Global–Local Information Fusion Network for Road Extraction: Bridging the Gap in Accurate Road Segmentation in China

Road extraction is crucial in urban planning, rescue operations, and military applications. Compared to traditional methods, using deep learning for road from remote sensing images has demonstrated unique advantages. However, previous convolutional neural networks (CNN)-based methods have had limited receptivity failed effectively capture long-distance features. On the other hand, transformer-based good global information-capturing capabilities, but face challenges extracting edge information. Additionally, existing excellent lack validation Chinese region. To address these issues, this paper proposes a novel model called global–local information fusion network (GLNet). In model, (GIE) module integrates contextual relationships, local (LIE) accurately captures information, (IF) combines output features both branches generate final results. Further, series of experiments on two different datasets with geographic robustness demonstrate that our outperforms state-of-the-art models tasks China. CHN6-CUG dataset, overall accuracy (OA) intersection over union (IoU) reach 97.49% 63.27%, respectively, while RDCME OA IoU 98.73% 84.97%, respectively. These research results hold significant implications traffic, humanitarian rescue, environmental monitoring, particularly context