The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
975
take advantage of post-mission improvements to the spacecraft
ephemerides (Rappaport et al., 1999). Gridded Magellan
altimetry data are also read in, for use as a source of vertical
control. All these data sets are exported to SOCET SET by
means of ISIS translation programs.
SOCET processing optionally begins with the collection of
tiepoints in the overlap areas between BIDRs and bundle-
adjustment of the spacecraft trajectories. Some of the points
may be constrained to lie at elevations given by altimetry. The
trajectories are typically adjusted by introducing offset and
linear drifts in the three orthogonal directions in-track, cross
track, and radial. Adjustment parameters obtained from
measurements on BIDRs can then be used in processing
mosaicked data sets containing those BIDRs.
Automatic DTM generation is achieved using SOCET SET’s
Automatic Terrain Extraction (ATE) module (Zhang and Miller,
1997). Although a lower limit on the useful spacing of DTM
data that could be collected from the 75 m/pixel images would
be 225 m (3 pixels) per post, we routinely generate DTMs at
675 m/post, mainly to allow for greater averaging over speckle
noise in the SAR images. Prior to running ATE, we "seed" the
DTMs with manually collected points on ridge and valley lines,
or with reliable altimetry data. This greatly improves the
success rate of the automatic matching step, and generally
limits the need for manual editing to bland areas, where the
matcher fails entirely, and to ‘blunders’ found at the image
edges where elevation values are extrapolated. The FMAP
mosaics are normally used for ATE, avoiding the need to define
a large number of image pairs made up of the smaller and more
numerous BIDRs. The automatically generated DTM is viewed
in stereo along with the images and interactively edited. Where
possible, editing is based on the FMAPs, but some seam areas
may need to be edited based on the individual BIDRs. After
interactive editing, individual DTMs are combined into a single
DTM for the entire map area. The merged DTM then requires
additional interactive editing to replace gaps (due to missing
data in orbits) with corresponding altimetry data. Finally, the
BIDR images may be orthorectified and mosaicked, yielding an
image base that registers more precisely to the DTM than the
standard FMAP. The DTM and orthomosaic may be exported
in various formats for analysis in ISIS and the production of
publication-quality maps with other software such as ArcGIS
and Adobe Illustrator. When mapping with FMAP mosaics, we
produce 1:1,500,000-scale topographic maps, with a contour
interval of 200 meters, with orthomosaic base, nomenclature,
and collar information. We also produce color-coded shaded-
relief/elevation maps (Figure 1) because they portray subtle
topographic relations that assist with analyses of tectonic
deformation, stratigraphic interpretation, flow direction, mass
wasting, etc. The color-coding is chosen to show as much
information as possible within a given map area, and hence is
not necessarily consistent planetwide.
Figure 1. Example map product: a color-coded shaded relief
map of the 12°xl2° Joliot-Curie (06S066) FMAP quadrangle on
Venus. Stereo data collected at 675 m/post have been edited
interactively and merged with Magellan altimetry data.
Enlargement at right gives an idea of the comparative detail
level of the altimetric and stereo data sets (smooth and rugged
strips). Full size Magellan stereo-derived maps are available at
http://webgis.wr.usgs. gov/pigwad/down/venus_topo.htm
2.3 Testing and Validation
We have validated our Magellan stereomapping techniques as
carefully as possible, given the limited availability of other data
with which to compare the results (Howington-Kraus et al.,
2006). As a first step, mapping of a small area showed that
ATE combined the high speed of automatic matching available
in MST with much higher DTM resolution and gave results
consistent with DSW-V, as expected given the commonality in
the sensor model code used. Mapping of a larger area, the
12°xl2° FMAP quadrangle 06S066 (Joliot-Curie) allowed us to
develop procedures for making accurate controlled products.
Magellan mosaics were previously known to contain
discontinuities of as much as several km between data from
orbits whose ephemerides had been calculated in separate
solution blocks. Stereo viewing of the mosaics reveals these
discontinuities as apparent “cliffs” in many cases. New
ephemerides were computed after the mission on the basis of an
improved gravity model, with a claimed reduction in position
errors by 1.5 orders of magnitude (Rappaport et al., 1999). Our
test mapping showed that the use of the improved ephemerides
reduced discontinuities in the DTMs significantly, as well as
presumably improving absolute positional accuracy, but that
bundle adjustment based on image tiepoints was required in
order to achieve pixel-level consistency. In both of these initial
tests, some manual editing was required in relatively flat and
featureless areas. Improved results were obtained in such areas
by using the Cycle 2 images, in which the spacecraft was on the
opposite side relative to Cycles 1 and 3. This provides greater
stereo parallax, but also reverses any topographic shading,
making stereo matching more difficult where such shading is
present. In the lowlands of Venus, where relief and shading are
low, the opposite-side imaging proved to be ideal.
Mapping at high latitudes initially failed because of apparent
multi-kilometer offsets between the images being used, even
though the same BIDRs aligned properly at low latitudes. This
problem was eventually traced to the use of an insufficiently
precise value of the radar wavelength in the sensor model.
Smaller discrepancies in cross-track coordinates were traced to
the different atmospheric refraction models used in the DSW-V
and the Magellan processor used to make the BIDRs. The
Magellan model contains a bug that renders along-track
adjustment of spacecraft positions impossible, so the DSW-V