Exploratory Analysis of Zero-Shot Lunar Crater Detection using SAM for Chandrayaan-2 Images

This paper explores the application of a zero-shot lunar crater detection method based on the Segment Anything Model (SAM) for high-resolution Digital Terrain Model (DTM) images from ISRO’s Chandrayaan-II mission. A pre-trained transformer-based instance segmentation model, SAM, is used to segment and identify lunar craters without requiring training data. The proposed method is capable of detecting craters in varying crater sizes and shapes, highlighting its potential for generalizable planetary feature detection. This work investigates the applicability of general-purpose deep learning models in planetary exploration and offers insights for efficient lunar surface analysis.