Extremely dark (low reflectance) materials called dark mantle deposits (DMDs) are known to exist on the Moon’s surface. These are thought to be mantle materials which have been brought up the Moon’s surface by magma eruptions and then deposited. The distribution and chemical composition of DMDs therefore carry important information that can teach us about the Moon’s evolutionary process. However, extremely low reflectance of DMDs gives false positive when we detect dark regions in exploration data including shadows .

JJLPEDA and the National Institute of Advanced Industrial Science and Technology (AIST) are investigating methods to automatically detect shadows from images of the Moon’s surface. As training data, we use results determined by humans as shadows in observation data from the lunar orbit “KAGUYA”. Using deep learning methods utilized in image recognition, we apply this training data to observation data under various conditions, from which we evaluate the results and improve our methods.

Investigating these methods will likely make it possible to automatically exclude shadowed regions and identify hard-to-find dark mantle deposits (DMDs) scattered about on the Moon’s surface when conducting scientific research using vast amounts of lunar observation data.