Clustering Guided Domain-Specific Pretrained Foundation Model Very High-Resolution Arctic Remote Sensing

2026-05-28

AI Summary

This study addresses the limited representational capacity of existing general-purpose remote sensing foundation models in high-resolution Arctic image analysis, primarily due to data distribution biases and scene redundancy. To overcome this, the authors propose a domain-specific self-supervised pretraining paradigm. They first apply affinity propagation clustering based on spectral and metadata features to curate a diverse Arctic dataset of 3 million image chips from an original corpus of 267 TB. Using this dataset, they pretrain a ViT-Large model with a fixed Masked Autoencoder (MAE) architecture to yield a specialized feature encoder tailored for Arctic scenes. Evaluated across four labeled benchmarks, the resulting model significantly enhances transfer performance, achieving average F1 score improvements of 5–8 percentage points over ImageNet-initialized ViT and at least 15 percentage points over Prithvi-EO-2.0, thereby demonstrating the efficacy and superiority of domain-specific self-supervised pretraining.

Abstract

This study introduces a novel Arctic-focused remote sensing foundation model (RSFM) by combining diversity-aware regional-scale image curation with masked autoencoder (MAE) self-supervised pretraining of a Vision Transformer (ViT) encoder for very-high-spatial-resolution (VHSR) satellite image analysis. Spectral and acquisition-metadata descriptors were used in a scalable affinity-propagation clustering workflow to select approximately 3 million chips from 267 TB of Vantor VHSR imagery This curation strategy was designed to reduce oversampling of visually repetitive or low-information areas while preserving broad scene diversity across the study domain. We pretrained a ViT-Large encoder on the curated corpus using a domain-adapted MAE reconstruction objective, producing Arctic-specific transformer weights for downstream feature mapping. The pretrained encoder was integrated into an existing location-aware detection and segmentation framework and evaluated across four hand-labeled Arctic datasets. Compared to ImageNet-initialized ViT-Large baseline, Arctic MAE pretraining produced consistent improvements in foreground mean F1 scores of 0.87, 0.72, 0.93, and 0.87, for infrastructure, IWP, RTS, and TCNs, with approximately 5-8 percentage increase. The proposed model also outperformed Prithvi-EO-2.0 in all downstream comparisons, with the smallest gain corresponding to at least a 15 percentage improvement mean F1, suggesting that domain-specific self-supervised pretraining on curated Arctic VHSR imagery provides more transferable representations for fine-scale Arctic mapping than a general-purpose Earth observation foundation model. These results demonstrate that optimizing the pretraining data distribution at regional scale, while keeping the architecture and MAE objective fixed, can produce a reusable Arctic-domain encoder for multiple VHSR remote sensing applications.

Problem

• Arctic remote sensing
• very-high-spatial-resolution
• foundation model
• domain-specific pretraining
• image clustering

Innovation

• domain-specific pretraining
• clustering-guided curation
• masked autoencoder
• Vision Transformer
• very-high-resolution remote sensing