:.~t
~-
($):
y
REO
(lnltared}
...t:::-
lD
\.N
~
I
11~4 2.
:~:~·~~~=
Decision on
proper width
-
---- --1
I
Time:
Cost !Sl:
J
Printout
._, Reproduction
J Time:
Cost (S l
Point de l'image situe ala colonne j
de la ligne de balayage i
f
>-_ non
___,__...! Retour
Retour
Retour
oui
- Nb. d'oliviers augmente
de I
Olivier
Retour
point (i, j)
Fig. 3 - Organigramme de detection des oliviers utilisant
les proprietes de texture locale.
LJ94
BASIC DOCUMENTATION
Geol., pedaL:, geog.,
h}'dr . macs, etc .
--
!
PRIOR ZONING
r-
Coll ect ion and Preparation
of Cadastral r~aps
Cutting Out/
Difficulty
Flight Plans
I Collection of Declarations by
Producers and Classification
Preliminary Ground Examination
by Phot oin terpreter s
Total Black and White Aerophotographic Coverage
Growing Zones
Constitution of a
Ground Truth
Reference Set
t
I
I
--j
Aerial Triangulation
J,
Construct.~on of Orthophotos
and plotting thereon of
Cadastral 'Boundar ies
I
~Examination
rl
~
"'>-
.......::0:
z
0
.....
u
:><:
....
...:
1
w
0:
a..
0:
w
of the Anomalies
....
z.....
of the Parcel Boundaries
.t
Cadastral Corrections on
L
Orthophotographs
I
L
- f!
l
8
"'0
w
0:
w
....
tr.l
Te""" An• ly• i'
Automatic Counting
Identification
Correlation
Counting/Parcel
'
f===
~
I
11
Thematic Map on the
Orthophotographic scale
.
r--- --
I
J
Photograph
II
"
I Quality Control:;on· Photo-
I graph
1..-t_
by Sample"
"
.
I -;~ncth_!_r Me.hcd _ _ _ _...;J I
r-cw~::
I
~
II
f.:> r .;
~
1 Digitalization
II
,,ol
~~
::
T
?R-SENTAT
.ON
of the
I~;;:;~~
ar.d
on
Listings, Thematic
·_;;~...:.c.~-<---------'
r·1aps
FQR ES.~T~A7S~L~IS~H~I~N~G~.:T~HE~R~E~G~I~STER OF OLIVE CULTIVATION
I Inventvry
I--
Orthophoto/1adastral
"
Ground Control
Request for Correc.: ions j
or Ex.:ni:-,aticn by
;t-
!Requ·; s t
0
u
*
Automatic Spectral
-.;.-, Ana Lysis
Identifi"ti6n Olive Trees
l
' I on
l
I
....a..
.;.
T
.! Quality Control by Parcel
i
I
z
....
I
'i
I
.... :..:
a..
•
Digitalization (3
Spectral componen t s)
~
Tracing on Orthophoto
of t he Results
I_ _ -
a...c.n
o:;;>
w w
Partial IRC Aerophotograph.i c Coverage
Low Relief Areas
I
Counting/Parcel
(a) manual
(b) ser.Ji -automat ic
I
r-
0:
LU
1-
Manual Ph~tointerpretation
(a) t r ad i t ion a l
(b ) statistic al
~
I
.
-
\\
~_j j
495