HR: 1340h
AN: P23A-0184    [Abstracts]
TI: Machine Identification of Martian Craters Using Digital Elevation Data
AU: * Bue, B
EM: bbue@cs.purdue.edu
AF: Purdue Univ., Dept. of Computer Sci., West Lafayette, IN 47907 United States
AU: Stepinski, T F
EM: tom@lpi.usra.edu
AF: Lunar and Planetary Institute, 3600 Bay Area Blvd., Houston, TX 77058 United States
AB: Impact craters are among the most studied features on Martian surface. Their importance stems from the worth of information that a detailed analysis of their number and morphology can bring forth. Because building manually a comprehensive dataset of craters is a laborious process, there have been many previous attempts to develop an automatic, image-based crater identifier. The resulting identifiers suffer from low efficiency and remain in an experimental stage. We have developed a DEM-based, fully autonomous crater identifier that takes an arbitrarily large Martian site as an input and produces a catalog of craters as an output. Using the topography data we calculate a topographic profile curvature that is thresholded to produce a binary image, pixels having maximum negative curvature are labeled black, the remaining pixels are labeled white. The black pixels outline craters because crater rims are the most convex feature in the Martian landscape. The Hough Transform (HT) is used for an actual recognition of craters in the binary image. The image is first segmented (without cutting the craters) into a large number of smaller images using the ``flood" algorithm that identifies basins. This segmentation makes possible the application of highly inefficient HT to large sites. The identifier is applied to a 10⁶ km² site located around the Herschel crater. According to the Barlow catalog, this site contains 485 craters >5 km. Our identifier finds 1099 segments, 628 of them are classified as craters >5 km. Overall, there is an excellent agreement between the two catalogs, although the specific statistics are still pending due to the difficulties in recalculating the MDIM 1 coordinate system used in the Barlow catalog to the MDIM 2.1 coordinate system used by our identifier.
DE: 5494 Instruments and techniques
DE: 6225 Mars
SC: Planetary Sciences [P]
MN: Fall Meeting 2005