×

You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Title
New perspectives to save cultural heritage
Author
Altan, M. Orhan

CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey
208
to underline that a correct geometric positioning of the objects
is as important as the recognition of their modifications.
The self-developed RFM algorithm was applied to orthoproject
both images. It should be noted that the Rational Function
Model is very unstable and depends on many parameters, such
as the distribution of the GCPs, the range of the elevation
values, the sensor attitude and the Tikonov coefficient: all these
parameters have to be evaluated for each different image. Poor
estimations of these parameters can lead both to a decrease in
the positioning accuracy and to the occurrence of asynthops in
the final image (raw distortions).
A detailed statistical analysis was carried out to evaluate which
map scale the obtained orthoimages are suitable for, taking into
consideration the map tolerances (in Italy the usually accepted
value is of 0.2 mm at the map scale and the tolerance is
interpreted as two times the RMSE of the coordinates). This
means that a 1:10000 map has a 2 m tolerance and a 1:5000
map has a 1 m tolerance. Residuals can been considered as
statistical variables (one for each test) and their mean and
standard deviation can been calculated.
CHK
AE
(pixel)
AN
(pixel)
RMS
(pixel)
1
0,16
-0,82
0,83
2
-0,26
0,63
0,68
3
-0,67
0,82
1,06
4
0,87
-2,15
2,32
5
0,02
0,07
0,07
6
-1,95
0,01
1,95
7
0,51
-0,03
0,52
8
0,73
-0,03
0,73
9
-1,24
0,97
1,57
10
-1,06
1,08
1,52
11
2,24
0,46
2,29
12
0,55
1,42
1,52
13
-0,06
-1,01
1,01
14
0,04
1,64
1,64
15
0,56
-1,47
1,57
V
o v
0,03
1,0069
0,11
1,0817
RMSE
1,43
Table 2. Residuals on CHK over the SPOT5 image
Fitting statistical tests can be made on each residual distribution
in order to understand whether they would fit a normal
statistical distribution or whether they are affected by systematic
or raw errors. In particular, the % 2 test was performed. The
residuals (v) successfully passed this test in both cases.
In such situations, according to the Tchebycheff theorem
(/i — 2(7 < V < /A + 2(7 ), at least 95% of the residuals fall
into following ranges for the Spot5 orthoprojected images:
-4.95 -5.14 These results suggest that Spot5 orthoimages are not suitable
for a 1:10000 scale map updating; however tolerance values can
be accepted for a 1:25000 scale map.
CHK
AE
(pixel)
AN
(pixel)
RMS
(pixel)
1
1,20
-1,82
2,18
2
1,23
-1,73
2,12
3
-0,19
-0,06
0,20
4
1,57
-0,26
1,59
5
0,16
0,10
0,19
6
-1,53
0,35
1,57
7
-0,92
1,31
1,60
8
0,87
0,12
0,88
9
0,50
0,18
0,53
10
-0,71
0,27
0,76
11
-0,13
1,06
1,07
12
0,23
0,47
0,53
13
1,05
1,79
2,07
14
0,54
-0,05
0,54
15
-0,05
0,16
0,17
16
0,86
-1,16
1,44
1L_
0.2531
0.1271
0.8724
0.9521
RMSE
1,32
Table 3. Residuals on CHKs over the QuickBird image
As far as QuickBird orthoimages are concerned, their
Tchebycheff relation appears, as shown below:
-0.91 -1.08 Such results show that 95% of the residuals are lower than
1.25m and suggest a potential use of Quickbird orthimages for
upgrading 1:10000 scale maps (whose tolerance is about 2.0 m).
Another tolerance evaluation method can be considered
according to the technical rules that are adopted for the
production of the used 1:10000 regional reference cartography.
They indicate that the following relationship has to be satisfied:
RMS, = [{e™ k - E° rtho j + (n™ k - N° rtho J J< 4.0m (4)
However also this approach shows that Spot 5 orthoimages
seem to be not suitable for a 1:10000 scale map updating.
It is important to note that a more rigorous test should take into
consideration the check points precision q t (that is the same of
the test cartography), according to te following relation:
<7 v (5)
where (J u = precision of the up-to-date cartography.
Such an approach is less limiting and so it has been cautionally
decided to consider <7 T equal to zero.
4. CARTOGRAPHIC UPDATING
One of the main aims of this paper is to establish whether the
geometric information of the orthoprojected remotely sensed
images are suitable for cartographic updating (at a scale that
depends both on the geometric resolution of the image and on