基于高光谱跨场景迁移学习的集成对齐子空间自适应算法

doi:10.12305/j.issn.1001-506X.2025.02.04

系统工程与电子技术 ›› 2025, Vol. 47 ›› Issue (2): 369-375.doi: 10.12305/j.issn.1001-506X.2025.02.04

• 电子技术 • 上一篇

基于高光谱跨场景迁移学习的集成对齐子空间自适应算法

宋怡佳¹, 王海燕², 冯伟¹^,*, 全英汇¹

1. 西安电子科技大学电子工程学院, 陕西西安 710071
2. 陕西省林业科学院, 陕西西安 710016

收稿日期:2024-01-04 出版日期:2025-02-25 发布日期:2025-03-18
通讯作者: 冯伟
作者简介:宋怡佳 (1997—), 女, 硕士研究生, 主要研究方向为迁移学习、遥感大数据、跨场景分类、嵌入式人工智能
王海燕 (1973—), 女, 高级工程师, 本科, 主要研究方向为经济林培育与利用、林业综合监测、林业碳汇监测
冯伟 (1985—), 女, 副教授, 博士, 主要研究方向为遥感大数据多阶协同信息挖掘、多源数据深度集成感知与智能干扰识别、机器学习机理与方法、天地空一体化遥感智能应用
全英汇 (1981—), 男, 教授, 博士, 主要研究方向为智能感知、敏捷雷达
基金资助:
国家自然科学基金(62201438);国家自然科学基金(12005169);国家自然科学基金(62331019);陕西省自然科学基础研究计划(2021JC-23);陕西省林业科技创新重点项目(SXLK2022-02-8);陕西省社会科学界联合会项目(2022HZ1759);宁夏自然科学基金优秀青年项目(2024AAC05057)

An aligned subspace adaptive ensemble algorithm based on hyperspectral cross-scene transfer learning

Yijia SONG¹, Haiyan WANG², Wei FENG¹^,*, Yinghui QUAN¹

1. School of Electronic Engineering, Xidian University, Xi'an 710071, China
2. Shaanxi Academy of Forest, Xi'an 710016, China

Received:2024-01-04 Online:2025-02-25 Published:2025-03-18
Contact: Wei FENG

摘要/Abstract

摘要：

针对跨场景分类问题, 结合集成学习和域自适应思想, 提出一种对齐子空间自适应集成学习(aligned subspace adaptive ensemble learning, ASAEL)算法。首先, 对原始数据进行多次随机抽样以解决样本不平衡问题。然后, 将源域和目标域进行几何配准和统计配准, 以构建一个公共的子空间。最后, 对目标场景中的数据进行分类, 在保留有效信息的前提下, 通过计算多次分类结果, 得出最终的分类标签。所提ASAEL算法可以解决迁移过程中产生投影的不确定性和随机性。在两个数据集上的实验结果表明: 与传统的机器学习和域自适应方法相比, 所提算法的准确性显著提高。

关键词: 跨场景分类, 域自适应, 集成学习, 高光谱, 迁移学习

Abstract:

To solve the cross-scene classification problem, an aligned subspace adaptive ensemble learning algorithm combining ensemble learning and domain adaptive is proposed. Firstly, multiple random sampling is performed on the original data to solve the sample imbalance problem. Then, the source and target domains are geometrically and statistically aligned to construct a common subspace. Finally, the target scene data are classified, and the final classification labels are derived by calculating the classification results of multiple times while preserving the valid information. The proposed aligned subspace adaptive ensemble learning algorithm can solve the uncertainty and randomness of projection during the transfer process. Experimental results carried out on two datasets show that the proposed algorithm has a significant improvement in accuracy compared to traditional machine learning and domain adaptive methods.

Key words: cross-scene classification, domain adaptive, ensemble learning, hyperspectral, transfer learning

中图分类号:

TP751

宋怡佳, 王海燕, 冯伟, 全英汇. 基于高光谱跨场景迁移学习的集成对齐子空间自适应算法[J]. 系统工程与电子技术, 2025, 47(2): 369-375.

Yijia SONG, Haiyan WANG, Wei FENG, Yinghui QUAN. An aligned subspace adaptive ensemble algorithm based on hyperspectral cross-scene transfer learning[J]. Systems Engineering and Electronics, 2025, 47(2): 369-375.

