AUTOMATIC IDENTIFICATION OF CHANGES ON THE LUNAR SURFACE

Since June 2009, the Lunar Reconnaissance Orbiter (LRO) has maintained a stable polar orbit enabling the twin Narrow Angle Cameras (NACs) to acquire high-resolution observations of the lunar surface (pixel scale of 0.25 to 2 m/pixel). This orbital configuration facilitates occasional repeat coverage with nearly identical lighting geometries. These before and after observations, referred to in this study as temporal pairs, enable the identification of changes to the surface using a series of automated change detection techniques.

Initial manual inspection of temporal pairs resulted in the discovery of many changes across the lunar surface [1]. However, this manual process is time consuming (2-4 hours per temporal pair) and each analyst must apply their own judgment on whether they have discovered a real change or an artifact in the image pair. This manual approach resulted in the identification of 650 surface changes as well as 19 resolved craters.

Leveraging image processing techniques developed by the LROC team, we started automatically scanning and identifying temporal changes. The new automated algorithm locates changes based on reflectance variations and changes in surface texture. The program provides a list of potential new features for later manual inspection and classification (disturbance lacking resolvable crater or crater with a rim diameter of x meters). This new approach reduces the time to manually inspect a temporal pair by over a factor of 200 by providing cropped cutouts with the putative changes centered in a thumbnail.

The LROC NACs have already collected thousands of temporal pair observations and will continue to do so over the remaining extended mission. Highest fidelity change detection comes from temporal pairs with nearly identical lighting geometries (i.e. sub-solar points within 3°). In our initial scan of 1645 temporal pairs, nearly 8000 surface changes were identified along with over 50 new impact craters with diameters ranging from 1.7 m to 16 m. Continued scanning of temporal pairs will enable refinements to the current cratering rate, quantify the risk of small secondary impacts, and enhance our knowledge of the ejecta formation.

Reference: [1]: S.D. Thompson et al. (2014) Recent Impacts on the Moon, LPSC, #2769.