DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation

With advancements in remote sensing technologies, high-resolution imagery has become increasingly accessible, supporting applications in urban planning, environmental monitoring, and precision agriculture. However, semantic segmentation of such imagery remains challenging due to complex spatial stru...

Full description

Saved in:
Bibliographic Details
Main Authors: Weijun Meng, Lianlei Shan, Sugang Ma, Dan Liu, Bin Hu
Format: Article
Language:English
Published: MDPI AG 2025-03-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/17/7/1119
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849769424915005440
author Weijun Meng
Lianlei Shan
Sugang Ma
Dan Liu
Bin Hu
author_facet Weijun Meng
Lianlei Shan
Sugang Ma
Dan Liu
Bin Hu
author_sort Weijun Meng
collection DOAJ
description With advancements in remote sensing technologies, high-resolution imagery has become increasingly accessible, supporting applications in urban planning, environmental monitoring, and precision agriculture. However, semantic segmentation of such imagery remains challenging due to complex spatial structures, fine-grained details, and land cover variations. Existing methods often struggle with ineffective feature representation, suboptimal fusion of global and local information, and high computational costs, limiting segmentation accuracy and efficiency. To address these challenges, we propose the dual-level network (DLNet), an enhanced framework incorporating self-attention and cross-attention mechanisms for improved multi-scale feature extraction and fusion. The self-attention module captures long-range dependencies to enhance contextual understanding, while the cross-attention module facilitates bidirectional interaction between global and local features, improving spatial coherence and segmentation quality. Additionally, DLNet optimizes computational efficiency by balancing feature refinement and memory consumption, making it suitable for large-scale remote sensing applications. Extensive experiments on benchmark datasets, including DeepGlobe and Inria Aerial, demonstrate that DLNet achieves state-of-the-art segmentation accuracy while maintaining computational efficiency. On the DeepGlobe dataset, DLNet achieves a <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>76.9</mn><mo>%</mo></mrow></semantics></math></inline-formula> mean intersection over union (mIoU), outperforming existing models such as GLNet (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>71.6</mn><mo>%</mo></mrow></semantics></math></inline-formula>) and EHSNet (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>76.3</mn><mo>%</mo></mrow></semantics></math></inline-formula>), while requiring lower memory (1443 MB) and maintaining a competitive inference speed of 518.3 ms per image. On the Inria Aerial dataset, DLNet attains an mIoU of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>73.6</mn><mo>%</mo></mrow></semantics></math></inline-formula>, surpassing GLNet (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>71.2</mn><mo>%</mo></mrow></semantics></math></inline-formula>) while reducing computational cost and achieving an inference speed of 119.4 ms per image. These results highlight DLNet’s effectiveness in achieving precise and efficient segmentation in high-resolution remote sensing imagery.
format Article
id doaj-art-4028ebb2abce46218a9715bdcf62d990
institution DOAJ
issn 2072-4292
language English
publishDate 2025-03-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-4028ebb2abce46218a9715bdcf62d9902025-08-20T03:03:25ZengMDPI AGRemote Sensing2072-42922025-03-01177111910.3390/rs17071119DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing SegmentationWeijun Meng0Lianlei Shan1Sugang Ma2Dan Liu3Bin Hu4School of Computer Science and Technology, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaSchool of Computer Science and Technology, University of Chinese Academy of Sciences, Beijing 101408, ChinaSchool of Computer Science and Technology, Xi’an University of Posts and Telecommunications, Xi’an 710121, ChinaDepartment of Management, Kean University, Union, NJ 07083, USADepartment of Computer Science and Technology, Kean University, Union, NJ 07083, USAWith advancements in remote sensing technologies, high-resolution imagery has become increasingly accessible, supporting applications in urban planning, environmental monitoring, and precision agriculture. However, semantic segmentation of such imagery remains challenging due to complex spatial structures, fine-grained details, and land cover variations. Existing methods often struggle with ineffective feature representation, suboptimal fusion of global and local information, and high computational costs, limiting segmentation accuracy and efficiency. To address these challenges, we propose the dual-level network (DLNet), an enhanced framework incorporating self-attention and cross-attention mechanisms for improved multi-scale feature extraction and fusion. The self-attention module captures long-range dependencies to enhance contextual understanding, while the cross-attention module facilitates bidirectional interaction between global and local features, improving spatial coherence and segmentation quality. Additionally, DLNet optimizes computational efficiency by balancing feature refinement and memory consumption, making it suitable for large-scale remote sensing applications. Extensive experiments on benchmark datasets, including DeepGlobe and Inria Aerial, demonstrate that DLNet achieves state-of-the-art segmentation accuracy while maintaining computational efficiency. On the DeepGlobe dataset, DLNet achieves a <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>76.9</mn><mo>%</mo></mrow></semantics></math></inline-formula> mean intersection over union (mIoU), outperforming existing models such as GLNet (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>71.6</mn><mo>%</mo></mrow></semantics></math></inline-formula>) and EHSNet (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>76.3</mn><mo>%</mo></mrow></semantics></math></inline-formula>), while requiring lower memory (1443 MB) and maintaining a competitive inference speed of 518.3 ms per image. On the Inria Aerial dataset, DLNet attains an mIoU of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>73.6</mn><mo>%</mo></mrow></semantics></math></inline-formula>, surpassing GLNet (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>71.2</mn><mo>%</mo></mrow></semantics></math></inline-formula>) while reducing computational cost and achieving an inference speed of 119.4 ms per image. These results highlight DLNet’s effectiveness in achieving precise and efficient segmentation in high-resolution remote sensing imagery.https://www.mdpi.com/2072-4292/17/7/1119high-resolution imageryremote sensingsemantic segmentationself-attentioncross-attention
spellingShingle Weijun Meng
Lianlei Shan
Sugang Ma
Dan Liu
Bin Hu
DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation
Remote Sensing
high-resolution imagery
remote sensing
semantic segmentation
self-attention
cross-attention
title DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation
title_full DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation
title_fullStr DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation
title_full_unstemmed DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation
title_short DLNet: A Dual-Level Network with Self- and Cross-Attention for High-Resolution Remote Sensing Segmentation
title_sort dlnet a dual level network with self and cross attention for high resolution remote sensing segmentation
topic high-resolution imagery
remote sensing
semantic segmentation
self-attention
cross-attention
url https://www.mdpi.com/2072-4292/17/7/1119
work_keys_str_mv AT weijunmeng dlnetaduallevelnetworkwithselfandcrossattentionforhighresolutionremotesensingsegmentation
AT lianleishan dlnetaduallevelnetworkwithselfandcrossattentionforhighresolutionremotesensingsegmentation
AT sugangma dlnetaduallevelnetworkwithselfandcrossattentionforhighresolutionremotesensingsegmentation
AT danliu dlnetaduallevelnetworkwithselfandcrossattentionforhighresolutionremotesensingsegmentation
AT binhu dlnetaduallevelnetworkwithselfandcrossattentionforhighresolutionremotesensingsegmentation