PE&RS July 2019 - page 488

TPS-Based DOM Registration Results
Due to the limited resolution of the SLDEM2015 compared with
the LROC NAC images, geometric inconsistencies still exist
between the seamlessly mosaicked adjacent subarea DOMs.
The upper two subfigures in Figure 7 show examples of the
geometric inconsistencies between Part 1 and 2 as well as Part
9 and 10 mosaicked DOMs, which were effectively reduced
by the image registration process based on the TPS model as
indicated in subfigures (c) and (d) in Figure 7. The registra-
tion was conducted subarea by subarea using the proce-
dure detailed in the section “TPS-Based Large-Area Image
Registration.”A quantitative evaluation of the registration re-
sults can also be realized by measuring the differences of the
check point pair coordinates between any two overlapping
subarea DOM mosaics. Parts 1 and 2 are taken as examples.
Ten evenly distributed check point pairs were automatically
matched in the overlapping region, as shown in Figure 8.
The deviations of the check point coordinates in the latitu-
dinal and longitudinal directions are listed in Table 4. After
the TPS-based registration, the planar deviations are reduced
to about 1 pixel, and the largest difference is 2.69 m, which is
no more than 2 pixels of the output DOM, reflecting a high-
precision registration.
Landing Area DOM and Potential Applications
After subarea planar block adjustment and TPS-based image
registration of the subarea DOMs, a seamless DOM mosaic of
the entire Chang’e-5 planned landing area was produced. The
generated radiometrically homogeneous and geometrically
seamless DOM mosaic is shown in Figure 9 (zoomed-out view).
This final DOM mosaic has the image size of 224 721 columns
and 44 945 rows with a ground sample distance of 1.5 m.
This high-resolution 10-gigapixel DOM has many potential
applications for detailed morphological and geological stud-
ies. For example, using the high-resolution map, craters can
be precisely measured to determine the age of surface units
(Michael and Neukum 2010); small, particularly fresh craters
(e.g., flat-bottomed, central-mound, and concentric craters)
can be used to estimate the depth of the lunar regolith (Bart et
al. 2011; Di et al. 2016); rocks/boulders on the ejecta of a cra-
ter can be identified and the spatial density used to estimate
the formation time of the crater (Li et al. 2017); and so on.
More important, distribution pattern analyses of crater rays,
crater chains, and boulders are significant in helping identify
source locations of the exposed features (e.g., rock and soil
samples to be collected by the lander), which will directly
contribute to the major scientific objective of the sample
return mission.
(a) (b) (c)
(d) (e) (f)
Figure 6. The geometric differences between NAC DOMs and SLDEM2015 before (upper three) and after (lower three) the
planar block adjustment. The NAC DOMs in subfigures (a) and (d) are part of m1221740903r in Part 1, (b) and (e) are part of
m1175809506l in Part 5, and (c) and (f) are part of m1145135367l in Part 10.
Table 3. Image numbers, control point, and tie point numbers and control point and tie point precisions of the overall block
adjustment using all the selected NAC images involved in the subarea block adjustment. The unit for precision assessment is an
NAC image pixel assumed to be 1.5 m.
Image
Number
Control
Point
number
Tie
Point
Number
RMS Errors of Control
Points (Pixel)
Maximum Errors of
Control Points (Pixel)
RMS Errors of
Tie Point (Pixel)
Maximum Errors of
Tie Point (Pixel)
x y xy x y xy x y xy x y xy
750 95 437 999 23.69 26.13 35.27 49.87 78.93 93.37 0.27 0.37 0.46 0.94 5.62 5.7
488 July 2019 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING