u 2004
—— —
ver the
ips are
where
d from
eration
um
ino
ndaries
Military
> whole
as been
0 maps
-Boaga.
> DTM
logy of
amount
es for a
with a
lifferent
both of
and the
ed fora
image.
Legrated
ind not
ated on
he IGM
1e laser
ith laser
al result
f sparse
meters,
; clearly
ribe the
»ver the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
m a.s.l.
10004
800-1
7004
6004
400
300+
200-1
100
0-7
T T
0 1000 2000 300 4000 5000 6000 7 8000 9000 10000 m
= hi
x
Figure 4. Integration of LIDAR and cartographic DTM. The transect shows the coherence between the two models. Blue dots
represent the IGM DTM, red dots the laser data
Before integrating the different models, a pre-processing has
been indispensable in order to get the suitability and coherence
of elevation data. Each laser strip has been compared with the
GPS-RTK ground survey and, for each one, it has been
determined the height shift with respect to the ground truth. It
has emerged that the shift is not constant between adjacent
strips because of problems in GPS signal reception at the flight
moment. Each strip has been corrected of the proper elevation
and all the strips have been adjusted together so as to obtain a
coherent final model .
Even the IGM DTM, that has a nominal height accuracy equal
to 7-10 m, has been compared with the GPS-RTK ground
survey. The average shift between the two data, equal to 3,50 m,
has been applied to the DTM.
The reliability of Lidar DTM is lower than the nominal
accuracy of the system (about 0.3 m), due both to the
morphology of ground (slopes up to 45?) and to the applied
corrections; however its accuracy, around 1-2 m, is higher than
that one of the DTM obtained from the IGM.
The final result of these adjustment is a correct and uniform
model over the whole area. Figure 4 shows the entire model and
a profile over the integration area (black line over the DTM).
4. ORTHOIMAGES GENERATION
Orthorectification of the QuickBird image has been performed
using PCI Geomatica OrthoEngine v8.2 software. The software
adopts different geometric correction models; among them, the
parametric rigorous model and the rational polynomial model
are the most accurate.
It should be noted that the rigorous model can be applied to the
entire scene and not only to a little part of it. Moreover, from
previous studies performed by our group, it has been found that
at least 20 ground control points evenly distributed over the
Whole scene are needed for an accurate orthorectification with
the rigorous model. On the contrary, QuickBird image is
delivered together with Rational Polynomial Coefficients
947
(RPC), that, theoretically, allow an accurate orthorectification
without ground control points.
Because of the availability of points from a RTK-GPS survey
not over the entire image but only over the study area, we chose
to ortorectify the image using the Rational Polynomial Model
with coefficients delivered with metadata.
Actually, previous studies performed using PCI Geomatica
showed that the software, when GCPs are not used, applies
erroneous transformation parameters from UTM-WGS84 to
Gauss_Boaga system, that leads to a shift of several tens of
meters from the correct position, both in East and North
coordinates.
For this reason, in order to eliminate the shift and the little
rotations it is necessary to use some ground control points
placed around the study area.
The orthorectification tests have been performed using different
DTMs to evaluate the influence of the morphology change on
the orthorectification procedure. First test has been carried out
with the IGM DTM, the second one with the integrated DTM.
For both tests the geocoding procedure has been led with four
ground control points placed as shown in figure 5.
Figure 5. Subset of the study area of an orthorectified image.
Yellow dots represent the GCPs location
