Scholarlink

Precise crater detection on the moon is important for geological interpretation, dating by stratigraphy, and mission planning for future lunar missions. Manual mapping of craters, while efficient, is labour-intensive and inadequate to manage the volume of data from present-day planetary missions. The present research introduces an automated deep learning process for crater detection using Chandrayaan-2 mission data. The model is developed on top of the YOLOv12 object detection structure, with the addition of a ResNet101 backbone, and is tested on two datasets: optical imagery from the Orbiter High Resolution Camera (OHRC) and Terrain Mapping Camera-2 (TMC-2) Digital Elevation Model (DEM) elevation data. OHRC data-trained model had a precision of 0.746 and a recall of 0.668, whereas TMC-2 DEM data-trained model had a precision of 0.809 and a recall of 0.519. The above results show that the framework is strong enough to catch crater features for diverse terrains and imaging modes. The approach shows strong potential for real-time, scalable lunar surface characterisation and supports the future of automated planetary mapping.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Lunar Crater Detection Using Chandrayaan-2 TMC-2 DEM and OHRC Imagery with Yolo V12

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mimansa Sinha,
  • Malaya Shekhar,
  • Md Arbab,
  • Sanchita Paul

摘要

Precise crater detection on the moon is important for geological interpretation, dating by stratigraphy, and mission planning for future lunar missions. Manual mapping of craters, while efficient, is labour-intensive and inadequate to manage the volume of data from present-day planetary missions. The present research introduces an automated deep learning process for crater detection using Chandrayaan-2 mission data. The model is developed on top of the YOLOv12 object detection structure, with the addition of a ResNet101 backbone, and is tested on two datasets: optical imagery from the Orbiter High Resolution Camera (OHRC) and Terrain Mapping Camera-2 (TMC-2) Digital Elevation Model (DEM) elevation data. OHRC data-trained model had a precision of 0.746 and a recall of 0.668, whereas TMC-2 DEM data-trained model had a precision of 0.809 and a recall of 0.519. The above results show that the framework is strong enough to catch crater features for diverse terrains and imaging modes. The approach shows strong potential for real-time, scalable lunar surface characterisation and supports the future of automated planetary mapping.