HRU-Net: High-Resolution Remote Sensing Image Road Extraction Based on Multi-Scale Fusion

Road extraction from high-resolution satellite images has become a significant focus in the field of remote sensing image analysis. However, factors such as shadow occlusion and spectral confusion hinder accuracy consistency road images. To overcome these challenges, this paper presents multi-scale fusion-based framework, HRU-Net, which exploits various scales resolutions features generated during encoding decoding processes. First, phase, we develop feature fusion module with upsampling capabilities (UMR module) to capture fine details, enhancing shadowed areas boundaries. Next, design multi-feature (MPF obtain spatial information, enabling better differentiation between roads objects similar characteristics. The network simultaneously integrates information downsampling process, producing maps through progressive cross-layer connections, thereby more effective prediction tasks. We conduct comparative experiments quantitative evaluations proposed HRU-Net framework against existing algorithms (U-Net, ResNet, DeepLabV3, ResUnet, HRNet) using Massachusetts Dataset. On basis, selects three models HRNet, HRU-Net) on DeepGlobe experimental results demonstrate that outperforms its counterparts terms mean intersection over union. In summary, model skillfully different resolution maps, effectively addressing challenges discontinuous reduced caused by factors. complex scenarios, accurately extracts comprehensive regions.