• Department
  • Courses
  • Sections
  • Projects
• Resources
  • Wiki
  • News
  • Forums
  • Surveys
  • Mailing Lists
• Help

• Projects
• CS
• Lunar Crater Detection & Recognition

Crater detection and avoidance is critical for spacecraft landings and mapping of lunar terrain. Resources for crater detection are limited and new approaches are always welcome. The Lunar Mapping and Modeling Portal (LMMP) team at JPL tasked our LCDR team to develop an algorithm that will automatically detect lunar craters using images from the Lunar Reconnaissance Orbiter Camera (LROC). Our approach to crater detection was to combine several algorithms and  achieve a high accuracy percentage. Approaches for detecting candidate craters include template matching, circle hough transform, and extraction of highlights and shadows. Accuracy is prioritized over performance, per request by JPL.

The team developed a SW system named Ringtoss. A user selects a region on the moon. Ringtoss then downloads an image from the LROC repository for the selected region of interest. This image is then processed by our three different detection methods. Each potential crater is added to a list, which undergoes further processing to remove duplicate detected craters. The combined candidate list is then provided to a machine learning algorithm for improved detection and recognition. Finally, depth and diameter are calculated for each detected crater. Results are provided in text output and stored in a DB as well.