图/表 10

图1

表1

图2

表2

图3

图4

表3

表4

图5

图6

参考文献 30

1	何同弟, 李见为. 基于自适应稀疏表示的高光谱遥感图像分类[J]. 系统工程与电子技术, 2013, 35 (9): 1994- 1998. doi: 10.3969/j.issn.1001-506X.2013.09.32
	HE T D , LI J W . Hyperspectral remote sensing image classification based on adaptive sparse representation[J]. Systems Engineering and Electronics, 2013, 35 (9): 1994- 1998. doi: 10.3969/j.issn.1001-506X.2013.09.32
2	冯伟, 龙以君, 全英汇, 等. 基于SMOTE和深度迁移卷积神经网络的多类不平衡遥感图像分类算法研究[J]. 系统工程与电子技术, 2023, 45 (12): 3715- 3725.
	FENG W , LONG Y J , QUAN Y H , et al. Multi-class imba-lance remote sensing image classification based on SMOTE and deep transfer convolutional neural network[J]. Systems Engineering and Electronics, 2023, 45 (12): 3715- 3725.
3	MENG Z H, YE M C, YAO F T, et al. Cross-scene hyperspectral image classification based on cycle-consistent adversarial networks[C]//Proc. of the IEEE International Geoscience and Remote Sensing Symposium, 2022: 1912-1915.
4	于纯妍, 徐铭阳, 宋梅萍, 等. 对抗与蒸馏耦合的高光谱遥感域自适应分类方法[J]. 遥感学报, 2024, 28 (1): 231- 246.
	YU C Y , XU M Y , SONG M P , et al. Unsupervised domain adaptive classification for hyperspectral remote sensing by adversary coupled with distillation[J]. National Remote Sensing Bulletin, 2024, 28 (1): 231- 246.
5	CHEN J B, YE M C, LU H J, et al. Cross-scene relationship mining with learning graph net for hyperspectral image classification[C]//Proc. of the 2nd International Conference on Artificial Intelligence and Computer Engineering, 2021: 519-524.
6	ZHANG Y X , LI W , TAO R , et al. Cross-scene hyperspectral image classification with discriminative cooperative alignment[J]. IEEE Trans. on Geoscience and Remote Sensing, 2021, 59 (11): 9646- 9660. doi: 10.1109/TGRS.2020.3046756
7	SUN B, SAENKO K. Deep CORAL: correlation alignment for deep domain adaptation[EB/OL]. [2023-11-22]. https://arXiv, 2016: 1607.01719.
8	FENG W , QUAN Y H , DAUPHIN G , et al. Semi-supervised rotation forest based on ensemble margin theory for the classification of hyperspectral image with limited training data[J]. Information Sciences, 2021, 575, 611- 638. doi: 10.1016/j.ins.2021.06.059
9	PAN S J , YANG Q . A survey on transfer learning[J]. IEEE Trans. on Knowledge and Data Engineering, 2009, 22 (10): 1345- 1359.
10	PAN S J , TSANG I W , KWOK J T , et al. Domain adaptation via transfer component analysis[J]. IEEE Trans. on Knowledge and Data Engineering, 2011, 22 (2): 199- 210.
11	LONG M S, WANG J, DING G G, et al. Transfer feature learning with joint distribution adaptation[C]//Proc. of the IEEE International Conference on Computer Vision, 2013: 2200-2207.
12	TZENG E, HOFFMAN J, ZHANG N, et al. Deep domain confusion: maximizing for domain invariance[EB/OL]. [2023-12-01]. https://arXiv, 2014: 1412.3474.
13	LONG M S, CAO Y, WANG J M, et al. Learning transferable features with deep adaptation networks[EB/OL]. [2023-12-01]. https://arXiv, 2015: 1502.02791.
14	WANG J D, CHEN Y Q, HU L S, et al. Stratified transfer learning for cross-domain activity recognition[C]//Proc. of the IEEE International Conference on Pervasive Computing and Communications, 2018.
15	SCHÖLKOPF B, PLATT J, HOFMANN T. A kernel method for the two-sample-problem[C]//Proc. of the Advances in Neural Information Processing Systems, 2007: 513-520.
