The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
995
£(cr ,) = (<?,- +CT*‘) +
4 Var(h) 4 R n
- tan 2 0, (tan 2 S„ + tan 2 S ± )
(2)
where Var(h) is the height variance.
Therefore the local surface roughness can be derived by
applying correction terms as follows
roughness nns = 0.5cja corr 2 -^j-tan 2 0,(tan 2 S„ + tan 2 S ± ) ^
In (3), CT C orr is the corrected pulse width which is recorded in the
MOLA PEDR data record where correction refers to filter
characteristics and threshold settings, as determined by the
receiver model (one sigma value, with the minimum limited by
the filter response; Neumann et al., 2003). The extracted a corr
value is corrected in two stages: (1) Removal of noisy tracks
which is shown in Figure 1 a. compared with some noisy free
reference planes such as HRSC DTMs; (b) slope correction by
Eq (3). However the extracted a corr values or local surface
roughness even after noisy track removal and slope correction
using MOLA height values which is corrected by crossover
check (Neumann et al., 2001) revealed strong across track
artefacts which can be observed in Figure 1 b) and c). The
reason for these across track artifacts is that some of the vertical
range errors within individual tracks were not corrected.
The height of the surface can be determined from the MOLA
measurements and the spacecraft tracking information (Abshire
et al., 2000) as
K - [««os 2 + * 2 - ZSAgs cos 2 m 2 - R,„ < 4 >
where R MCS is the radius of the MGS spacecraft orbit, R^ is
the radius of reference. The error in R MCS or possibly error of
the off nadir angle produces a deviation from the surface and
creates a false across-track slope effect.
(a) height values
using MOLA
uncorrected
planet radius
(b) The received
pulse width after
removing bad
orbits
(c) local roughness
after slope
correction using
500m MOLA
Gridded DTM
Figure 1. The influence of noisy MOLA tracks on topographic
height values, received pulse widths and local
surface roughness
Now there is one more correction term in equation (3)
roughness rmi = 0.5c
tan 2 0, (tan 2 S luncorrt
+ tan 2 S„ + tan 2 S x )
(5)
where S Luncorrt is the false slope which is produced by some
tracking information.
In this study, the slope correction procedure can be performed
in two ways.
1) Apply a slope correction using the height differences
between un-crossover corrected MOLA tracks and the MOLA
Grid DTM.
2) Apply a slope correction in the along track and across
direction using HRSC & CTX DTMs. The slope values are
extracted using nearest height points in the across and along end
side of individual MOLA footprints. This has the added
advantage of correcting for a local slope error. One very
important factor is the value of the nominal divergence angle.
Abshire et al. (2000) stated this value as 93grad. However,
Neumann et al., (2003) used an approximation consisting of
fitting 0.5 times the nominal angle from inflight data. The
nominal divergence angle largely influences the amount of
slope correction. We tried to find optimal values using
saturations of the slope correction and height variation of
HiRISE DTMs and found the optimal values as 33 grad.
3. RESULTS AND ASSESSMENT
Our study site is in the Athabasca valley. The main channel of
the Athabasca valley is covered by HRSC orbit 2099 and CTX
stereo pair (P01_001540_1889 and P02_001804_1889). Part
of the valley is covered by HiRISE stereo image pair
(PSP 001540 1890 and PSP_002371_1890). First of all,
various stereo DTMs were generated and then the correction
term was applied to the MOLA beam pulse width. Finally,
corrected MOLA local surface roughness tracks were compared
with ortho images and the standard deviation of height values
derived from the HiRISE DTM.
3.1 Stereo processing results
Over the central part of the Athabasca area, four different
resolution DTMs were extracted (Figure 3). These consists of a
50m HRSC stereo DTM; 18m CTX stereo DTM; 3.5m HiRISE
DTM for the whole HiRISE stereo area and a 70cm DTM over
some selected area.
Coarse resolution DTMs and ORIs were used for the geodetic
control of the next stage as described in the previous section, so
that registration between different DTMs and ORIs can be
guaranteed so long as the non-rigorous sensor model is
working. When the ORI products are superimposed from
different products they appear to be seamlessly co-registered.
This suggests that the geodetic control accuracy in this research
is sufficient to allow completion of this analysis (Figure 2).
Close-up images over a few areas (Figure 3(b)-(f)) using
different data resolutions, clearly show the kind of information
which is available from individual DTM sources. For example
50m resolution HRSC usually show the general outline of the
overall slope.