GLD-Road:A global-local decoding road network extraction model for remote sensing images

2025-06-11

AI Summary

Addressing the challenge of simultaneously achieving high accuracy, computational efficiency, and topological connectivity in road network extraction from remote sensing imagery, this paper proposes a two-stage decoding framework that synergistically combines global coarse detection with local refinement. We innovatively design a global node detection and connection prediction module (Connect Module), integrated with local window search and iterative refinement mechanisms, enabling end-to-end topology-aware road extraction. The method significantly reduces computational overhead: it achieves absolute improvements of 1.9% and 0.67% in APLS scores on the City-Scale and SpaceNet3 datasets, respectively; attains 40% faster inference speed than Sat2Graph and 92% faster than RNGDet++; and maintains both high geometric accuracy and structural integrity of road networks, supporting real-time applications.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Remote Sensing / Geospatial AI
• Search and Optimization: Local Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Road networks are crucial for mapping, autonomous driving, and disaster response. While manual annotation is costly, deep learning offers efficient extraction. Current methods include postprocessing (prone to errors), global parallel (fast but misses nodes), and local iterative (accurate but slow). We propose GLD-Road, a two-stage model combining global efficiency and local precision. First, it detects road nodes and connects them via a Connect Module. Then, it iteratively refines broken roads using local searches, drastically reducing computation. Experiments show GLD-Road outperforms state-of-the-art methods, improving APLS by 1.9% (City-Scale) and 0.67% (SpaceNet3). It also reduces retrieval time by 40% vs. Sat2Graph (global) and 92% vs. RNGDet++ (local). The experimental results are available at https://github.com/ucas-dlg/GLD-Road.

Problem

• Efficiently extract road networks from remote sensing images
• Combine global speed with local accuracy in road extraction
• Reduce computational time while maintaining high extraction precision

Innovation

• Two-stage model combining global and local decoding
• Connect Module for efficient road node linking
• Local iterative search for precise road refinement