基于混合光谱增强与多尺度空间聚合的高光谱图像分类方法

doi:10.13229/j.cnki.jdxbgxb.20240252

吉林大学学报(工学版) ›› 2025, Vol. 55 ›› Issue (11): 3727-3735.doi: 10.13229/j.cnki.jdxbgxb.20240252

• 计算机科学与技术 • 上一篇

基于混合光谱增强与多尺度空间聚合的高光谱图像分类方法

欧阳宁^1,²(),黄辰钰²,林乐平^1,²()

^1.桂林电子科技大学认知无线电与信息处理省部共建教育部重点实验室，广西桂林 541004
^2.桂林电子科技大学信息与通信学院，广西桂林 541004

收稿日期:2024-03-12 出版日期:2025-11-01 发布日期:2026-02-03
通讯作者: 林乐平 E-mail:ouyangning@guet.edu.cn;linleping@guet.edu.cn
作者简介:欧阳宁（1972-），男，教授，硕士. 研究方向：数字图像处理，智能信息处理.E-mail： ouyangning@guet.edu.cn
基金资助:
国家自然科学基金项目(62001133);广西科技基地和人才专项项目(桂科 AD19110060);广西自然科学基金项目(2017GXNSFBA198212);广西无线宽带通信与信号处理重点实验室基金项目(GXKL06200114)

Hyperspectral image classification based on hybrid spectral enhancement and multi-scale spatial aggregation

Ning OUYANG^1,²(),Chen-yu HUANG²,Le-ping LIN^1,²()

^1.Key Laboratory of Cognitive Radio and Information Processing of Ministry of Education，Guilin University of Electronic Technology，Guilin 541004，China
^2.School of Information and Communication，Guilin University of Electronic Technology，Guilin 541004，China

Received:2024-03-12 Online:2025-11-01 Published:2026-02-03
Contact: Le-ping LIN E-mail:ouyangning@guet.edu.cn;linleping@guet.edu.cn

摘要/Abstract

摘要：

由于高光谱图像存在同物异谱和异物同谱现象，仅依赖光谱信息无法充分表征高光谱图像的特征，因此可引入空间信息以更准确地捕捉物体特征。为此，本文提出一种基于混合光谱增强与多尺度空间聚合的高光谱图像分类方法。该方法设计了混合光谱增强模块，利用小波变换构建光谱的多尺度局部特征，通过Transformer架构生成光谱的全局特征，以增强光谱特征的类内一致性。同时，设计了多尺度空间聚合模块，用于提取空间特征固有的多尺度信息，并建立不同尺度间的交互关系，以生成更具鲁棒性的土地覆盖表示，从而进一步提升分类性能。实验结果表明：本文方法相较于其他先进网络表现出显著的优越性，能有效获取更丰富的光谱信息和空间特征表示。

关键词: 高光谱图像分类, 混合光谱增强模块, 小波变换, 多尺度空间聚合模块

Abstract:

Due to the presence of the same object with different spectrum and the same spectrum with different objects in hyperspectral images， spectral information alone cannot fully reflect the features of hyperspectral images， and spatial information can be introduced to capture the features of objects more accurately. Therefore， a hyperspectral image classification method based on hybrid spectral enhancement and multi-scale spatial aggregation is proposed in this paper. In this method， a hybrid spectral enhancement module is designed， multi-scale local features of the spectrum are constructed using wavelet transform， and global features of the spectrum are generated by Transformer architecture， so as to enhance the intra-class consistency of the spectral features. At the same time， a multi-scale spatial aggregation module is designed to extract the inherent multi-scale information of spatial features and establish the interaction between different scales， so as to generate a more robust land cover representation， thereby further improve the classification performance. The experimental results show that the proposed method is superior to other advanced networks， indicating that the method can effectively obtain more abundant spectral information and spatial feature representation.

Key words: hyperspectral image classification, hybrid spectral enhancement module, wavelet transform, multi-scale spatial aggregation module

中图分类号:

TP753

欧阳宁,黄辰钰,林乐平. 基于混合光谱增强与多尺度空间聚合的高光谱图像分类方法[J]. 吉林大学学报(工学版), 2025, 55(11): 3727-3735.

Ning OUYANG,Chen-yu HUANG,Le-ping LIN. Hyperspectral image classification based on hybrid spectral enhancement and multi-scale spatial aggregation[J]. Journal of Jilin University(Engineering and Technology Edition), 2025, 55(11): 3727-3735.

