Computer-Assisted Detection of Collapse Pits in LROC NAC Images
Abstract
Pits in mare basalts and impact melt deposits provide unique environments for human shelters and preservation of geologic information. Due to their steep walls, pits are most distinguishable when the Sun is high (pit walls are casting shadows and impact crater walls are not). Because of the large number of NAC images acquired every day (>350), each typically with 5000 samples and 52,224 lines, it is not feasible to carefully search each image manually, so we developed a shadow detection algorithm (Pitscan) which analyzes an image in thirty seconds. It locates blocks of pixels that are below a digital number (DN) cutoff value, indicating that the block of pixels is "in shadow", and then runs a DN profile in the direction of solar lighting, comparing average DN values of the up-Sun and down-Sun sides. If the up-Sun average DN is higher than the down-Sun average, the shadow is assumed to be from a positive relief feature, and ignored. Otherwise, Pitscan saves a 200 x 200 pixel sub-image for later manual review. The algorithm currently generates ~150 false positives for each successful pit identification. This number would be unacceptable for an algorithm designed to catalog a common feature, but since the logic is merely intended to assist humans in locating an unusual type of feature, the false alarm rate is acceptable, and the current version allows a human to effectively check 10,000 NAC images for pits (over 2500 gigapixels) per hour. The false negative rate is not yet known, however Pitscan detected every pit in a test on a small subset of the images known to contain pits. Pitscan is only effective when the Sun is within 50° of the zenith. When the Sun is closer to the horizon crater walls often cast shadows, resulting in unacceptable numbers of false positives. Due to the Sun angle limit, only regions within 50° latitude of the equator are searchable. To date, 25.42% of the Moon has been imaged within this constraint. Early versions of Pitscan found more than 150 small (average diameter 15m) pits in impact melt deposits of Copernican craters [1]. More recently, improvements to the algorithm revealed two new large mare pits, similar to the three pits discovered in Kaguya images [2]. One is in Schlüter crater, a mare-filled crater near Orientale basin, with a 20 x 40m opening, approximately 60 m deep. The second new pit is in Lacus Mortis (44.96°N, 25.61°E) in a tectonically complex region west of Burg crater, This pit is the largest mare pit found to date, with an opening approximately 100 x 150 m, and a floor more than 90 m below the surrounding terrain. Most interesting from an exploration point of view is the fact that the east wall appears to have collapsed, leaving a relatively smooth ~22° slope from the surrounding mare down to the pit floor. Computer-assisted feature detection is an effective method of locating rare features in the extremely large high-resolution NAC dataset. Pitscan enabled the discovery of unknown collapse pits both in the mare and highlands. These pits are an important resource for future surface exploration, both by providing access to pristine cross-sections of the near-surface and by providing radiation and micrometorite shielding for human outposts. [1] Wagner, R.V. et al. (2012), LPSC XLIII, #2266 [2] Haruyama, J. et al. (2010), LPSC XLI, #1285
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2012
- Bibcode:
- 2012AGUFM.P53A2042W
- Keywords:
-
- 1914 INFORMATICS / Data mining;
- 5464 PLANETARY SCIENCES: SOLID SURFACE PLANETS / Remote sensing;
- 6250 PLANETARY SCIENCES: SOLAR SYSTEM OBJECTS / Moon