International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
process would be required even if the photometric function
were precisely known (e.g., even for visible band images
without "magic airbrush" processing; see section 3 below
and Kirk et al., 2003b) because the image contains a uniform
additive offset that affects its constant. This offset comes in
part from atmospheric haze in the VIS image, which varies
with time and is not known a priori. Figure 2 shows the
photoclinometric DEM from the example of Fig. 1, with a
MOLA DEM of the same area for comparison. The
photoclinometry adds local details while preserving long-
wavelength topography such as the height of the 10-km
wide mesa at the bottom. The detailed DEM can be used to
simulate the VIS image with realistic photometric function
pv but no albedo variations. Dividing the VIS image (after
subtraction of a constant haze value) by the simulation
yields a map of albedo variations that can reveal subtle
features, e.g., the dark slopes on the sides of the mesa in Fig.
2d. The DEM can also be used to map absolute and
directional slopes.
We extended the analysis shown in Figs. 1-2 to neighboring
THEMIS VIS/IR triplets in order to produce topographic,
slope, and albedo maps of essentially the entire MER-A
(Spirit) landing ellipse, and also the MER-B (Opportunity)
ellipse. These landing sites are among the first areas on Mars
for which nearly complete VIS coverage is available, but
current mission plans will lead to the eventual collection of
global IR imagery and VIS images of about half the planet.
MOLA-resolved features that could be used to calibrate
photoclinometry were present in only a few of the images, so
the remainder were calibrated to have similar slopes on
small-scale features. As it happened, Spirit landed within
the images shown here, about 2 km west of the triangle of
hills seen at the top center. Our THEMIS maps were useful in
the early days of the mission for locating the landing point,
to ~100-m accuracy by tracing sightlines to features visible
from the lander. Hypothetical lander locations were also
tested by using the DEM to simulate the appearance of the
visible horizon. The high resolution of our maps (compared,
e.g., to MOLA) was critical for this application.
3. MARS EXPRESS HRSC
3.1 Source Data
In early January 2004, the ESA Mars Express mission started
Its science phase in orbit around Mars. Imaging and
mapping the Martian surface by the High Resolution Stereo
Camera (HRSC) is one of the main goals of Mars Express.
The HRSC experiment (Albertz et al., 1992; Neukum et al.,
2004) is a pushbroom scanning instrument with 9 CCD line
detectors mounted in parallel on the focal plane. Its unique
feature is the ability to nearly simultaneously obtain
imaging data of a specific site at high resolution, with
along-track triple stereo, with four colors, and at five
different phase angles, thus avoiding any time-dependent
variations of the observation conditions. An additional
Super-Resolution Channel (HRSC-SRC, a framing device) is
yielding nested-in images in the meter-range thus serving as
the sharpening eye for detailed photogeologic studies. The
spatial resolution from the nominal periapsis altitude of 250
km is 10 m/pixel for the HRSC proper and 2.3 m/pixel for
the SRC. The SRC images are normally acquired vertically,
but a subset are oblique because of spacecraft maneuvers to
fill gaps in image coverage.
3.2 Stereo Mapping Methodology
Our approach to processing HRSC data is largely
independent of those used by other members of the camera
team, which are described in several papers in this volume
(Hauber et al., 2004; Oberst et al., 2004; Ebner et al., 2004;
Heipke et al., 2004; Dorrer et al., 2004). We start with images
in VICAR format that have been radiometrically calibrated at
the German Aerospace Center (DLR) in Berlin. These images
are ingested into the USGS in-house digital cartographic
836
software ISIS (Eliason, 1997; Gaddis et al., 1997; Torson and
Becker, 1997; see also http://isis.astrogeology.usgs.gov).
[In particular, the supporting cartographic information in the
VICAR labels is converted into its ISIS equivalents. Be-
cause ISIS, unlike VICAR, does not currently accommodate
changing line exposure times within a scanner image, the
images are if necessary broken into multiple files that cach
have a constant exposure time.
