Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification

Recently, transformers have garnered significant attention due to their exceptional capability to capture long-range dependencies in data. A critical factor contributing to their superior performance is their ability to operate over large receptive fields. As such, a natural question arises as to ho...

Full description

Saved in:
Bibliographic Details
Main Authors: Rong Liu, Zhilin Li, Jiaqi Yang, Jian Sun, Quanwei Liu
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/11007459/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850161258281566208
author Rong Liu
Zhilin Li
Jiaqi Yang
Jian Sun
Quanwei Liu
author_facet Rong Liu
Zhilin Li
Jiaqi Yang
Jian Sun
Quanwei Liu
author_sort Rong Liu
collection DOAJ
description Recently, transformers have garnered significant attention due to their exceptional capability to capture long-range dependencies in data. A critical factor contributing to their superior performance is their ability to operate over large receptive fields. As such, a natural question arises as to how to expand the receptive fields in convolutional neural networks to achieve the superior performance comparable with that of transformers. Large kernel convolution provides the inspiration for the above issue. To explore the potential of large kernel convolution, we propose a hyperspectral image (HSI) classification algorithm in this article that utilizes a large kernel convolution module combined with multiscale coattention and an adaptive geometric feature (AGF) classifier, named Hyper-LKCNet. By integrating this feature enhancement module, our method effectively adjusts the contributions of various spectral and spatial features, ensuring that the network captures critical but easily overlooked information across both dimensions and improving the performance to classify HSI. The AGF classifier, derived by neural collapse theory, alleviates the sample imbalance problem and incorporates the label smoothing focal loss function to enhance generalization ability. Extensive experiments on four HSI datasets demonstrate that the proposed method outperforms the state-of-the-art approaches. In addition, our algorithm maintains a low parameter count and reduced floating point of operations.
format Article
id doaj-art-be417d8488804b6cb615139dbaf8e2d3
institution OA Journals
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-be417d8488804b6cb615139dbaf8e2d32025-08-20T02:22:55ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118139501396610.1109/JSTARS.2025.357195411007459Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image ClassificationRong Liu0https://orcid.org/0000-0002-4642-9086Zhilin Li1https://orcid.org/0009-0002-5817-5615Jiaqi Yang2https://orcid.org/0000-0001-8322-9270Jian Sun3Quanwei Liu4https://orcid.org/0000-0003-4976-1858School of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaSchool of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaState Key Laboratory of Information Engineering in Surveying Mapping, and Remote Sensing, Wuhan University, Wuhan, ChinaEnshi Urban Planning and Design Institute Company Ltd., Enshi, ChinaCollege of Science and Engineering, James Cook University, Cairns, QLD, AustraliaRecently, transformers have garnered significant attention due to their exceptional capability to capture long-range dependencies in data. A critical factor contributing to their superior performance is their ability to operate over large receptive fields. As such, a natural question arises as to how to expand the receptive fields in convolutional neural networks to achieve the superior performance comparable with that of transformers. Large kernel convolution provides the inspiration for the above issue. To explore the potential of large kernel convolution, we propose a hyperspectral image (HSI) classification algorithm in this article that utilizes a large kernel convolution module combined with multiscale coattention and an adaptive geometric feature (AGF) classifier, named Hyper-LKCNet. By integrating this feature enhancement module, our method effectively adjusts the contributions of various spectral and spatial features, ensuring that the network captures critical but easily overlooked information across both dimensions and improving the performance to classify HSI. The AGF classifier, derived by neural collapse theory, alleviates the sample imbalance problem and incorporates the label smoothing focal loss function to enhance generalization ability. Extensive experiments on four HSI datasets demonstrate that the proposed method outperforms the state-of-the-art approaches. In addition, our algorithm maintains a low parameter count and reduced floating point of operations.https://ieeexplore.ieee.org/document/11007459/Attention mechanismhyperspectral image (HSI) classificationimbalanced datalarge kernel convolution
spellingShingle Rong Liu
Zhilin Li
Jiaqi Yang
Jian Sun
Quanwei Liu
Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Attention mechanism
hyperspectral image (HSI) classification
imbalanced data
large kernel convolution
title Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification
title_full Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification
title_fullStr Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification
title_full_unstemmed Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification
title_short Hyper-LKCNet: Exploring the Utilization of Large Kernel Convolution for Hyperspectral Image Classification
title_sort hyper lkcnet exploring the utilization of large kernel convolution for hyperspectral image classification
topic Attention mechanism
hyperspectral image (HSI) classification
imbalanced data
large kernel convolution
url https://ieeexplore.ieee.org/document/11007459/
work_keys_str_mv AT rongliu hyperlkcnetexploringtheutilizationoflargekernelconvolutionforhyperspectralimageclassification
AT zhilinli hyperlkcnetexploringtheutilizationoflargekernelconvolutionforhyperspectralimageclassification
AT jiaqiyang hyperlkcnetexploringtheutilizationoflargekernelconvolutionforhyperspectralimageclassification
AT jiansun hyperlkcnetexploringtheutilizationoflargekernelconvolutionforhyperspectralimageclassification
AT quanweiliu hyperlkcnetexploringtheutilizationoflargekernelconvolutionforhyperspectralimageclassification