图/表 10

图1

图2

图3

图4

表1

表2

表3

图5

图6

表4

参考文献 21

[1]	Yang K, Sun H, Zou C B, et al. Cross-attention spectral-spatial network for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-14.
[2]	Ahmad M, Khan A M, Mazzara M, et al. Hybrid dense network with attention mechanism for hyperspectral image classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 3948-3957.
[3]	欧阳宁,李祖锋,林乐平.基于多层次空-谱融合网络的高光谱图像分类[J].吉林大学学报:工学版,2022,52(10):2438-2446.
	Ouyang Ning, Li Zu-feng, Lin Le-ping. Hyperspectral image classification based on a multi-level spatial-spectral fusion network[J]. Journal of Jilin University (Engineering and Technology Edition), 2022,52(10):2438-2446.
[4]	欧阳宁,李祖锋,林乐平.基于空谱分组卷积密集网络的高光谱图像分类[J].计算机工程与设计,2022,43(7):2031-2039.
	Ouyang Ning, Li Zu-feng, Lin Le-ping. Hyperspectral image classification based on a spatial-spectral group convolution dense network[J]. Computer Engineering and Design, 2022,43(7):2031-2039.
[5]	Hamida A B, Benoit A, Lambert P, et al. 3-D deep learning approach for remote sensing image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(8): 4420-4434.
[6]	Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16×16 words: Transformers for image recognition at scale[J]. Arxiv Preprint, 2010, 11: 201011929.
[7]	Hong D F, Han Z, Yao J, et al. Spectralformer: rethinking hyperspectral image classification with transformers[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-15.
[8]	Sun L, Zhao G R, Zheng Y H, et al. Spectral–spatial feature tokenization Transformer for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-14.
[9]	Uddin M P, Mamun M A, Hossain M A. Feature extraction for hyperspectral image classification[C]∥ IEEE Region 10 Humanitarian Technology Conference. Piscataway, NJ: IEEE. 2017: 379-382.
[10]	Zhang D S. Wavelet Transform[J]. Fundamentals of Image Data Mining: Analysis, Features, Classification and Retrieval, 2019: 35-44.
[11]	Tang Y G, Xie X L, Yu Y H. Hyperspectral classification of two-branch joint networks based on gaussian pyramid multiscale and wavelet transform[J]. IEEE Access, 2022, 10: 56876-56887.
[12]	Ahmad M, Ghous U, Usama M, et al. WaveFormer: spectral-spatial wavelet Transformer for hyperspectral image classification[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21: 1-5.
[13]	Wang D, Du B, Zhang L P, et al. Adaptive spectral-spatial multiscale contextual feature extraction for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(3): 2461-2477.
[14]	Gao H M, Yang Y, Li C M, et al. Multiscale residual network with mixed depthwise convolution for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(4): 3396-3408.
[15]	Leng Q M, Yang H O, Jiang J J, et al. Adaptive multiscale segmentations for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(8): 5847-5860.
[16]	Ghaderizadeh S, Abbasi M D, Sharifi A, et al. Multiscale dual-branch residual spectral-spatial network with attention for hyperspectral image classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 5455-5467.
[17]	Wang X, Tan K, Du P J, et al. A unified multiscale learning framework for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-19.
[18]	Baumgardner M F, Biehl L L, Landgrebe D A. 220 band aviris hyperspectral image data set: June 12, 1992 Indian pine test site 3[J]. Purdue University Research Repository, 2015, 10(7): 991.
[19]	Zhong Y F, Hu X, Luo C, et al. WHU-Hi: UAV-borne hyperspectral with high spatial resolution (H₂) benchmark datasets and classifier for precise crop identification based on deep convolutional neural network with CRF[J]. Remote Sensing of Environment, 2020, 250: 112012.
[20]	Zhong Y F, Wang X Y, Xu Y, et al. Mini-UAV-borne hyperspectral remote sensing: From observation and processing to applications[J]. IEEE Geoscience and Remote Sensing Magazine, 2018, 6(4): 46-62.
[21]	Mei S H, Li X G, Liu X, et al. Hyperspectral image classification using attention-based bidirectional long short-term memory network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-12.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

基线	小波变换	多尺度空间聚合模块	OA/%	AA/%	Kappa×100
SpectralFormer	×	×	80.46	72.49	77.88
	√	×	82.48	73.36	80.19
	×	√	87.72	83.31	86.00
	√	√	94.80	91.90	94.04

类别	ABLSTM	CNN3D	ViT	SpectralFormer	SSFTT	HEMA
1	26.19	57.14	68.96	74.19	100.00	100.00
2	59.92	80.73	67.68	75.84	83.33	96.16
3	48.90	66.48	66.75	63.25	74.96	87.59
4	38.63	39.08	52.04	46.04	74.10	80.76
5	84.86	95.39	80.94	91.00	92.45	97.43
6	90.94	98.30	93.56	92.93	98.10	98.88
7	18.75	34.28	30.76	50.00	48.00	92.30
8	84.84	98.98	92.82	97.75	99.49	100.00
9	4.54	10.86	13.33	20.83	100.00	83.33
10	69.80	77.44	63.88	76.57	73.29	86.22
11	73.08	78.49	82.11	90.50	87.30	97.53
12	50.00	48.94	42.17	63.05	56.38	98.52
13	92.00	94.59	83.13	88.60	90.90	97.90
14	97.69	97.46	98.07	97.66	97.18	97.83
15	69.84	72.93	71.15	71.97	81.86	95.32
16	60.00	81.63	64.51	59.70	64.51	60.60
OA/%	69.90	77.89	74.96	80.46	83.77	94.80
AA/%	60.62	70.80	66.99	72.49	83.62	91.90
Kappa×100	65.84	74.89	71.58	77.88	91.54	94.04

