[论文解读] Deep learning universal crater detection using Segment Anything Model (SAM)

Iraklis Giannakis, Anshuman Bhardwaj+2 more|arXiv (Cornell University)|Apr 16, 2023

Astro and Planetary Science被引用 14

一句话总结

The paper proposes a universal crater detection pipeline that uses the Segment Anything Model (SAM) to segment crater-like masks across diverse planetary data, followed by circular-elliptical filtering and ellipse fitting to estimate crater location and size without task-specific training.

ABSTRACT

Craters are amongst the most important morphological features in planetary exploration. To that extent, detecting, mapping and counting craters is a mainstream process in planetary science, done primarily manually, which is a very laborious and time-consuming process. Recently, machine learning (ML) and computer vision have been successfully applied for both detecting craters and estimating their size. Existing ML approaches for automated crater detection have been trained in specific types of data e.g. digital elevation model (DEM), images and associated metadata for orbiters such as the Lunar Reconnaissance Orbiter Camera (LROC) etc.. Due to that, each of the resulting ML schemes is applicable and reliable only to the type of data used during the training process. Data from different sources, angles and setups can compromise the reliability of these ML schemes. In this paper we present a universal crater detection scheme that is based on the recently proposed Segment Anything Model (SAM) from META AI. SAM is a prompt-able segmentation system with zero-shot generalization to unfamiliar objects and images without the need for additional training. Using SAM we can successfully identify crater-looking objects in any type of data (e,g, raw satellite images Level-1 and 2 products, DEMs etc.) for different setups (e.g. Lunar, Mars) and different capturing angles. Moreover, using shape indexes, we only keep the segmentation masks of crater-like features. These masks are subsequently fitted with an ellipse, recovering both the location and the size/geometry of the detected craters.

研究动机与目标

• Motivate a universal, data-agnostic crater detection method that works across celestial bodies and data types.
• Leverage SAM for prompt-able segmentation to detect crater-like features without fine-tuning.
• Filter segmentation masks by geometric shape and fit ellipses to recover crater sizes and positions.

提出的方法

• Apply SAM to segment masks on any input planetary image, DEM, or multispectral data.
• Filter masks by circularity and ellipticity using m and n indices and a fitted ellipse.
• Apply Canny edges to remaining masks and fit circles/ellipses to edges to estimate crater geometry (a, b, center).
• Post-process to remove duplicates and false positives.

Figure 1: A) Mars Express HRSC natural colour image of Ocrus Patera B) Segmentation of the input image using Segment Anything Model (SAM) [ 21 ] .

实验结果

研究问题

• RQ1 Can SAM provide zero-shot, universal crater segmentation across diverse datasets without task-specific training?
• RQ2 How effective are circular-elliptical shape indices in isolating crater-like features from SAM masks?
• RQ3 What is the accuracy of crater center and size estimation after ellipse fitting across different data types and bodies?

主要发现

• SAM can identify crater-looking objects in various data types and setups without additional training.
• Mask filtering with circular-elliptical indices followed by ellipse fitting yields crater centers and axes, enabling size estimation.
• Case studies show successful crater detection on Mars, Moon, and Phobos data across different sensors and resolutions.
• Limitations include potential false positives from non-crater circular features and reliance on threshold tuning for indices.

Figure 2: A) The remaining segmentation masks from Fig. 1 after filtering out the non-circular/elliptical classes using geometrical indexes. B) Canny filter is applied to each one of the remaining masks, and the edges are fitted with circles and ellipses.