Lunar and Planetary Science XXXIV (2003) 2084.pdf

TEN-METER SCALE TOPOGRAPHY AND ROUGHNESS OF EXPLORATION ROVERS LANDING SITES AND POLAR REGIONS. Anton B. Ivanov, MS168-414, Jet Propulsion Laboratory, Caltech, Pasadena, CA, 91109, USA, anton.ivanov@jpl..gov.

Introduction We have presented topography calculated using MOC Red The Mars Orbiter Camera (MOC) has been operating Wide Angle camera in [2], where we have proved validity of on board of the (MGS) spacecraft the proposed algorithm and in this work we concentrate on the since 1998. It consists of three cameras - Red and Blue Narrow angle camera imaging. In the MGS Extended mission Wide Angle cameras (FOV=140 deg.) and Narrow Angle phase MOC has targeted numerous sites for off-nadir imaging camera (FOV=0.44 deg.). The Wide Angle camera allows by its narrow-angle camera. First set of data considered here surface resolution down to 230 m/pixel and the Narrow Angle consists of images taken during specific targeted observations camera - down to 1.5 m/pixel. This work is a continuation (``ROTO maneuvers'' ) for MER landing site observations. of the project, which we have reported previously [2]. Since The second group of images comes from the polar region then we have refined and improved our stereo correlation of Mars, taken while MGS spacecraft was in the ``Relay-16'' algorithm and have processed many more stereo pairs. We will mode. discuss results of our stereo pair analysis located in the Mars MER landing sites Over the course of Mapping and Ex- Exploration rovers (MER) landing sites and address feasibility tended phases of the MGS mission MOC camera has been of recovering topography from stereo pairs (especially in the targeted to take stereo images of selected landing sites. Land- polar regions), taken during MGS ``Relay-16'' mode. ing site targets included Viking Landers, , Method Pathfinder and future candidate sites for Mars Exploration The basis for stereo image processing described in this Rovers (MER). The best stereo pairs were obtained during work are the image correlation tools developed as a part of the the Extended mission, when MGS spacecraft was pointed VICAR (Video Image Communication And Retrieval) soft- off-nadir to obtain high quality stereo pairs. ware suite at the Multimission Image Processing Laboratory Here we present an example stereo pair (E02-00665 / (MIPL) at JPL. VICAR has been developed since 1966 to E03-01453) results. This stereo pair is of very high quality digitally process multi-dimensional imaging data. and illustrates common problems that we have encountered. We employed VICAR tools used for geometric rectifi- We have collected more than 1 million tiepoints on two cation (GEOMV) and automatic tiepoint matching (program images. Fig. 1 shows the actual image E02-01453 and shaded TRACKER3) with a properly calibrated camera model (SPICE relief map, constructed from the retrieved topography. All I-kernel, [3]) for MOC Wide and Narrow angle cameras in- topographic features can be clearly identified in shaded relief strument. The basis for robust recovery of tiepoints from two map. However, there are some small scale undulations in images is the Gruen correlation algorithm [1], which has been the image, which are believed to be due to very small scale implemented in VICAR [5]. oscillations of the spacecraft. The cause of these oscillations TRACKER3 routine provides the most useful interface for is unknown, but one of the possible explanations is the wobble the feature tracking and automatic tiepoint identification for of the solar panels. We are currently developing algorithms to stereo processing. This program takes two images as input and remove this noise. This will enable us to analyze roughness automatically finds all tiepoints in the images with accuracy of landing sites at 10m horizontal scale and compare with down to about 0.2 of a pixel. The tiepoints can be used for 300m horizontal scale MOLA topography. These results are either referencing target image to the reference image, or for consistent with analysis of the same stereo pair performed by detecting change between time separated image sequence. Kirk et. al [4]. Recovery of topography information from stereo pairs South polar region During the extended mission phase the followed standard processing scheme : MGS spacecraft has endured so called ``Relay-16'' mode in order to save fuel and prolong operations in Martian orbit. 1. Perform radiometric calibration of the images (using ◦ ISIS tools for MOC processing) The spacecraft is pointed 16 off the nadir orientation. Con- sequently, all MOC images taken during this period of time 2. Perform rectification of the images (project into the have approximately 18◦ emission angle. Although, this angle same reference frame) using the best known camera is smaller than the optimal separation angle for stereo, we model. Pointing knowledge is very important at this decided to study feasibility of reconstructing topography from step. this set of images. The results of stereo processing are shown in Fig.2. This quality of stereo pair is very high and noise is 3. Use TRACKER3 routine for to automatically obtain minimal. Note that the same effects due to wobbling of the tiepoints from both images. spacecraft are very well pronounced. Summary 4. Use XOVER subroutine for all tiepoints to obtain location of intersecting rays projected from the camera. We have employed new stereo image processing capa- bilities developed for the VICAR image processing system. Data analysis Ten-meter scale topography, that is derived from Narrow An- Lunar and Planetary Science XXXIV (2003) 2084.pdf

