International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
Type of parcel Topography
Scale of the
Variation of the
orthoimage area between the extracted data and
(9/0) the reference data (m)
Rural — Large estates Flat terrain 1/10 000 0.14 : 4.3
Rural — medium estates Rugged terrain 1/5 000 0.22 6.8
(10 — 20ha)
Rural — medium estates Rugged terrain 1/5 000 1.77 8.5
(5 — 10ha)
Rural — small estates Flat terrain 1/5 000 3.7 3.7
(smaller than Sha)
Peri-urban estates Flat terrain 1/2 500 22.6 4.33
Urban estates Flat terrain 1/2 500 Do not apply Do not apply
Mean variation of the distance
Table 1. Results obtained for each type of properties.
However some of the vertices have a difference of localization
of less then one meter.
There is a significant difference of result depending on the size
of the property. Medium properties between 10 and 20ha
presented a variation of surface of 0.22%, yet estates of 5 —10ha
presented a variation of 1.77%. The difference of vertice
location between the datasets was of 6.8m for properties of 10 —
20ha. and of 8.5m for properties of 5 to 10 hectares. Field
borders in rugged topography presented a very low density of
vegetation, making the identification on the image more
difficult (figure 4). When the border is under dense vegetation,
the distance variation between the digitized border and the one
survey can be as high as 11m (figure 5). In the case where no
physical border exists the difference can be as high as 13m.
Adding to this case, when there is no land use change, it is
almost impossible to identify the border; the surveyor can only
estimate the location of the vertice.
For small estates in flat areas the variation of the area detected
was of 3.7% and the average variation of the distance between
the borders identified in the image and the survey one was of
3.78m.
4.2 Urban properties
By peri-urban parcel, we consider properties of 500 — 1600m?
with a combination of residences and survival cultures. Area
variation was of 22.6% and the variation between borders was
4.33m. Most of the cases presented no physical boundaries,
only ‘legal invisible’ fences. The presence of fruit trees, make
the identification less accurate as we had to generalize the
location of the vertice under dense canopy.
Where land consolidation is present, a variation of 1.48m of
length is obtained.
Urban identification on the orthoimage was limited as it can be
observed on figure 6 and 7. The areas of the buildings surveyed
were smaller than 500m?. The borders were barely identifiable
in the orthoimage, making the delimitation of boundaries almost
impossible. We have to note that building forms are highly
changed between what can be identified in the image and the
real aspect of the building as we can observe in figure 7. This is
mainly due because of similar spectral response of the building
and its entourage, where no contrast is identified the accuracy of
image identification drops. Details of 8m of length could not
be identified in the satellite image data.
5. ANALYSIS AND DISCUSSION
As.the area covered by the orthoimage is elevated and the use of
digital data presents some advantages, different preliminary
work can be done with the image. In this particular case, the
use of satellite images can be implemented since the beginning
of the cadastral activities, for example in:
- planning activities,
- preliminary analysis for the selection of methodologies to be
used,
- preliminary definition of cadastral polygons and municipality
boundaries,
- to update the cartography, and
- for land-use/land-cover studies and general studies of the
region surveyed.
Nowadays in Guatemala, the planning of cadastral activities is
made with outdated cartography, 1/50 000 maps and in many
cases without any previous field supervision. The design of
cadastral polygons is completed using official cartography and
photogrammetric products from the 80' ad adjusted as the
cadastral mapping advances. The delimitation of municipalities,
which cadastral polygons cannot surpass, is done in a
preliminary step with invalid maps and is adjusted on the field
inspection making the planning step costly and time consuming.’
SPOT 5 orthoimage brings better elements than the ones used at
the present .
Eight GCPs were provided to relocate the image to a 2.5m
spatial resolution of the orthoimage. Ideally a GCP should be
as small as a single pixel and an appropriate number should be
well distributed in the study area (Campbell, 1996, Korgaja et
dl, 2003). In tropical woodlands, steep topography and for
distant regions, to obtain appropriate GCPs is a challenging
task. The nature of the study area will influence the
georeferenciation of the orthoimage and the resulted accuracy
border identification.
The image used in this study presented some limitations, the
presence of clouds and haze restricts the area of utility and the
identification of parcel borders is more complex. It is
recommended to work with enhanced images, as boundaries are
most of the time linear, more easily identified. Depending on
the average size of the properties, an adequate scale was chosen.
Yet, as the image quality for cadastral mapping was not the
most adequate one, most of the scaled images presented only
the area of interest (cfr. Table 1). We cannot suggest an
appropriate scale for preliminary or definitive cadastral plans, as
the election of this one will depend on the nature and size of the
properties — latifundium, medium and small estates, peri-urban
and urban properties —, topography, atmospheric effects and
precision needed. However, in this study we obtained more
accurate results for medium scales (1/20 000 to 1/5 000) than
for small ones (1/2 500), this is mainly due to the resampling
method applied to the stereopair images of 5m spatial resolution
to obtain a 2.5m orthoimage.
As we can observe, the variation of the parcel area between the
referenced data and the one derived form the image
interpretation depends directly of the size and topography of the
property. The variation of area was considered very low for
latifundium estates with no topographic variation and for
medium estates of 10 — 20ha. However, properties in rugged
terrain, small agricultural estates and peri-urban properties
Inter
—
prese
variat
land «
For I:
variat
extra
image
rugge
variat
live |
locate
with
bounc
legal
areas
plant:
The
numt
prese
ident
bounce
less a
accid
cover
be lin
The
meth
prope
medi
urbar
impoi
their
urban
areas
dense
limite
data
logis!
estim
sourc
resol
time
photo
from
exam
most
And
comb
ident
As 1
ident
8.5m
diffei
the u:
meas
meas
Fisca
Poly
the d
for le
is 0.
toler:
diffe;
for 1
toler:
foru
diffe:
