04
USE OF ORTHOPHOTOS AS GROUND TRUTH IN IKONOS IMAGE PROCESSING
F. Blasco””, M.F. Bellan”, V. Barbaroussi/, and G. Miliaresis*
" CESBIO-LaDyBio, 18, avenue Edouard Belin, bpi 2801, 31041 Toulouse cedex 4, France -
Francois.Blasco(acict. fr
b CESBIO-LaDyBio (ex LET), 18, avenue Edouard Belin, bpi 2801, 31041 Toulouse cedex 4, France
- Marie-France.Bellan@cict. fr
‘ Topography Dpt, National Technical University of Athens, 9, H. Polytechniou, Athens, 15773 Greece —
vbarbar(@survey.ntua.gr
Commission PS, WG IV/ 7
KEY WORDS: Remote Sensing, Spatial, Interpretation, Landscape
ABSTRACT:
This is the first attempt to use very high spatial resolution satellite sensor data (IKONOS) on Greek landscapes. The
IKONOS imagery analysed in this study covers a hilly area close to Corinth Gulf, behind an industrial harbor. The
processing seeks to retrieve the main units of a Mediterranean, more or less degraded or preserved, vegetation. The
validation of the results obtained through any satellite imagery processing is only possible with the help of ground data.
In Greece as in many Mediterranean regions, field work may be difficult due to low accessibility. The present paper deals
with the use of aerial photographs as ground data in the assistance of an IKONOS image processing. 1,280 sampling
points are located on orthophotos. Their interpretation is compared to the result of a pre-classifying of IKONOS image,
allowing to noticeably improving the classification.
I. INTRODUCTION
In Mediterranean countries, where human activity has
happened for a long time, the vegetation cover reveals
both environmental factors like geomorphology,
elevation and ever bioclimatic element, and the
intensity of human pressure due to grazing, cultivation,
fires, etc.. These results in some complexity relied to
the juxtaposition of dense vegetation in the most
preserved places with more or less degraded types
elsewhere. SPOT HRV data, with 10 or 20m spatial
resolution, do not totally express such a variability
within a broad unit * Mediterranean vegetation types at
low elevation *. The very high ground resolution of
IKONOS data is a new step towards a detailed image of
the vegetation cover, close to an aerial photograph but
with the geometric quality, the homogeneity and
periodicity proper to satellite imagery. However,
whichever scale or detail may be available, ground data
acquisition is always necessary to carry on, to evaluate
and to improve digital processing. Especially in
complex landscapes such as those commonly found in
Mediterranean regions, automated classifying makes
arise serious problems, even on high spatial resolution
satellite imagery. However, field work is sometimes
difficult to carry out, due to time consuming or hard
access. This paper points out the role of air photographs
in ground data acquisition for helping digital processing
of very high ground resolution satellite sensor data such
as IKONOS imagery. The study area is located in
Greece, Corinthian Province, where an IKONOS image
covers the surroundings of Agioi Theodoroi harbor. It is
the first attempt of using these very high spatial
" Corresponding author.
63]
resolution data on Greek landscapes. * Space Imaging
Europe ‘ provided for this purpose a PAN/MSI product,
with Im ground resolution and four spectral channels.
2. METHODOLOGY
2.1 Previous Works
IKONOS, first commercial high spatial resolution Earth
imaging satellite, was launched on 24 September 1999
from USA. Since this date, the high quality of its
products has been appreciated by many users. Through
visual interpretation, interesting results were already
obtained (Sotaro Tanaka and Toshiro Sugimur 2001,
Imbernon 2003). Other works focus in textural analysis,
especially on forest applications (Franklin et al. 2001).
In the present study, classical digital processing is
applied in order to classify the vegetation types.
However, in spite of a general knowledge of the
vegetation cover in this part of Greece, accurate ground
data are a prerequisite to retrieve important ecological
information from this kind of data. Especially in
Mediterranean areas, the landscape is often complex
and a good satellite imagery processing is strongly
dependent of a thorough field work (Merzouk et al.
1998) covering well known training areas. In Greece,
some attempts were carried out with SPOT 1
(Barbaroussi 1988). Then, it was found suitable to
explore the benefit of using very high spatial resolution
data for a better analyses of mediterranean landscapes.
Unfortunately, in such hilly regions as these found in
Corinth, training sites are difficult to access and to
