UniRoute: Unified Routing Mixture-of-Experts for Modality-Adaptive Remote Sensing Change Detection

2026-01-21

AI Summary

This work addresses the limitations of existing remote sensing change detection methods, which rely on specialized models and struggle to adaptively switch between homogeneous and heterogeneous modalities, while fixed differencing operations often introduce artifacts under cross-modal or geometrically misaligned conditions. To overcome these challenges, the authors propose UniRoute, a unified framework that reformulates feature extraction and fusion as a conditional routing problem to enable modality-adaptive change detection. Key innovations include the AR2-MoE module, which decouples local details from global semantics, and the MDR-MoE module, which adaptively selects the optimal fusion operator at the pixel level. Additionally, a consistency-aware self-distillation (CASD) strategy is introduced to enhance training stability under data scarcity. Experiments demonstrate that UniRoute achieves both high accuracy and efficiency across five public datasets using a single, unified architecture.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Current remote sensing change detection (CD) methods mainly rely on specialized models, which limits the scalability toward modality-adaptive Earth observation. For homogeneous CD, precise boundary delineation relies on fine-grained spatial cues and local pixel interactions, whereas heterogeneous CD instead requires broader contextual information to suppress speckle noise and geometric distortions. Moreover, difference operator (e.g., subtraction) works well for aligned homogeneous images but introduces artifacts in cross-modal or geometrically misaligned scenarios. Across different modality settings, specialized models based on static backbones or fixed difference operations often prove insufficient. To address this challenge, we propose UniRoute, a unified framework for modality-adaptive learning by reformulating feature extraction and fusion as conditional routing problems. We introduce an Adaptive Receptive Field Routing MoE (AR2-MoE) module to disentangle local spatial details from global semantic context, and a Modality-Aware Difference Routing MoE (MDR-MoE) module to adaptively select the most suitable fusion primitive at each pixel. In addition, we propose a Consistency-Aware Self-Distillation (CASD) strategy that stabilizes unified training under data-scarce heterogeneous settings by enforcing multi-level consistency. Extensive experiments on five public datasets demonstrate that UniRoute achieves strong overall performance, with a favorable accuracy-efficiency trade-off under a unified deployment setting.

Problem

• remote sensing change detection
• modality-adaptive
• heterogeneous imagery
• difference operator
• model scalability

Innovation

• Mixture-of-Experts
• Modality-Adaptive
• Conditional Routing
• Change Detection
• Self-Distillation