36th DPS Meeting, 8-12 November 2004
Session 47 Mars Surface and Dust
Oral, Friday, November 12, 2004, 3:30-5:00pm, Lewis

[Previous] | [Session 47] | [Next]

[47.05] A Comparison of Automated and Manual Crater Counting Techniques in Images of Elysium Planitia.

C. S. Plesko (UCSC, LANL), S. P. Brumby (LANL), E. Asphaug (UCSC)

Surveys of impact craters yield a wealth of information about Martian geology, providing clues to the relative age, local composition and erosional history of the surface. Martian craters are also of intrinsic geophysical interest, given that the processes by which they form are not entirely clear, especially cratering in ice-saturated regoliths (Plesko et al. 2004, AGU) which appear common on Mars (Squyres and Carr 1986).

However, the deluge of data over the last decade has made comprehensive manual counts prohibitive, except in select regions. Given that most small craters on Mars may be secondaries from a few very recent impact events (McEwen et al. in press, Icarus 2004), using select regions for age dating introduces considerable potential for sampling error.

Automation is thus an enabling planetary science technology. In contrast to machine counts, human counts are prone to human decision making, thus not intrinsically reproducible. One can address human "noise" by averaging over many human counts (Kanefsky et al. 2001), but this multiplies the already laborious effort required. In this study, we test automated crater counting algorithms developed with the Los Alamos National Laboratory genetic programming suite GENIE (Harvey et al., 2002) against established manual counts of craters in Elysium Planitia, using MOC and THEMIS data. We intend to establish the validity of our method against well-regarded hand counts (Hartmann et al. 2000), and then apply it generally to larger regions of Mars.

Previous work on automated crater counting used customized algorithms (Bierhaus et al. 2003, Burl et al.. 2001). Algorithms generated by genetic programming have the advantage of requiring little time or user effort to generate, so it is relatively easy to generate a suite of algorithms for varied terrain types, or to compare results from multiple algorithms for improved accuracy (Plesko et al. 2003).

[Previous] | [Session 47] | [Next]

Bulletin of the American Astronomical Society, 36 #4
© 2004. The American Astronomical Soceity.