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
3
x10* Cross-Track Offset vs. SCS Temperature
+ data
ese lite d Curve
. 4
+. *
Cross-Track Offset (degrees)
2 4 6 8 19 12 14 16 18 20 22
SCS Temperature (°C)
Figure 7. Best-fit curves for the cross-track offset between the
NAC-L to NAC-R. Red line is a best-fit curve (2"* order Fourier
series). Y-axis units are 10? degrees.
The relative offset correction was applied using a second order
Fourier series with the following general form:
f(x) 2 ag * a, cos(xw) * b, sin(xw) (1)
*à» cos(2xw) * b5sin(2xw)
where ay, a;, 42, b;, by, w= derived constants
3.2 WAC Distortion Model and Pointing Correction
Unlike the NAC, which can be directly tied to human artefacts
on the lunar surface, the WAC in-flight geometric calibration
was based on registration with map projected NAC images that
have been processed with the latest calibration updates (see
section 3.1). To limit topographic variation images were
selected over the relatively flat Mare Imbrium region (Figure 8).
In all, 729 WAC images (96 monochrome, 633 color) were co-
registered to 1,212 NAC observations, thus collecting over 6.5
million data points for deriving improved pointing and camera
distortion models.
Figure 8. Distribution of NAC images used for "ground truth"
in calibrating the WAC instrument. The NAC observations are
overlaid on the WAC derived topographic model, GLD100
[Scholten et al., 2012].
3.2.1 Image Registration: To provide a "ground truth,"
NAC observation acquired over the Mare Imbrium region were
map projected at 25 meters per pixel. WAC images, acquired
under similar lighting conditions that overlapped these NAC
images were oversampled and projected at the same pixel scale.
Both images were map projected using the highest resolution
digital terrain model, GLD100 [Scholten et al., 2012], and the
latest ephemeris derived from radiometric data and altimetric
crossovers [Mazarico et al., 2011].
Using a pattern-matching algorithm found in Integrated
Software for Imagers and Spectrometers (ISIS) package
compiled by the Astrogeology Research Group of the United
States Geologic Survey (USGS) [Anderson et al., 2002] WAC
images were registered to the NAC images (“Truth”) covering
the same geographical region. Specifically, a region or pattern
chip (in this case a 20 sample by 20 line region) was extracted
from the map projected WAC image. The pattern chip was used
to identify a matching region in the search chip. The search chip
was a larger area found in the NAC image (in this case a 80
sample by 80 line region). The pattern chip was scanned across
and compared to sub regions of the search chip. A goodness of
fit (GOF) was calculated for each point in the search chip by
computing:
GOF = | cov( pattern,subregion) |
| var( pattern) x var(subregion )
(2)
cov — covariance function
var = variance function
pattern = n x m pattern chip
subregion =n x m sub-region of the search region
where
Upon walking the pattern chip through the search chip and
calculating the corresponding goodness value for cach point, the
pixel with the highest correlation value represents the position
in the search chip that best matches the pattern chip. This result,
however, was only good to one pixel accuracy. In most cases,
the point may lie between a set of pixels. To match at the sub-
pixel level, a surface model was generated over the matrix of
GOF values. The maximum point of this surface estimates the
true registration position of the pattern chip in the search chip.
This process was repeated at multiple locations over each map-
projected pair. Due to the large number of points (^ 6.5 M), any
mis-registration has very little impact on the distortion
modelling as a whole.
The co-registration information was passed to a second ISIS
program that identifies the location of the distorted and
undistorted, or corrected, pixel. Due to the wide angle optics
present on the WAC, images are distorted resulting in the
location of pixels altered from their ideal point on the CCD.
This effect increases the further the pixel is from the boresight,
or where the optical axis of the lens intersects the focal plane.
The program reads in the registration information and identifies
where on the WAC focal plane the distorted (from the WAC
image) and corrected (from the NAC image) pixel is located
(Figure 9). This information can then be used to identify the
distortion present in the WAC optics.
482