Mapping Lunar Swirls with Machine Learning: The Application of Unsupervised and Supervised Image Classification Algorithms in Reiner Gamma and Mare Ingenii

Abstract Lunar swirls are recognized as broad, bright albedo features in various regions of the Moon. These often separated by dark off-swirl lanes or terminate against background, such lunar maria. Prior mapping has been done primarily contrast, which is prone to subjectivity. Closer examination on-swirl areas shows that they not uniform, making boundary between on- and difficult map with certainty. We have applied machine learning techniques address these issues identifying number swirl units then them based on actual reflectance, I/F data. Using LROC NAC paired stereo images converted reflectance at a range incidence angles, we both unsupervised K-means clustering supervised Maximum Likelihood Classification algorithms classify portions Reiner Gamma Mare Ingenii. Results show classification maps reasonable match representative albedos for two study regions. A third transitionary unit, termed diffuse-swirl, present cumulative distribution plots values. Overall, find use provides independent confirmation location their interrelation. More importantly, remove subjectivity using quantitative information. The data statistics generated from also value future studies placing limits categorizing different