From this point onward, our approach to topographic
mapping with the HRSC images is similar to those we have
used for a wide range of planetary datasets (Kirk et al.,
2000a), and, in particular, for the Mars Global Surveyor
MOC images as described in detail in a recent paper (Kirk et
al., 2003a). We usec ISIS for mission-specific steps (data
ingestion and, for most instruments, calibration, though for
HRSC the latter has been performed in VICAR), as well as
"2D" processing such as map-projection and image mosaick-
ing. The ISIS projection capability now includes ortho-
rectification, and detailed photometric models of the surface
and atmosphere (Kirk et al., 2000b, 2001) can be used to
correct the images for variations in illumination and atmo-
spheric haze before mosaicking. Two dimensional photo-
clinometry is also implemented in ISIS (Kirk et al., 2003b).
Our commercial digital photogrammetric workstation
running BAE Systems SOCET SET ® software (Miller and
Walker, 1993; 1995) is used for "3D" processing steps such
as control of the images and automatic extraction and
manual editing of DEMs. SOCET SET includes a pushbroom
scanner sensor model that is physically realistic but
"generic" enough to describe the individual HRSC scanner
lines, as well as most images from the Mars Global Surveyor
Mars Orbiter Camera (MOC). SRC images can be imported
and used with the frame sensor model. We have written
software to import these and other types of images from ISIS
into SOCET SET and to translate the orientation data in ISIS
into the necessary format. In this process, each HRSC
detector array must be treated as separate single-line camera,
oriented at an appropriate pitch angle to model the true
geometry of the instrument. Because SOCET SET is unaware
that the images from the different HRSC lines are collected at
the same time, the intrinsic robustness of the multiline
stereo system is unfortunately lost. This is a relatively
minor disadvantage, given the availability of the MOLA
global topographic dataset, which provides control with an
accuracy on the order of 10 m vertically and 100 m
horizontally (Smith et al., 2001; Neumann et al., 2001). The
disadvantage is more than made up for in practice by the
ability of SOCET SET to perform bundle adjustments that
include images from sensors of different types (e.g., SRC and
MOC), and to produce DEMs from stereopairs of mixed type.
Nevertheless, we plan to extend the bundle-adjustment
software being developed for the THEMIS IR scanner
(Archinal et al., 2004) to model the HRSC with proper
accounting for the constraints between image lines. This
software may also eventually be extended to model the high-
frequency pointing variations ("jitter") that affect many
MOC narrow-angle (NA) images. Because the SOCET SET
adjustment software we are now using includes only
smoothly varying pointing corrections, we must use ad hoc
processing by spatial filtering to remove the jitter-related
artifacts from MOC DEMs.
The MOC NA images are well suited in terms of resolution
for stereomapping in conjunction with SRC images. The
fundamental resolution of the camera is 1.4 m/pixel, but
images are most often obtained by pixel summation at
resolutions close to 3 or 6 m (Malin and Edgett, 2001).
More than 50,000 NA images have been obtained so far (see
http://www.msss.com/mars images/index.html and http://
ida.wr.usgs.gov/), and we have been able to locate vertical or
near-vertical images that overlap several of the handful of
oblique SRC image sets that have been obtained while the
Mars Express spacecraft is rolled off of its normal vertical
orientation. A larger number of pairs combining off-vertical
MOC images with the hundreds of vertical SRC images are
Interi
Figu
Mars
imag
Olyr
comi
throi
0397
5x9
stere
likely
part
m/pi:
SRC
two |
contc
these
33
Phot
DEM
imag
no h
prev
deter
contr
to be
topo,
Calit
trial
spher
agree
an a
imag
depei
comr
If the
has a
at be:
but if
avail.
ed to
Beca
with
usabl
albed
also |
4.1 |
The 1
Janua
rover
rover
et al,
inclu
morp
are re
