TA
g 550 m was
r orientation
2.2 m, RMS
m. DTM has
pared on the
6 000.
S and CNES
generation.
] lA - (size
Bavarian test
/ith metadata
acquired on
's and nearly
e matching
image stereo
id backward
as a spatial
nd sampling
racy of the
Fthe HRS is
! sequence is
n regarding
star sensors)
ivered data;
ith detailed
point, with
the RMSE °
rm:
le 1: 25 000
alization of
2. THE METODOLOGY OF GEOMETRIC
CORRECTION OF PANCHROMATIC STEREO HRV
SPOT 4 DATA
2.1 Specification of methodology
In order to determine the factors influencing the result of
orientation of stereoscopic scene HRG SPOT 4, the series
of methodical research work was performed. The
following factors were considered:
- Filtration of raw SPOT 4 data in spectral and
frequency domain;
- A Accuracy of measurement and identification of
control points and check points;
- . Accuracy of measurement of image coordinates;
- Accuracy of delimitation of satellite orbital
parameters;
- Distribution of ground control points.
The specific of ‘measurements the each of
photogrammetric point on stereo images SPOT 4 (the
semi-automatic correlation method) did not require the
detailed analysis of influence of radiometric quality of
these images on accuracy of relative and absolute
orientations. However, the ability of identification of
control points and check points was examined on original
scenes and on the images after radiometric corrections.
The accuracy of control points depends on accuracy of
their measurement in field as well as on accuracy of
identification on SPOT 4 images.
The image coordinates of control points, check points and
tie points were measured by correlation methods using
TRIFID software installed on the SSK Pro workstation,
under the Intergraph photogrammetric environment
(Image Station Multisensor Triangulation - SPOT).
The analysis of results of bundle adjustment shows that
the range of estimated accuracy of photogrammetric
adjustment is good when the standard deviation of
adjustment does not change. The choice of accuracy of
measurement outside of this range worsens the credibility
of estimation of adjustment. For the tested panchromatic
stereo SPOT 4 scene the accuracy of photogrammetric
measurement was accepted with range from 4pm to 6um.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
If GCP's are symmetrically distributed then twelve points are
sufficient for orientation of stereoscopic SPOT 4 scene.
The influence of accuracy of delimitation of satellite orbital
parameters on absolute orientation results has been examined
for established construction of bundle and determined number
and distribution of control points as well as tie points on
stereoscopic SPOT 4scene. This accuracy is characterized by
standard deviation. It has been confirmed that the results of
adjustment being the function of these standard deviations do
not change significantly; therefore the approximate values of
sensor parameters for orientation of stereoscopic scene HRV
SPOT 4 were used.
2.2 Results of geometric corrections
The results of geometric correction of panchromatic
stereoscopic scene HRV SPOT 4 for above-mentioned analyses
were received, in reference to full scene and to the separated
fragments. Results are shown in Table 1.
In orientation process, except the approximate parameters of
orbit, the control points measured on topographical map in scale .
1:10,000 permits for orientation the panchromatic stereoscopic
scene of HRG SPOT 4 with accuracy below half pixel. The use
of control points measured in field by the GPS techniques or
appointed in space triangulation process on the basis of aerial
photographs in scale 1:26,000 permits to increase the
orientation accuracy to 1/3 of pixel.
3. THE ACCURACY OF GEOMETRIC CORRECTION
OF BLOCK SCENES HRS SPOT 5
Each scene has been oriented independently using from 15 to 36
ground control points. Table 2 shows the numbers of ground
control points and checkpoints on rectified scenes. Many natural
GCP's have not been well identified. The planimetric accuracy
of control points and checkpoints is about 1m. There is also an
identification error that can reach up to 1m. For good B/H ratio
RMSz=1.5m.
Results of orientation with the use of Toutin model (PCI
Geomatica v. 9.1 software) and ground control points are shown
in Table 3. The accuracy of orientation of block panchromatic
scenes HRS SPOT 5 is 2m in planimetry and 2m in height
independently form terrain type.
s ; : RMSE in meters on the:
Range of Number of control Number of Control points Check Doms
elaboration points check points X Y 7 Y Z
Full scene 12 15 222 23 +} +50 +20 + 4.7
First test region 14 7 $21 4 £17 312 +44 = 33 +34
Second test region 16 13 210 +15 +00 | 130 +30 t3]
Table 1. The accuracy of orientation of stereoscopic scene SPOT 4
01 02 03 04
Scene No. (5 062-252/1 18) (5 062-252/1 28) (5 062-253 18) (5 062-253 28)
Ground Control Points 15 15 36 36
Check Points 5 3 16 16
Table 2. Numbers of GCP's and checkpoints used for orientation of the scenes
