An efficient dual-branch segmentation network for remote sensing images
IEICE Transactions on Information and Systems · 2026
DOI: 10.1587/transinf.2026edp7031 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEICE Transactions on Information and Systems, 2026. Verfasst von Hao CHEN und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Remote sensing segmentation underpins land classification and resource management. However, significant variation in object scales, high scene complexity, and blurred boundaries in remote sensing images make accurate segmentation particularly challenging. Traditional convolutional neural networks are limited in long-range modeling capability, while Transformer-based architectures often incur high computational costs, leading to a trade-off between prediction quality and inference speed. To address these issues, we propose an efficient dual-branch segmentation network for remote sensing images. The main branch extracts fine-grained local spatial features via a U-Net backbone, while an auxiliary branch based on the visual state space model captures long-range global context. To achieve adaptive and effective feature integration, a learnable gated fusion mechanism is designed to dynamically fuse multi-level features. Furthermore, a lightweight atrous spatial pyramid pooling module is introduced to enhance multi-scale object perception without introducing heavy computation. We conduct experimental validations on two publicly available remote sensing datasets, verifying that the proposed model achieves favorable segmentation accuracy and possesses great potential for practical deployment.
Zitieren
APA
CHEN, H., & HE, L. (2026). An efficient dual-branch segmentation network for remote sensing images. IEICE Transactions on Information and Systems. https://doi.org/10.1587/transinf.2026edp7031
Vancouver
CHEN H, HE L. An efficient dual-branch segmentation network for remote sensing images. IEICE Transactions on Information and Systems. 2026. doi:10.1587/transinf.2026edp7031
BibTeX
@article{CHEN2026,
author = {CHEN, Hao and HE, Long},
title = {An efficient dual-branch segmentation network for remote sensing images},
journal = {IEICE Transactions on Information and Systems},
year = {2026},
doi = {10.1587/transinf.2026edp7031}
}
EUVIMED