WATCH: Wide-Area Archaeological Site Tracking for Change Detection

Girmaw Abebe Tadesse, Titien Bartette, Andrew Hassanali, Allen Kim, Jonathan Chemla, Andrew Zolli, Yves Ubelmann, Caleb Robinson, Inbal Becker-Reshef, Juan Lavista Ferres

arXiv:2605.08160 · 2026-05-12 공개 · arXiv · PDF

foundation-models self-supervised-learning satellite-imagery weakly-supervised change-detection watch-framework planetcope-satellite temporal-embedding-distance

Abstract

Monitoring archaeological sites at scale is vital for protecting cultural heritage, yet pinpointing when disturbances occur remains difficult because visual cues are subtle and ground-truth data are sparse. We introduce WATCH, a framework for month-level change-event localization over PlanetScope satellite mosaics (2017-2024, 4.7 m/px) that supports three complementary scoring approaches: (i) Temporal Embedding Distance (TED), a training-free method that scores month-to-month deviations from a local temporal reference; (ii) Self-Supervised Change Detection (SSCD), an ensemble of reconstruction, forecasting, and latent-novelty signals; and (iii) a Weakly Supervised (WS) temporal localization model trained with sparse event-month labels. We benchmark WATCH on 1,943 archaeological sites in Afghanistan using embeddings from six foundation models (CLIP, GeoRSCLIP, SatMAE, Prithvi-EO-2.0, DINOv3, and Satlas-Pretrain) alongside a handcrafted spectral and texture baseline, and assess cross-regional generalization on sites in Syria, Turkey, Pakistan, and Egypt. The unsupervised approaches (TED, SSCD) consistently outperform the weakly supervised alternative. TED with SatMAE achieves the highest exact-month recall (55% at m=0), while TED with GeoRSCLIP, CLIP, or Satlas-Pretrain reaches 92.5% within a three-month tolerance (m=3). Handcrafted features remain competitive for exact-month detection under weak supervision. Our directional margin analysis reveals systematic temporal biases: SSCD paired with GeoRSCLIP or Prithvi-EO-2.0 exhibits the strongest early-warning profile, detecting anomalies before the recorded event, while TED favors confirmation-oriented detection after a change has materialized. These results show that satellite imagery combined with foundation-model embeddings enables scalable, decision-relevant heritage monitoring. Code: https://github.com/microsoft/WATCH

한국어 요약

📋 한 줄 요약

**[원격 탐사 / 변화 탐지]** 광역 위성 시계열에서 월 단위 변화 시점을 찾는 WATCH 프레임워크로 1,943개 아프가니스탄 고고학 유적지를 벤치마킹, SatMAE 기반 TED가 정확 월 회수율 55% 달성.

🎯 핵심 기여도

💡 핵심 아이디어

유적 훼손 같은 변화는 시각 단서가 미세하고 라벨이 매우 희소하다. 본 연구는 “학습 없는 임베딩 거리 기반 시점 탐지(TED)”와 “재구성·예측·잠재 신호의 자기지도 앙상블(SSCD)”을 통해 라벨 의존성을 줄이고, 파운데이션 모델 임베딩의 시계열적 일관성을 직접 활용해 변화 시점을 찾는다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 라벨이 거의 없는 광역 모니터링 시나리오에서 파운데이션 모델 임베딩의 “학습 없는 활용”이 강력한 베이스라인임을 입증, 문화유산 보호의 운영 가능한 도구를 제공한다. **한계**: PlanetScope의 시공간 해상도와 구름·계절 변화에 영향을 받으며, 비고고학 변화(농업·공사) 구분과 사건 “종류” 분류에는 추가 모델링이 필요하다.

🚀 실용적 활용