TEN-METER SCALE TOPOGRAPHY OF MER LANDING SITES AND SOUTH POLAR REGION.

gle stereo pairs, is consistent with the MOLA topography. References Detailed analysis of the Narrow Angle images revealed previ- ously unknown oscillations of the spacecraft. Narrow Angle [1] GRUEN, A., AND BALTSAVIAS, E. Adaptive least processing results are consistent with the results obtained by squares correlation with geometrical constraints. In SPIE Kirk et al.[4]. We are currently working on algorithms to Computer Vision for Robots (Cannes, 1985), vol. 595. remove spacecraft noise and analyze surface roughness. Work is currently underway to analyze more stereo pairs as [2] IVANOV, A. B., AND LORRE, J. J. Analysis of Mars more data is being released and also introduce THEMIS visual Orbiter Camera Stereo Pairs. In 33rd LPSC, March 11- 15, 2002, Houston, Texas, abstract no.1845 (Mar. 2002), subsystem images into our processing. Data resulting from the vol. 33, pp. 1845--+. stereo pair analysis will be of high value for investigations of small scale topographic structures on Mars and future landing [3] KIRK, R. L., BECKER, T. L., ELIASON, E. M., ANDER- sites selection process. SON, J., AND SODERBLOM, L. A. Geometric Calibration Acknowledgments of the Mars Orbiter Cameras and Coalignment with the This work has been funded by the Interplanetary Network Mars Orbiter Laser Altimeter. In Lunar and Planetary and Information Systems Directorate (IPN-ISD) at the Jet Institute Conference (Mar. 2001), vol. 32, pp. 1863+. Propulsion Laboratory, through the Continuous Improvement Program (CIP). We would like to thank Larry Preheim for his [4] KIRK, R. L., HOWINGTON-KRAUS, E., AND ARCHI- support. We thank Jean Lorre for his advice on general stereo NAL, B. A. Topographic Analysis of Candidate Mars analysis and image matching algorithms. We also would Exploration Rover Landing Sites from MOC Narrow An- like to thank Randy Kirk and Boris Semenov for numerous gle Stereoimages. In 33rd LPSC, March 11-15, 2002, discussions on the subject and for allowing us to use the latest Houston, Texas, abstract no.1988 (Mar. 2002), vol. 33, MOC camera calibration models. pp. 1988--+.

[5] LORRE, J. J. Function minimization with partially correct data via simulated annealing. In Applications of Digital Image Processing (1988), vol. 595, SPIE, pp. 398--403.

Figure 2: Original MOC Narrow angle image M10/01422, Figure 1: Actual MOC image (E02-01453, top panel) and reconstructed shaded relief map and comparison with the corresponding shaded relief map of DEM reconstructed from MOLA data. This DEM is located on the south polar resid- Narrow Angle stereo pair (E02-00665 and E03-01453, middle ual ice cap. Thick line and blue dots are same as Fig. 1 for panel). This DEM is located in the crater. Compar- stereo pair (M10/01422 - E09/01101). This DEM is of very ison with the actual MOLA topography (thick black line) is high quality and only lacks data in the areas with no contrast, shown on the bottom panel. Blue dots represent topography where the image matcher was not able to nd good correla- sampled from reconstructed DEM. Note small undulations in tion. Note pronounced “waves” introduced by the wobbling the recovered topography due to ne scale oscillations of the of the spacecraft. spacecraft. Track sampled from the stereo DEM is offset slightly in height for clarity.