Sat2Flow: A Structure-Aware Diffusion Framework for Human Flow Generation from Satellite Imagery

2025-08-26

AI Summary

Existing OD flow forecasting methods rely on costly, sparsely available auxiliary data (e.g., POIs, census statistics) and are sensitive to regional indexing order, lacking topological robustness. This paper proposes the first end-to-end framework that generates structurally consistent OD flow matrices solely from satellite imagery. We design a multi-kernel encoder to capture multi-scale land-use features and introduce permutation-aware diffusion modeling coupled with equivariant contrastive learning to achieve topology-invariant representation under arbitrary regional renumbering. Evaluated on real-world urban datasets, our method significantly outperforms both physics-based and data-driven baselines. It preserves flow distribution fidelity and spatial structural consistency even under exponential regional reordering. The approach enables auxiliary-feature-free, globally scalable human mobility prediction—establishing a novel paradigm for low-cost, high-robustness urban mobility modeling.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Planning with Language Models
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Origin-Destination (OD) flow matrices are essential for urban mobility analysis, underpinning applications in traffic forecasting, infrastructure planning, and policy design. However, existing methods suffer from two critical limitations: (1) reliance on auxiliary features (e.g., Points of Interest, socioeconomic statistics) that are costly to collect and have limited spatial coverage; and (2) sensitivity to spatial topology, where minor index reordering of urban regions (e.g., census tract relabeling) disrupts structural coherence in generated flows. To address these challenges, we propose Sat2Flow, a latent structure-aware diffusion-based framework that generates structurally coherent OD flows using solely satellite imagery as input. Our approach introduces a multi-kernel encoder to capture diverse regional interactions and employs a permutation-aware diffusion process that aligns latent representations across different regional orderings. Through a joint contrastive training objective that bridges satellite-derived features with OD patterns, combined with equivariant diffusion training that enforces structural consistency, Sat2Flow ensures topological robustness under arbitrary regional reindexing. Experimental results on real-world urban datasets demonstrate that Sat2Flow outperforms both physics-based and data-driven baselines in numerical accuracy while preserving empirical distributions and spatial structures under index permutations. Sat2Flow offers a globally scalable solution for OD flow generation in data-scarce urban environments, eliminating region-specific auxiliary data dependencies while maintaining structural invariance for robust mobility modeling.

Problem

• Generates human flow matrices using only satellite imagery
• Addresses sensitivity to spatial topology and index reordering
• Eliminates dependency on costly auxiliary urban data features

Innovation

• Satellite imagery input for flow generation
• Permutation-aware diffusion for structural coherence
• Multi-kernel encoder capturing regional interactions