16	ZHANG Y X , LI W , ZHANG M M , et al. Topological structure and semantic information transfer network for cross-scene hyperspectral image classification[J]. IEEE Trans. on Neural Networks and Learning Systems, 2023, 34 (6): 2817- 2830. doi: 10.1109/TNNLS.2021.3109872
17	ZHAO C H , QIN B A , FENG S , et al. An unsupervised domain adaptation method towards multi-level features and decision boundaries for cross-scene hyperspectral image classification[J]. IEEE Trans. on Geoscience and Remote Sensing, 2022, 5546216.
18	YU C Y , LIU C Y , SONG M P , et al. Unsupervised domain adaptation with content-wise alignment for hyperspectral ima-gery classification[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19, 5511705.
19	吕勤哲, 全英汇, 沙明辉, 等. 基于集成深度学习的有源干扰智能分类[J]. 系统工程与电子技术, 2022, 44 (12): 3595- 3602. doi: 10.12305/j.issn.1001-506X.2022.12.02
	LYU Q Z , QUAN Y H , SHA M H , et al. Ensemble deep learning-based intelligent classification of active jamming[J]. Systems Engineering and Electronics, 2022, 44 (12): 3595- 3602. doi: 10.12305/j.issn.1001-506X.2022.12.02
20	BREIMAN L . Bagging predictors[M]. New York: Machine Learning, 1996.
21	DONG S X , FENG W , QUAN Y H , et al. Deep ensemble CNN method based on sample expansion for hyperspectral image classification[J]. IEEE Trans. on Geoscience and Remote Sensing, 2022, 60, 5531815.
22	FENG W , HUANG W J , BAO W X . Imbalanced hyperspectral image classification with an adaptive ensemble method based on SMOTE and rotation forest with differentiated sampling rates[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16 (12): 1879- 1883. doi: 10.1109/LGRS.2019.2913387
23	陈华杰, 吕丹妮, 周枭, 等. 遥感图像小样本舰船识别跨域迁移学习算法[J]. 遥感学报, 2024, 28 (3): 793- 804.
	CHEN H J , LYU D N , ZHOU X , et al. Cross-domain transfer learning algorithm for few-shot ship recognition in remote-sensing images[J]. National Remote Sensing Bulletin, 2024, 28 (3): 793- 804.
24	FENG W , DAUPHIN G , HUANG W J , et al. Dynamic synthetic minority over-sampling technique-based rotation forest for the classification of imbalanced hyperspectral data[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12 (7): 2159- 2169. doi: 10.1109/JSTARS.2019.2922297
25	FRANCISCO C , ANTONIO J R , MARÍA J , et al. Addressing imbalance in multilabel classification: measures and random resampling algorithms[J]. Neurocomputing, 2015, 163, 3- 16. doi: 10.1016/j.neucom.2014.08.091
26	DEBES C , MERENTITIS A , HEREMANS R , et al. Hyperspectral and LiDAR data fusion: outcome of the 2013 GRSS data fusion contest[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7 (6): 2405- 2418. doi: 10.1109/JSTARS.2014.2305441
27	LE SAUX B , YOKOYA N , HANSCH R , et al. 2018 IEEE GRSS data fusion contest: multimodal land use classification[J]. IEEE Geoscience and Remote Sensing Magazine, 2018, 6 (1): 52- 54. doi: 10.1109/MGRS.2018.2798161
28	KARANTZALOS K, KARAKIZI C, KANDYLAKIS Z, et al. HyRANK hyperspectral satellite dataset I (version V001)[data set][J]. Journal of Photogrammetry and Remote Sensing, 2018. DOI: 10.5281/zenodo.1222202.
29	JOHAN A K S . Support vector machines: a nonlinear modelling and control perspective[J]. European Journal of Control, 2001, 7 (2): 311- 327.
30	COVER T , HART P . Nearest neighbor pattern classification[J]. IEEE Trans. on Information Theory, 1967, 13 (1): 21- 27. doi: 10.1109/TIT.1967.1053964

