bul 2004
ammetric
have to
] find no
phase of
ntee the
"rroneous
; in some
ent study
ich are to
the GIS.
automatic
nages by
se. PRO,
d whith a
nphasized
essary to
e, ground
ructure of
Ocess of
tr adapted
:'oefficient
determine
, thereby
adation of
A will be
ol points
rovince of
:a with a
1 and flat
90-2800 m
es with 10
2-11-1991
km.
| points to
points to
oints have
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Figure 1. Study area
(SPOT image dropped over DEM).
4.3 Sofware
SPOT-HRV data were processed two photogrammetric
aplications: Socet Set (Bae Systems) and OrthoBase Pro
(Erdas Imagine). The rest of the procedures were carried out
using the GIS, ArcView 3.2, with the modules 3D Analyst
and Spatial Analyst (ESRI, 1998).
The photogrammetric applications only
characteristics are commented below.
variying
* Socet Set (Leica Geosystems)
Socet Set allows work with a specific module for SPOT
data. The DEM may be generated as either a TIN or as a
raster Uniform Regular Grid, or URG.
* Erdas Imagine 8.5 with OrthoBASE Pro
OrthoBASE Pro has a specific module to work with SPOT
data, but ASTER is only supported by means a generic
module introducing the values for angles, B/H ratio, etc. The
DEM may be generated only as a vector structure, a
Triangulated Irregular Network or TIN.
4.4 DEM generation: extraction of elevations
In the photogrammetric process, the stereo-matching consists
basically of locating homologous points in the images. In the fit
to the SPOT images, RMSE values of 0.5 pixels may be
attained from a small number of ground control points as long
they are appropriately distributed spatially. The process has a
relatively high precision since the collinearity equations allow
one to obtain a direct relationship between the coordinates of
the image and the object. One thus obtains the relative
orientation and the model coordinates, and the calibration
parameters may be included in the reduction of. systematic
errors. The Socet Set application used performs the orientation
of a stereoscopic SPOT pair with the module MST (Multisensor
Triangulation), and the identification of homologous points can
be performed by area based matching.
The elevation was calculated using an iterative algorithm which
begins with the top level of the image pyramid (that of poorest
257
resolution), and advances to the highest image resolution. The
DEM may be generated as either a vector structure - a
triangulated irregular network, TIN (Peucker et al., 1978) - or as
a raster structure - a uniform regular grid, URG. The latter does
not require the position to be stored since it is implicit in the
structure itself. The TIN structure may be adapted to the type of
relief, i.e., to changes in the topography of the surface. We
constructed a URG-DEM and chose a pixel size of 20 m. The
automatic extraction of DEM is facilitated if the specific sensor
model information is available.
In order to guarantee the best possible DEM that can provide
SPOT-HRV images, we have analyzed the influence of some
aspects, such as number and spatial distribution of GCP, the
data structure. (TIN or URG), and the sample interval;
depending on the software used, the algorithms and correlation
coefficient threshold can also be tested.
We have conducted several experiments to determine the
optimal value of influential aspects (Table 2). We constructed
ninety SPOT derived DEM (see the results section, Table 4) .
o N° of
Variable Ne of nid. DEM by
Test s Range of values DEM by ne
analyzed OrthoBas
¥ SocetSet
] number of CP? 5...20 16 16
2 dd ef 4 distributions 4 4
3 data structure TIN ®/ URGE 2 2
ES 100, 80, 60, 40,
4 size of grid 20, 15, 10 m 7 7
5 algorithm of several 2 12
matching
ene : SocetSet: 0.5...1
coefficient of
6 comelation OrthoBase: 11 8
s 0.6...0.95
42 49
DEMs 91
generated:
* Control Points.
* Triangulated Irregular Network.
* Uniform Regular Grid.
Table 2. DEM generated from SPOT-HRV images.
4.5 Accuracy and realibility
DEM accuracy is estimated by a comparison with DEM Z-
values, and by contrasting many check points with “true”
elevations. The pairwise comparisons allow the calculation of
the Mean Error (ME), Root Mean Squere Error (RMSE),
Standard Deviation (SD) or similar statistics.
It’s obvious that reliability in the process is not a constant but
depends on several factors. The number of chek points is an
important factor in reliability because it conditions the range of
stochastic variations on the SD values (Li, 1991). Another
factor is obvious: The accuracy of check points must be
sufficient for the control objectives.
The estimate of errors in DEM is usually made by following the
USGS recommendation of a minimum of 28 check points. Li
showed, however, that many more points are needed to achieve
a reliability closer to what is accepted in most statistical tests.
The expression that relates reliability to number of check points
is:
