Automated Granulometric Analysis of High Spatial Resolution Images for Application to In Situ Martian Sedimentology Studies, 15-R8191
Victoria E. Hamilton
Inclusive Dates: 10/18/10 – 02/18/11
Background — Sediments form by the mechanical and/or chemical breakdown of their parent rocks. Physical properties of sediments, such as their particle size distribution and particle morphology, are indicators of mineral composition, distance transported and depositional environment. A geologist can make the observations needed to determine the above characteristics by simple visual inspection of sediments (or sedimentary rocks) in the field, but this is clearly not possible on Mars or other planetary surfaces. An Earth-bound geologist must rely on photographic images of the sediments and rocks to determine particle size and distribution (granulometry) characteristics. Published granulometric analysis of Martian sediments using images acquired by the Mars Exploration Rovers (MER) Spirit and Opportunity have relied on painstaking manual analysis of a very small subset of all available data. The primary objective of this effort was to test whether commercial, automated particle-counting software (designed for mining applications) may enable more rapid, comprehensive and quantitative analyses of granulometry from images. If so, such software could allow the analysis of a larger fraction of all data, leading to better statistics, and thus better constraints on the processes that produce and modify sediments. The immediate application is to granulometric studies of Mars during the upcoming NASA Mars Science Laboratory (MSL) mission, which has multiple cameras designed to image the surface of Mars at spatial resolutions down to ˜20-30 μm.
Approach — SwRI researchers conducted lab work to produce and image control samples as well as field work to collect images of natural sedimentary materials under a variety of illumination conditions and from a variety of natural geologic settings. Researchers also selected images of Martian sediments from recent rover missions for re-analysis. A subset of these lab, field and Mars images was subsequently processed using an automated image processing software package to determine how well particle shape and particle size distributions can be retrieved, depending on the amount of user intervention and the image quality.
Accomplishments — The following goals were accomplished:
Collected a high-resolution data set of sediments from different geologic environments under varying observation conditions.
Demonstrated the use of automated granulometry software.
Determined optimal observations for automated analysis.
Demonstrated automated analysis on images of Martian sediments.
Despite finding that automated analyses needed a greater amount of user intervention than initially estimated (largely due to user inexperience), it was found that the amount of time required to analyze an image could be decreased by up to 50 percent over manual methods, enabling a doubling of productivity and data analyzed. The influence of image processing on results was not investigated, but based on the results, some image processing (e.g., reduction of contrast, image filtering) could lead to additional increases in efficiency of the automated particle analysis process.