1001
The International Archives of the Photogrammetrv. Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
EO polynomials of the reference strip starts from zero. For the
other non-reference strips, it starts from the offsets. The initial
value of the EO polynomial coefficients can be figured out by
least-squares fitting of the line-by-line telemetry EO data.
3. BUNDLE ADJUSTMENT OF HiRISE STEREO IMAGES
This bundle adjustment integrates HiRISE EO parameters and tie
points on HiRISE stereo images. EO parameters have been
retrieved from the SPICE kernel and stored line by line. A
hierarchical-matching algorithm has been developed at the
Mapping and GIS Laboratory at The Ohio State University (OSU)
for mapping purposes. The tie points are automatically selected
from the matched terrain points on stereo images to make sure
they are evenly distributed. These tie points will be included in
the bundle adjustment as measurements after the interior
orientation procedure.
Different types of measurements will form observation equations
separately. For image tie points, they are related to the ground
coordinates and EO parameters via the collinear equations:
x | J ll (X t -X c l ) + af l2 <r i -r i ) + a , i3 {Z,-Z c l ) Q
' a i 3l (X i -X c i ) + a i n (Y i -Y c i ) + a‘ i3 (Z i -Z c i ) (5)
{ . a' 2 \(X j -X c J + a' 22 (Yj - YT) + a‘ 23 (Z i -Z c ¡) Q
a' 31 (X t - X%) + a' 32 (i' - 7 C ) + a' 33 (Z. - Z c ,)
where X i is the along-track coordinate of the detector on the
focal plane of the f h point which can be calculated using
HiRISE interior orientations; y i is the corresponding
cross-track image coordinate; X t ,Y j ,Z j are the ground
coordinates of the i' h point; X°¡,Y C j9 Z c t are the position
of the perspective center of the sensor at the i' h point;
a' n ...a l 33 are the elements of the rotation matrix formed by
the sensor pointing angles, and f is the focal length of the
sensor.
The EO parameters in Equation 4 determined through sensor
modeling are treated as weighted parameters to form a set of
pseudo observation equations, such that
pseudo observation + correction - parameter = 0 (6)
These equations are combined into the bundle adjustment system.
The adjustment is based on the least-squares principle
minimizing the sum of the weighted square of the residuals of the
observations and the pseudo observations. Since Equation 5 is
non-linear, linearization should be applied. Therefore the solution
must be iterative. The initial values of the EO polynomial
coefficients are from the least-squares fitting of telemetry EO
data. The initial ground positions of tie points are obtained
through an intersection using telemetry EO data.
In this bundle adjustment system, tie points are used as
measurements and telemetry EO parameters serve as pseudo
measurements; 36 EO polynomial coefficients and ground tie
point coordinates are unknowns. One tie point contributes to 4
measurements and 3 unknowns, therefore at least 36 tie points
are needed to make the problem solvable.
4. DATA PROCESSING RESULTS AT MER SITES
4.1 Bundle Adjustment without Ground Control
The MRO mapping data used in this study includes HiRISE
stereo images and MRO trajectory data including the positions of
the camera perspective center and pointing angles. Table 1
summarizes the main properties of the stereo images used in this
study.
Site Name
Victoria Crater
Husband Hill
Stereo
Left
Right
Left
Right
Acquisition Time
10/03
11/14
12/12
11/22
(M/D/Y)
/2006
/2006
/2006
/2006
Convergence Angle
9.8 degrees
19.8 degrees
Image Size
5000 by 4096
5000 by 4096
No. of Tie Points
136
178
Table 1. Sites studied in this research
Victoria Crater and Husband Hill, where the twin rovers in the
2003 Mars Exploration Rover (MER) mission are doing their
surface exploration, were selected as mapping areas in this study.
The planned comparison with ground mapping products and
further integration with ground data was the reason those sites
were chosen. Also, those two sites represent two typical
geological characteristics on the Martian surface. Husband Hill is
a hilly area with a lot of topographical variations while Victoria
Crater is 800 meters in diameter with a very smooth crater wall
and flat sandy surface outside the crater ring. MRO trajectory
data is extracted using Spacecraft, Planet, Instrument, C matrix