Assessing Foundation Models for Sea Ice Type Segmentation in Sentinel-1 SAR Imagery

2025-03-28

AI Summary

This study addresses the fine-grained sea-ice type segmentation task using Sentinel-1 SAR imagery, tackling challenges including strong speckle noise, highly variable backscatter signatures, and poor seasonal and spatial generalizability. Methodologically, we propose the first systematic foundation model selection framework and a dedicated evaluation benchmark tailored to polar SAR data, introducing polar-specific performance metrics. Leveraging dual-polarization SAR imagery, we adopt a transfer-fine-tuning strategy and comprehensively evaluate ten remote-sensing foundation models using F1-score, IoU, and cross-seasonal/cross-regional generalization capability. Results show Prithvi-600M achieves top performance, followed by CROMA. The study demonstrates that foundation models enable high-accuracy sea-ice segmentation even under few-shot settings, while revealing a significant domain shift bottleneck of existing pre-trained models in the SAR domain. This work establishes a novel paradigm for intelligent interpretation of polar remote sensing data.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Segmentation
• Intelligent Robots: Other Foundations and Applications

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Accurate segmentation of sea ice types is essential for mapping and operational forecasting of sea ice conditions for safe navigation and resource extraction in ice-covered waters, as well as for understanding polar climate processes. While deep learning methods have shown promise in automating sea ice segmentation, they often rely on extensive labeled datasets which require expert knowledge and are time-consuming to create. Recently, foundation models (FMs) have shown excellent results for segmenting remote sensing images by utilizing pre-training on large datasets using self-supervised techniques. However, their effectiveness for sea ice segmentation remains unexplored, especially given sea ice's complex structures, seasonal changes, and unique spectral signatures, as well as peculiar Synthetic Aperture Radar (SAR) imagery characteristics including banding and scalloping noise, and varying ice backscatter characteristics, which are often missing in standard remote sensing pre-training datasets. In particular, SAR images over polar regions are acquired using different modes than used to capture the images at lower latitudes by the same sensors that form training datasets for FMs. This study evaluates ten remote sensing FMs for sea ice type segmentation using Sentinel-1 SAR imagery, focusing on their seasonal and spatial generalization. Among the selected models, Prithvi-600M outperforms the baseline models, while CROMA achieves a very similar performance in F1-score. Our contributions include offering a systematic methodology for selecting FMs for sea ice data analysis, a comprehensive benchmarking study on performances of FMs for sea ice segmentation with tailored performance metrics, and insights into existing gaps and future directions for improving domain-specific models in polar applications using SAR data.

Problem

• Evaluate foundation models for sea ice type segmentation in SAR imagery
• Assess model performance under seasonal and spatial variations
• Address gaps in domain-specific models for polar SAR applications

Innovation

• Utilizes foundation models for sea ice segmentation
• Evaluates models using Sentinel-1 SAR imagery
• Focuses on seasonal and spatial generalization