HR: 0800h
AN: P31B-0432    [Abstracts]
TI: Rigorous Photogrammetric Processing of HiRISE Stereo Images for Topographic and Geomorphologic Analysis at MER landing sites
AU: * Li, R
EM: li.282@osu.edu
AF: The Ohio State University, 470 Hitchcock Hall, 2070 Neil Avenue, Columbus, OH 43210,
AU: Di, K
EM: di.2@osu.edu
AF: The Ohio State University, 470 Hitchcock Hall, 2070 Neil Avenue, Columbus, OH 43210,
AU: Hwangbo, J
EM: hwangbo.2@osu.edu
AF: The Ohio State University, 470 Hitchcock Hall, 2070 Neil Avenue, Columbus, OH 43210,
AU: Chen, Y
EM: chen.1256@osu.edu
AF: The Ohio State University, 470 Hitchcock Hall, 2070 Neil Avenue, Columbus, OH 43210,
AB: High-precision topographic information is critical to Mars surface exploration. Such information can be derived from both orbital and ground rover data. The availability of HiRISE stereo images makes a great progress in high resolution imaging and topographic and morphological information derivation. This presentation gives the necessary rigorous photogrammetric model for HiRISE stereo image processing and results achieved at the Mars Exploration Rover mission sites. HiRISE is a push-broom imaging sensor. For short segments of the orbital trajectory, 2nd-order polynomials can be used to model change in the exterior orientation (EO) parameters with respect to time. Since all of the 14 CCD lines (10 red, 2 blue-green and 2 NIR) share the same EO parameters at a specific time, only one set of polynomial parameters is needed to model the EO parameters for all the CCD arrays. The orbit's initial position and pointing angles are provided in the SPICE kernels. For any given ephemeris time, the EO parameters can be retrieved by interpolating the spacecraft's trajectory and pointing vectors. Based on the developed rigorous sensor model, we have developed a method for bundle adjustment (BA) of HiRISE stereo images that removes or reduces measurement inconsistency and improves mapping precision. We have also developed a hierarchical stereo matching process. Based on the original images, an image pyramid with 5 levels is constructed by sub-sampling of each previous level. Interest points are generated by Foerstner operator at every image scale. Matching starts from the images of the lowest resolution; results are transferred to the next higher level, with more interest points being extracted and matched. After matching the highest resolution images, evenly distributed matched interest points are selected as tie points between the stereo images. In the end, grid points (4-pixel spacing) are defined and matched to generate a DTM of the terrain. Using the HiRISE images TRA_000873_1780 and PSP_001414_1780 of Victoria Crater, 136 evenly distributed tie points and 135 check points were automatically selected for BA. Before BA, the mean residual in the along- track direction was 24.6 pixels with a standard deviation of 0.77 pixel, and the mean residual in the cross-track direction was 0.017 pixel with a standard deviation of 0.23 pixel. After BA, the mean residual in the along-track and cross-track directions are 0.30 pixel and 0.0047 pixel, respectively, with standard deviations of 0.37 pixel and 0.024 pixel. This improved precision ensures the high quality of topographic and morphological information to be derived. Currently, we are processing a stereo pair of HiRISE images that cover Husband Hill and the Home Plate area. The presentation should provide more detailed information and results.
DE: 0520 Data analysis: algorithms and implementation
DE: 0540 Image processing
DE: 6225 Mars
SC: Planetary Sciences [P]
MN: 2007 Fall Meeting