类别	ABLSTM	CNN3D	ViT	SpectralFormer	SSFTT	HEMA
1	94.15	99.04	96.16	93.34	99.61	99.21
2	75.67	68.80	70.93	87.46	81.12	97.19
3	25.96	39.53	49.13	91.15	53.96	92.15
4	97.65	98.51	98.45	98.66	98.72	99.28
5	30.99	56.97	30.10	41.39	57.08	48.84
6	91.01	84.21	96.77	96.86	98.61	99.68
7	98.92	98.66	99.92	99.97	99.96	99.95
8	81.01	82.22	84.56	90.22	91.11	94.33
9	28.37	68.43	59.33	90.53	86.68	85.34
OA/%	84.94	91.47	91.06	93.69	95.38	96.75
AA/%	69.30	77.37	76.15	85.51	85.20	90.67
Kappa×100	80.80	89.01	88.44	91.79	93.99	95.76

数据集		ABLS-TM	CNN-3D	ViT	SpectralFormer	SSFTT	HEMA
IN	训练	58.76	24.61	45.99	86.25	22.16	62.23
IN	测试	1.94	0.54	1.69	3.94	0.58	2.28
LK	训练	58.63	13.25	43.44	73.77	20.33	49.47
LK	测试	59.49	16.19	56.50	123.47	18.96	64.92

基于混合光谱增强与多尺度空间聚合的高光谱图像分类方法

Hyperspectral image classification based on hybrid spectral enhancement and multi-scale spatial aggregation

RICH HTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 21

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	赖丹晖,罗伟峰,袁旭东,邱子良. 复杂环境下多模态手势关键点特征提取算法[J]. 吉林大学学报(工学版), 2024, 54(8): 2288-2294.
[2]	欧阳宁,李祖锋,林乐平. 基于多层次空⁃谱融合网络的高光谱图像分类[J]. 吉林大学学报(工学版), 2022, 52(10): 2438-2446.
[3]	李喜艳,周夏冰,刘征. 基于高容量强鲁棒的图像水印算法[J]. 吉林大学学报(工学版), 2022, 52(1): 174-179.
[4]	刘国华,周文斌. 基于卷积神经网络的脉搏波时频域特征混叠分类[J]. 吉林大学学报(工学版), 2020, 50(5): 1818-1825.
[5]	李雄飞,宋璐,张小利. 基于协同经验小波变换的遥感图像融合[J]. 吉林大学学报(工学版), 2019, 49(4): 1307-1319.
[6]	车翔玖, 梁森. 一种基于大顶堆的SPIHT改进算法[J]. 吉林大学学报(工学版), 2016, 46(3): 865-869.
[7]	季彦婕, 陈晓实, 王炜, 胡波. 基于小波变换和粒子群小波神经网络组合模型的有效停车泊位短时预测[J]. 吉林大学学报(工学版), 2016, 46(2): 399-405.
[8]	肖钟捷. 基于小波空间特征谱熵的数字图像识别[J]. 吉林大学学报(工学版), 2015, 45(6): 1994-1998.
[9]	齐滨, 赵春晖, 王玉磊. 基于支持向量机与相关向量机的高光谱图像分类[J]. 吉林大学学报(工学版), 2013, 43(增刊1): 143-147.
[10]	刘媛媛, 陈贺新, 赵岩, 孙红岩. 一种新的DWT视频动态水印算法[J]. 吉林大学学报(工学版), 2013, 43(增刊1): 445-449.
[11]	张久文, 米进财, 张同峰. 基于双树复小波和广义高斯密度的纹理图像检索[J]. 吉林大学学报(工学版), 2013, 43(增刊1): 60-63.
[12]	包磊, 徐其志. 基于PCA变换和光谱补偿的遥感影像融合方法[J]. 吉林大学学报(工学版), 2013, 43(增刊1): 88-91.
[13]	夏英杰, 李金屏, 陈瑞. 基于变电站多模图像的自动集成配准方法[J]. 吉林大学学报(工学版), 2013, 43(增刊1): 47-50.
[14]	王昕, 李玮琳, 刘富. 小波域CT/MRI医学图像融合新方法[J]. 吉林大学学报(工学版), 2013, 43(增刊1): 25-28.
[15]	胡玉平, 王志坚, 张玲华, 尹华. 基于小波变换和混沌映射的自适应水印算法[J]. 吉林大学学报(工学版), 2012, 42(增刊1): 401-404.