Scholarlink

<p>High-resolution building extraction from remote sensing images plays a pivotal role in urban planning, disaster response, and land-use monitoring. However, complex urban environments–characterized by dense building distributions, diverse architectural styles, and blurred object boundaries–pose significant challenges for existing semantic segmentation models. To address these issues, we propose DE-Net, a novel semantic segmentation framework for precise building extraction from high-resolution unmanned aerial vehicle (UAV) imagery. DE-Net consists of three key components: a ConvNeXt-based feature backbone that captures hierarchical semantic features while preserving spatial details; a Spatial Resolution Adaptive Reconstruction Module (SRARM) that dynamically decodes features based on predicted density maps, enabling tailored upsampling strategies for dense and sparse regions; and a Multi-Scale Edge-Aware Fusion Module (MS-EAM) that extracts edge attention from each encoder stage and fuses them to enhance boundary localization. Extensive experiments conducted on a high-resolution UAV dataset demonstrate that DE-Net outperforms competing methods in terms of Intersection over Union (IoU), F1 score, and boundary IoU. Specifically, DE-Net achieves an IoU of 86.78% and an F1 score of 92.91%, surpassing the strongest competing IoU model, Swin-UNet, by 1.66 percentage points and the strongest competing F1 model, UPerNet, by 2.65 percentage points, thereby achieving state-of-the-art performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DE-Net: a density-aware and edge-enhanced network for high-resolution building segmentation

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Guanjun Huang,
  • Rui Wu,
  • Liang Qiao

摘要

High-resolution building extraction from remote sensing images plays a pivotal role in urban planning, disaster response, and land-use monitoring. However, complex urban environments–characterized by dense building distributions, diverse architectural styles, and blurred object boundaries–pose significant challenges for existing semantic segmentation models. To address these issues, we propose DE-Net, a novel semantic segmentation framework for precise building extraction from high-resolution unmanned aerial vehicle (UAV) imagery. DE-Net consists of three key components: a ConvNeXt-based feature backbone that captures hierarchical semantic features while preserving spatial details; a Spatial Resolution Adaptive Reconstruction Module (SRARM) that dynamically decodes features based on predicted density maps, enabling tailored upsampling strategies for dense and sparse regions; and a Multi-Scale Edge-Aware Fusion Module (MS-EAM) that extracts edge attention from each encoder stage and fuses them to enhance boundary localization. Extensive experiments conducted on a high-resolution UAV dataset demonstrate that DE-Net outperforms competing methods in terms of Intersection over Union (IoU), F1 score, and boundary IoU. Specifically, DE-Net achieves an IoU of 86.78% and an F1 score of 92.91%, surpassing the strongest competing IoU model, Swin-UNet, by 1.66 percentage points and the strongest competing F1 model, UPerNet, by 2.65 percentage points, thereby achieving state-of-the-art performance.