radiometric resolution (2116 dpi. 24-bit). Obtained bitmaps with
size of about 3090 x 2100 pixels were saved in TIFF file format.
The photogrammetric processing, carried out by a Socet Set
v5.] workstation, was concerned with a subset of 14 images
chosen in order to cover the temenos and dromos areas (Figure
5) Inner orientation was performed with an affine
transformation by using the image frame corners; values of the
residuals were less than one pixel.
Triangulation procedure, realized by bundle adjustment
analytical model, encountered some difficulties due to
unconventional strip geometry: indeed the irregular image
overlap does not fit with the Socet Set triangulation module
requirement of a strict distribution scheme for points connecting
contiguous models and strips. However, the process achieved
satisfying results with a maximum residual of 0.20 m on the
control point coordinates. Figure 5 shows the distribution of the
control points used for exterior orientation.
Next paragraphs describe the subsequent procedures of DTM
extraction and orthomosaicing.
z
Figure 5. Image coverage and distribution
of control points for the selected area.
3.1 DTM extraction
DTMs of the stereoscopic models were automatically extracted
by means of Socet Set image matching algorithm implemented
in the Automatic Terrain Extraction (ATE) module, based on
HRC (Hierarchical Relaxation Correlation) algorithm.
In particular, due to various morphologies that characterize
Soknopaiou Nesos area, “adaptive” correlation method and TIN
format were chosen, in order to adopt the most suitable
correlation strategy in each zone and make easier the successive
interactive editing operation.
Comparison between manual and automatic extracted DTMs,
depicted in Figure 6, points out an effective working of ATE
module: about 81 per cent of points shows an absolute value of
the difference lower than 30 cm. The biggest differences are
localized close to structures or in sandy areas and so they are
connected with sharp height variation or with homogeneous
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004
image texture, that makes difficult the image matching
procedure. =
The experiment was conducted on sample areas with different
kinds of morphology.
In order to improve the automatic process result in
correspondence of such areas, we experimented the opportunity
of using seed DTMSs to initialize ATE process.
i Yo. Lea 1 1 ERE
@ -17 to 1
B 110-0.
32696604 @ et a L
-03 to 03
& 0.3106 1
& 11012
3269620 L
3269580-] E
x
*
3269540 hs
T T T T
274000 274040 274080 274120
Figure 6. Classed plot of residuals between automatic and
manual DTMs (automatic minus manual, meters).
In particular, DTMs with about 1 m medium inter-post spacing
were extracted in each test area by using three different
automatic collection modalities:
a) without any pre-existing elevation data;
b)starting from seed points obtained by kinematic GPS
survey (only terrain information);
c)starting from seed points derived from a manually
measured DTM with grid spacing of about 3 m.
We report here the results obtained for a test area (Figure 7)
characterized by abrupt elevation changes due to a femenos
wall.
Figure 7. Test area around a temenos wall.
Intern
Figure
autom
one. |
impro
conclt
In ou
mistal
neighl
introd
condu
