Earth trained image features help classify Titan radar terrains
CETUS: How Far Do Representations Trained on Earth Transfer to Cassini SAR of Titan?
Computer Vision and Pattern RecognitionMachine Learning
Summary
Cassini spacecraft took radar pictures of Titan, one of Saturn’s moons, showing different types of land like dunes and lakes. The authors tested if computer models trained on Earth images could recognize these land types on Titan’s radar pictures. They found that Earth-trained models generally performed better than traditional image measurements when identifying terrain, though results varied based on training and settings. This means some Earth imagery knowledge can transfer to helping map other planets, but more tuning is needed for best results.
What this means in practice
- •For planetary mapping teams: Improve automated classification of planetary surfaces from radar images using Earth-trained image encoders.
- •For remote sensing platform developers: Develop SAR image analysis tools that leverage pretrained Earth image models to speed terrain interpretation on other planets.
Authors
Kevin Lee
Abstract
Cassini synthetic aperture radar (SAR) images reveal the dunes, plains, and lake basins of Titan, providing an instance of representations learned from Earth imagery for planetary terrain classification. Cross-domain Evaluation of Earth-to-Titan Transfer Using SAR (CETUS) compares features from DINOv2, DOFA and CROMA with classical image measurements and features from an untrained vision transformer on the U.S. Geological Survey's Cassini SAR mosaic. The classifiers learn terrain labels from an expert geomorphological map and predict those labels in geographically separate Titan regions. Under logistic regression settings, pretrained encoders achieve higher mean macro F1 than the combined classical features. Encoder rankings change when feature scaling, optimization, and regularization change together. Further training on Titan improves DINOv2 performance, degrades DOFA performance, and leads to mixed results for CROMA under the tested settings. Architectural and input processing differences prevent these comparisons from isolating the effect of pretraining. Classifier fitting and performance on individual terrain classes matter when assessing representation transfer for planetary mapping. Since the map draws partly on the same radar observations, the scores measure agreement with expert interpretation.