International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
[HRSC Strip [| Date — | Nadir Resolution [m] |
h2224 0001 | 2005-10-07 20.9 -31.6
h2213.0001 | 2005-10-04 20.5 -304
h8362 0000 | 2010-07-15 19.8 - 292
h2169 0001 | 2005-09-22 16.3 - 243
h5248.0000 | 2008-02-02 13.7 - 18.7
h5230.0000 | 2008-01-28 13.3 - 17.5
h5212.0000 | 2008-01-22 13.1 - 163
h2114.0000 | 2005-09-06 122 - 18.7
h2103.0000 | 2005-09-03 12.0 - 15.7
h2081.0000 | 2005-08-28 11.6- 144
h2070.0000 | 2005-08-25 11.6 - 13.7
h2059 0000 | 2005-08-22 11.6 - 13.2
h2037.0000 | 2005-08-16 12.2 - 12.7
h3193_0000 | 2006-07-05 13.8 - 18.5
h8383.0000 | 2010-07-21 18.3 -282
h3160.0000 | 2006-06-26 132 - 15.1
h2004.0000 | 2005-08-06 12.8 - 12.6
h1993_0000 | 2005-08-03 13.4 - 12.5
h1982.0000 | 2005-07-31 14.8 - 12.1
h1971.0000 | 2005-07-28 15.5 - 12.4
h1960.0000 | 2005-07-25 17.2 - 12.6
Table 1: HRSC strips used for the block
© nominal orientation
8 adjusted orientation
mean intersection error [m]
SSSSSSSSSS SSS SSS SSS
Fe Try hhh à FES SS
SM 23? AS SS Po N oF
S.C. SS
HRSC strips
Figure 7: Mean intersection error before and after single strip
bundle adjustment
For the stability of the complete block the spatial accuracy of tie
points in the overlapping areas is important. Hence for all 20 sub-
blocks with an overlapping area the mean intersection error was
calculated using the different orientation data sets (Fig. 8).
For the computation of the complete block as a whole, the number
of tie points for each sub-block was reduced to a target value of
8.000.
In all but one sub-block the block adjusted orientation data show
a clearly higher accuracy for the HRSC points in the overlapping
area than orientation data adjusted for single strips. In sub-block
h2169-h5248 the overlapping area is too small to find a sufficient
number of tie points. All sub-blocks with a mean intersection
error notably higher than 10m consist of two image strips with
different illumination conditions (Fig. 9).
Overall the adjustment of the exterior orientation in a regional
block does not only provides results with a high internal accu-
racy for each strip, but also the relative orientation of the strips is
clearly improved. Additionally the block as a whole is fitted to
a global reference system, the MOLA DTM. The block adjusted
orientation data are the most appropriate ones for the subsequent
derivation of a DTM mosaic or the rectification of ortho-images
© nominal orientation
B strip adjusted orientation
B block adjusted orientation
mean intersection error [m]
HRSC sub-blocks
Figure 8: Mean intersection error for overlapping areas in sub-
blocks calculated with different orientation values
(Dumke et al., 2008). For the depicted example of 21 image
strips, the HRSC DTM mosaic is shown in Fig. 10.
Figure 10: Shaded relief of a HRSC DTM mosaic generated with
the block adjusted orientation data
5 CONCLUSIONS AND FUTURE WORK
The workflow for systematic bundle adjustment of HRSC image
data was used to improve the exterior orientation of single strips
and a regional block. It has been shown that for 2329 single strips
a significant improvement was achieved. A block comprising 21
HRSC strips was simultaneously adjusted and the results were
evaluated to demonstrate the stability of the overlapping areas.
The automated implementation of the workflow and the parallel
processing using the performance of a university computing cen-
ter allow to locate stable overlapping areas in all available HRSC
strips in a systematic way. The results of this analysis can be used
to design larger regional blocks. It is also possible to investigate
blocks of strips connected to rings around the equator and if gaps
in the HRSC data are filled with images from other missions a
closed global block completely covering the Martian surface.
288
A SU ME E^ e
