DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images

Land use classification poses significant challenges when applied to remote sensing images. Due to the complex textures, spatial layouts, and scale variations of images, many methods solve these problems by ignoring the bias between low-level and high-level features caused by multiple semantic infor...

Full description

Saved in:
Bibliographic Details
Main Authors: Chenke Yue, Yin Zhang, Junhua Yan, Zhaolong Luo, Yong Liu, Pengyu Guo
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10902592/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850023259901263872
author Chenke Yue
Yin Zhang
Junhua Yan
Zhaolong Luo
Yong Liu
Pengyu Guo
author_facet Chenke Yue
Yin Zhang
Junhua Yan
Zhaolong Luo
Yong Liu
Pengyu Guo
author_sort Chenke Yue
collection DOAJ
description Land use classification poses significant challenges when applied to remote sensing images. Due to the complex textures, spatial layouts, and scale variations of images, many methods solve these problems by ignoring the bias between low-level and high-level features caused by multiple semantic information associated with each pixel and the effectiveness of multiscale fusion. To tackle the challenges, we propose a novel bidirectional feature enhancement network based on dynamic assembled kernels, which captures both low-level spatial and high-level semantic information of the corrected image through mutual guidance between deep and shallow features. Specifically, we employ high-level semantic features derived from the bilateral structure to compute the semantic deviation of each pixel in the low-level features. Meanwhile, we use the low-level features to resolve redundant information in the high-level components and enhance their global and local context through mutual guidance. On the other hand, we generate kernels by dynamically assembling elementary weight matrices stored in the weight library. The kernel construction is data driven, providing greater flexibility to multiscale features. We have conducted extensive objective and subjective comparative experiments, as well as ablation studies, on the International Society for Photogrammetry and Remote Sensing (ISPRS) Vaihingen, ISPRS Potsdam, and GaoFen Image Dataset. In conclusion, our method has demonstrated notable superiority over other prevailing methods, as evidenced by numerous experimental results.
format Article
id doaj-art-d4c4bf8800f248acbedac71a65d7acfe
institution DOAJ
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-d4c4bf8800f248acbedac71a65d7acfe2025-08-20T03:01:27ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01187117713310.1109/JSTARS.2025.354536510902592DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing ImagesChenke Yue0https://orcid.org/0000-0002-0122-0860Yin Zhang1https://orcid.org/0000-0002-1908-4000Junhua Yan2https://orcid.org/0000-0003-2042-2342Zhaolong Luo3Yong Liu4https://orcid.org/0000-0002-9141-6938Pengyu Guo5https://orcid.org/0000-0001-5592-0558Key Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology, Jiangsu, ChinaKey Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology, Jiangsu, ChinaKey Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology, Jiangsu, ChinaKey Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology, Jiangsu, ChinaNational Innovation Institute of Defense Technology, Academy of Military Sciences, Beijing, ChinaNational Innovation Institute of Defense Technology, Academy of Military Sciences, Beijing, ChinaLand use classification poses significant challenges when applied to remote sensing images. Due to the complex textures, spatial layouts, and scale variations of images, many methods solve these problems by ignoring the bias between low-level and high-level features caused by multiple semantic information associated with each pixel and the effectiveness of multiscale fusion. To tackle the challenges, we propose a novel bidirectional feature enhancement network based on dynamic assembled kernels, which captures both low-level spatial and high-level semantic information of the corrected image through mutual guidance between deep and shallow features. Specifically, we employ high-level semantic features derived from the bilateral structure to compute the semantic deviation of each pixel in the low-level features. Meanwhile, we use the low-level features to resolve redundant information in the high-level components and enhance their global and local context through mutual guidance. On the other hand, we generate kernels by dynamically assembling elementary weight matrices stored in the weight library. The kernel construction is data driven, providing greater flexibility to multiscale features. We have conducted extensive objective and subjective comparative experiments, as well as ablation studies, on the International Society for Photogrammetry and Remote Sensing (ISPRS) Vaihingen, ISPRS Potsdam, and GaoFen Image Dataset. In conclusion, our method has demonstrated notable superiority over other prevailing methods, as evidenced by numerous experimental results.https://ieeexplore.ieee.org/document/10902592/Dynamically assemblingglobal and local contextmultiscaleremote sensingsemantic bias
spellingShingle Chenke Yue
Yin Zhang
Junhua Yan
Zhaolong Luo
Yong Liu
Pengyu Guo
DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Dynamically assembling
global and local context
multiscale
remote sensing
semantic bias
title DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images
title_full DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images
title_fullStr DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images
title_full_unstemmed DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images
title_short DAKBNet: Multiscale Fusion With Dynamic Assembly Kernels and Bilateral Feature Enhancement for Land Use Classification of Remote Sensing Images
title_sort dakbnet multiscale fusion with dynamic assembly kernels and bilateral feature enhancement for land use classification of remote sensing images
topic Dynamically assembling
global and local context
multiscale
remote sensing
semantic bias
url https://ieeexplore.ieee.org/document/10902592/
work_keys_str_mv AT chenkeyue dakbnetmultiscalefusionwithdynamicassemblykernelsandbilateralfeatureenhancementforlanduseclassificationofremotesensingimages
AT yinzhang dakbnetmultiscalefusionwithdynamicassemblykernelsandbilateralfeatureenhancementforlanduseclassificationofremotesensingimages
AT junhuayan dakbnetmultiscalefusionwithdynamicassemblykernelsandbilateralfeatureenhancementforlanduseclassificationofremotesensingimages
AT zhaolongluo dakbnetmultiscalefusionwithdynamicassemblykernelsandbilateralfeatureenhancementforlanduseclassificationofremotesensingimages
AT yongliu dakbnetmultiscalefusionwithdynamicassemblykernelsandbilateralfeatureenhancementforlanduseclassificationofremotesensingimages
AT pengyuguo dakbnetmultiscalefusionwithdynamicassemblykernelsandbilateralfeatureenhancementforlanduseclassificationofremotesensingimages