2012
.owloon,
vations. For
them. This
imeter data.
parameters
n the Moon
nese robotic
1ifts in both
topography
ed works, a
nsformation
ce rotation
ed for the
models. The
t the Sinus
nployed for
remarks are
| by several
y, sterco-
the Moon’s
results were
on, both the
Smith et al.
aphic model
1ed by the
jined Earth-
tereoimages
llo, Mariner
rol Network
trol network
;s amount of
» new lunar
Chang E-1,
yhic models
camined the
ce frames in
yer of nearly
nissions. A
as used for
otions and
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
deformations. They found that the estimated relative rigid body
motion and deformation parameters between the two reference
frames are consistent (i.e., nearly zero estimates for the
translations, rotations, and shear parameters), while the three
strain parameters, which are similar in magnitude and sign,
reveal a statistically significant scale difference of about
0.9x10$ between the Chang’E-1 and SELENE reference
frames.
Ping et al. (2009) used more than 3 million topographic
measurements collected by the Chang’E-1 laser altimeter to
produce an accurate global lunar topographic model named
CLTM-s0l. A detailed comparison between CLTM-s01 and
other lunar topographic data, including the Clementine LiDAR
data and the ULCN 2005 was presented. Clementine LiDAR
has 72548 valid laser points, which is less than 296 of the
Chang’E-1 Laser Altimeter data. Clementine LiDAR data
didn’t cover the whole Moon. There are some areas with no or
sparse data, especially in the Polar regions. ULCN 2005
combines all the historical stereo photos (e.g. Apollo photos,
Clementine images), with the interpolated resolution of about
6.8 km and the elevation accuracy of laser measurements
approximate 100m, respectively. For the CLTM-s01 model, the
resolution is about 7 km and vertical accuracy is about 31 m,
respectively. The comparative analysis revealed that over the
large Maria regions on the near-side of the Moon, the
differences are very little within 200 m, however, over the far-
side of the Moon these differences are quite large. The
comparative results show that Chang'E-1 laser altimeter model
is an improvement of earlier models, including the Clementine
model and ULCN 2005, not only in data coverage and range
measuring accuracy, but also in spatial resolution.
Li et al.(2010) compared the DEMs (digital elevation model)
generated from the Chang’E-1 data with those from the
SELENE data using a wash-off relief map of the middle and
low latitude. The results show an identical trend with similar
data precision and spatial resolution. Li et al. (2010) also
examined the differences of the highest and lowest points
displayed in Chang'E-1 DEM and the SELENE DEM. For the
highest point there is only subtle difference between the two
DEMs, the point in SELENE DEM was about 100 m higher
than the similar point on the Chang'E-1 DEM model. However,
the plane position difference is up to 5.38 km for the lowest.
The lowest point in Chang'E-1 DEM model was over 100 m
higher than the lowest point in SELENE DEM model.
For surface comparison or matching between different
topographic data sets, vast of efforts have been performed in the
past. Williams (1999) studied the registration of three
dimensional data sets with rigid motions. The registration
process is comprised of two steps: correspondence selection and
motion estimation. Besl and Mckay (1992) developed an ICP
(Iterative Closest Point) algorithm for surface matching. The
basic theory of ICP is based on the search of pairs of the nearest
points in the two sets, estimating the rigid transformation, and
iteratively refining the transformation by repeatedly generating
pairs of corresponding points on the two sets by minimizing an
error metric. However, this algorithm required a lot of
calculation due to the exhaustive search of the nearest point.
Other researches proposed improved and accelerated algorithm
based on the ICP method (Park and Subbarao, 2003). Dijkman
and van den Heuvel (2002) presented a semi-automatic
registration method based on the Least Square Matching
method. The registration is performed using the parameters of
the models measured in different scans. Gruen and Akca (2005)
described an automatic method for surface registration using
template shaped targets. In this algorithm, seven parameters
including three transformations, three rotations, and one scale
factor could be obtained synchronously.
For the comparative analysis of lunar topographic models
derived from different sources, only simple and straightforward
methods were used in the past, and the comparisons were
mostly focused on the global scale for the whole Moon. This
research presents a detailed comparison of different lunar
topographic models in specific local regions based on a strict
least squares matching method.
3. COMPARISON OF LUNAR TOPOGRPHIC MODELS
DERIVED FROM CHANG’E-1 AND SELENE DATA
3.1 Overview of the Approach
Different lunar topographic models derived from the Chang’E-1
and SELENE laser altimeter data are used for comparative
analysis in this research. The framework of the comparative
analysis approach is illustrated in Figure 1. The least square
adjustment model integrates topographic data derived from the
Chang’E-1 and SELENE data using several conjugate points
through a strict mathematic model. Conjugate points were
carefully identified manually, which are obvious terrain features
(e.g., mount peaks or centers of craters) and evenly distributed
in the study region. After the least squares matching, seven
parameters (one scale factor, three transformations, and three
rotations) can be obtained. Finally, the detailed comparative
analyses between these two data sets are performed.
Topographic Models Topographic Models
from Chang’E-1 Data form SELENE Data
i i
Conjugate Points
i
Least Squares Matching
:
Seven Transformation Parameters
(One scale factor, three transformations,
three rotations)
| |
Chang’E-1Topographic Models
after Transformation
; |
Detailed Comparison Analysis
Figure 1. Framework of the least square adjustment approach
for the Chang’E-1 and SELENE laser altimeter data
3.2 Surface Matching Based on Least Squares Method
Due to the different times and the different sensors at different
positions, from which Chang’E-1 laser altimetry data and
SELENE laser altimetry data were obtained, the inconsistencies
must exist between the two data sources. Assume s (x, y, z) and
f (x, y, z) are conjugate regions of the Moon, from which
Chang'E-l data (search surface) and SELENE data (template
297
