The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
137
portion of the comer reflectors. The reflection from the comer
reflector is much stronger than neighbourhood.
Figure 2 shows the comer reflector (CR) used in this study. The
reflector consists of 3 sides of rectangular aluminium panels
with 50 cm length. The panels are attached perpendicular to
each other by bolts. We measured the coordinates of the CRs.
using GPS (accuracy is better than several centimeters). From
the comparison of the coordinates derived from TerraSAR-X
and CRs, pixel location accuracy of TerraSAR-X was evaluated.
Tokyo1
Tokyo2
Tsukuba
Acquisition
Date
16 th Feb.
2008
29 th Nov.
2007
23 th Oct.
2007
Mode
HS
HS
HS
Incidence Angle
39.4
42.2
52.7
Orbit
Ascending
Descending
Descending
Scene Center
35.7 N,
139.8 E
35.6 N,
139.8 E
36.1 N,
140.1 E
Table 5. Specification of used TerraSAR-X data.
Figure 1. TerraSAR-X EEC product of Tokyo on 29 th Nov.,
2007 and an enhanced portion of the comer reflectors.
3.2.1 Pixel location accuracy using science orbit: Table 6
shows the comparison of location accuracy of both SSC and
EEC product when science orbit information was used. This
analysis was done using the image of Tokyo of 29 th Nov. The
latest supplied processing system was applied to generate the
level 1B product.
The radar signal is subject to path delay due to the different
refractive indices of vacuum, ionosphere and troposphere. This
results in slant range error of the order of 2-3 m that depends on
the actual conditions in the passed media and on the length of
the signal path hence the incidence angle. This value is
annotated in the meta data annotation file.
The pixel location accuracy of the SSC product for slant range
and azimuth was estimated using the projected spacing value
for slant range and azimuth, respectively. Errors of slant range
are 0.4 m, and azimuth is -0.5 m. Therefore the result of the
pixel location accuracy is better than 1 m. Almost the same
results were obtained for Tsukuba site.
The pixel location accuracy for the northing (Y) direction was 1
m, though easting (X) direction of the EEC product was not
better than 1 m,. It is assumed that the vertical errors of DEM
affect the pixel location accuracy of the EEC product.
(a) SSC
TSX
GPS
Difference
Slant Range (pixel)
2781.46
2781.03
0.43
Azimuth (pixel)
2727.04
2727.62
-0.58
(b) EEC
TSX
GPS
Difference
X (m)
392133.5
392137.1
-3.6 m
Y (m)
3948968.6
3948967.6
1.0m
Table 6. The location accuracy of the SSC and EEC product
calculated using the image of 29 th Nov.
3.2.2 Pixel location accuracy comparison using another
orbit type:Table 7 shows the comparison of location accuracy
of EEC product by different orbit type. Even if smaller X
deviations for the predicted orbit observed, that doesn’t mean it
is more accurate because the errors of Y is much greater.
It reveals that in this case the pixel location accuracy is almost
same when using rapid and science orbit. This demonstrates
that JPL real-time GPS products used for the generation of the
rapid orbit product already provides high accuracies (Yoon et
al., submitted).
Predicted
Rapid
Science
X (m)
0.3
-3.6
-3.6
Y(m)
11.7
1.0
1.0
Table 7. Comparison of the location accuracy of the EEC
product by different orbit types. The acquisition
date is 29 th Nov.
Figure 2. Comer reflector used in this study.