序号	名称	样本数量
序号	名称	休斯顿2013	休斯顿2018
1	健康草坪	198	1 858
2	受压草坪	190	6 068
3	树木	188	3 331
4	水域	182	24
5	居民楼	196	6 480
6	非居民楼建筑	191	37 778
7	道路	193	7 695
8	总数	1 338	63 234

序号	名称	样本数量
序号	名称	休斯顿2013	休斯顿2018
1	密集城市肌理	1 262	288
2	矿产开发地点	204	67
3	非灌溉工地	614	542
4	果树	150	79
5	橄榄树	1 768	1 401
6	针叶林	361	500
7	茂密叶绿植被	5 035	3 793
8	稀疏叶绿植被	6 374	2 803
9	植被稀疏地区	1 754	404
10	岩石和沙地	492	487
11	水域	1 612	1 393
12	沿海水域	398	451
13	总数	20 024	12 208

序号	分类算法
序号	SVM	KNN	TCA	CORAL	DCA	ASAEL
1	0	0.16	86.02	74.92	95.43	100.00
2	9.67	10.12	87.83	1.53	76.68	96.78
3	83.10	76.55	43.11	93.76	38.67	38.19
4	62.50	79.17	100.00	0	0	0
5	2.98	20.69	66.36	41.31	40.14	74.92
6	44.00	24.34	15.46	28.88	61.11	73.59
7	0.04	1.78	34.29	8.54	78.99	92.35

序号	分类算法
序号	SVM	KNN	TCA	CORAL	DCA	ASAEL
1	0	17.01	27.78	4.17	5.21	3.13
2	0	89.55	86.57	83.58	0	0
3	0	42.44	64.02	6.09	38.93	45.94
4	0	27.85	45.57	3.80	0	0
5	0	1.00	61.17	1.07	0	0
6	0	45.60	22.00	14.2	23.80	2.40
7	0	78.33	61.67	83.55	66.57	75.90
8	100	29.65	47.48	63.22	69.71	63.18
9	0	43.07	8.17	8.91	97.03	99.50
10	0	23.20	5.75	3.70	18.07	17.25
11	0	100.00	100.00	100.00	100.00	100.00
12	0	100.00	97.34	100.00	100.00	100.00

[1]	王俊智, 仲伟志, 肖丽君, 王鑫, 朱秋明, 林志鹏. 基于迁移学习的室内波束选择优化方法[J]. 系统工程与电子技术, 2024, 46(3): 1109-1115.
[2]	赵晓枫, 牛家辉, 刘春桐, 夏玉婷. 基于三维注意力与混合卷积的高光谱图像分类[J]. 系统工程与电子技术, 2023, 45(9): 2673-2680.
[3]	庞世燕, 郝京京, 邢立宁, 谭旭. 融合域自适应和弱监督策略的遥感影像建筑物提取算法[J]. 系统工程与电子技术, 2023, 45(6): 1616-1623.
[4]	冯伟, 龙以君, 全英汇, 邢孟道. 基于SMOTE和深度迁移卷积神经网络的多类不平衡遥感图像分类算法研究[J]. 系统工程与电子技术, 2023, 45(12): 3715-3725.
[5]	杨立儒, 刘永祥, 杨威. 基于迁移学习的雷达杂波幅度统计模型选择[J]. 系统工程与电子技术, 2022, 44(8): 2457-2467.
[6]	闫世瑛, 颜克斐, 方伟, 陆恒杨. 基于差分进化邻域自适应的大规模多目标算法[J]. 系统工程与电子技术, 2022, 44(7): 2112-2124.
[7]	孙晶明, 虞盛康, 孙俊. 基于元学习的雷达小样本目标识别方法及改进[J]. 系统工程与电子技术, 2022, 44(6): 1839-1845.
[8]	周晓玲, 张朝霞, 鲁雅, 王倩, 王琨琨. 基于改进R-FCN的SAR图像识别[J]. 系统工程与电子技术, 2022, 44(4): 1202-1209.
[9]	何立, 沈亮, 李辉, 王壮, 唐文泉. 强化学习中的策略重用: 研究进展[J]. 系统工程与电子技术, 2022, 44(3): 884-899.
[10]	吴涛, 王伦文, 朱敬成. 基于迁移学习和注意力机制的伪装图像分割[J]. 系统工程与电子技术, 2022, 44(2): 376-384.
[11]	吕勤哲, 全英汇, 沙明辉, 董淑仙, 邢孟道. 基于集成深度学习的有源干扰智能分类[J]. 系统工程与电子技术, 2022, 44(12): 3595-3602.
[12]	孟晋丽, 王亚峰, 杨靖北, 王宁. 干扰极化捷变下的空域自适应滤波[J]. 系统工程与电子技术, 2022, 44(11): 3305-3312.
[13]	马骏, 杨镜宇, 邹立岩. 基于Stacking集成元模型的作战体系能力图谱生成方法[J]. 系统工程与电子技术, 2022, 44(1): 154-163.
[14]	陈善学, 王欣欣. 基于空间预处理联合稀疏表示高光谱图像分类[J]. 系统工程与电子技术, 2021, 43(9): 2422-2429.
[15]	丁嘉辉, 汤建龙, 于正洋. 轻量化的增量式集成学习算法设计[J]. 系统工程与电子技术, 2021, 43(4): 861-867.

基于高光谱跨场景迁移学习的集成对齐子空间自适应算法

An aligned subspace adaptive ensemble algorithm based on hyperspectral cross-scene transfer learning

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 30

相关文章 15

编辑推荐

Metrics

本文评价