v i. iS | i
BUNDLE ADJUSTMENT OF MARS HIRISE ORBITER STEREO IMAGES BASED ON
THE RIGOROUS SENSOR MODEL
Yunhang Chen
Mapping and GIS Laboratory, Dept, of Civil and Env. Eng. & Geodetic Science
The Ohio State University
470 Hitchcock Hall, 2070 Neil Avenue, Columbus, OH 43210-1275
chen. 1256@osu.edu
KEY WORDS: Digital Photogrammetry, Sensor Models, Adjustment, Topographic Mapping, High Resolution Image, Extraterrestrial
ABSTRACT:
Launched in August 2005, the Mars Reconnaissance Orbiter (MRO)’s primary goal is to obtain detailed morphologic information on
the Martian surface using its onboard HiRISE sensor with a 0.3-meter ground resolution. However, inconsistencies between HiRISE
stereo images must be removed based on the rigorous sensor model for the best level of accuracy. The new rigorous sensor model is
developed specially for HiRISE due to its unusual complex camera design so that the mosaic of multiple image strips could be
processed simultaneously under a uniform sensor model. A bundle adjustment (BA) based on this rigorous sensor model is
implemented on two stereo pairs covering Husband Hill and Victoria Crater. After BA, the significantly reduced back projection
residuals indicate that the inconsistencies are removed and the refined exterior orientation parameters are ready for mapping the
Martian surface. Furthermore, 4 ground points measured from geo-registered Spirit rover images are incorporated into the BA as
ground control points. Reduced inconsistencies on ground features between orbital intersection and measurements from rover images
after BA indicate the mapping potential of orbital-ground integration. Small motions of the spacecraft around its nominal pointing,
called jitter, has been detected on HiRISE trajectory data. The topographic effect of jitter is evaluated in this paper. Though its effect
is insignificant when mapping small areas, further investigation needs to be performed so that the effect can be removed or reduced
when mapping large areas.
1. INTRODUCTION
The Mars Reconnaissance Orbiter (MRO) has collected mapping
data at an unprecedented level of accuracy on the Martian surface.
Its primary scientific objectives are to characterize the surface,
subsurface and atmosphere of Mars and to identify potential
landing sites for future missions (NASA/JPL). HiRISE, the linear
pushbroom image sensor onboard MRO, provides images with
up to 0.3-meter resolution. This ultra-high resolution makes great
progress possible in the derivation of topographic information on
the Martian surface.
In order to support the Opportunity rover operation in the 2003
Mars Exploration Rover (MER) mission, a DEM of Victoria
Crater was generated by USGS using HiRISE stereo images. A
program was used to transform input raw images by projecting
them into ground coordinates and then back into the image
coordinates of an idealized HiRISE camera that can be modeled
in the commercially available photogrammetry system SOCET
SET. Photogrammetric processing of the transformed images is
conducted using SOCET SET; the DEM is then generated based
on the output results (Kirk, et al., 2007). However, in order to
reach the best achievable mapping accuracy, inconsistencies
between HiRISE stereo pairs must be removed based on the
rigorous sensor model. Subsequently, this paper proposes a
bundle adjustment to remove such inconsistencies using the
rigorous HiRISE sensor model. Instead of using the commercial
photogrammetric software used in the method described above,
we are using our own code to incorporate the rigorous sensor
model. Therefore, our results are free of errors caused by
topography variation and can achieve the best level of accuracy
can be reached.
999