融合光谱度量标记迁移和Tri-training的高光谱遥感图像半监督分类算法

doi:10.11959/j.issn.2096-0271.2022084

大数据 ›› 2023, Vol. 9 ›› Issue (6): 72-89.doi: 10.11959/j.issn.2096-0271.2022084

融合光谱度量标记迁移和Tri-training的高光谱遥感图像半监督分类算法

曹峰¹, 李文涛¹, 骆剑承², 李德玉¹^,³, 钱宇华¹^,³^,⁴, 白鹤翔¹, 张超¹

¹ 山西大学计算机与信息技术学院（大数据学院），山西太原 030006
² 中国科学院空天信息创新研究院遥感科学国家重点实验室，北京 100101
³ 山西大学计算智能与中文信息处理教育部重点实验室，山西太原 030006
⁴ 山西大学大数据与产业研究院，山西太原 030006

出版日期:2023-11-15 发布日期:2023-11-01
作者简介:曹峰（1980- ），男，博士，山西大学计算机与信息技术学院（大数据学院）副教授、大数据系主任，主要研究方向为人工智能与数据挖掘。
李文涛（1994- ），男，山西大学计算机与信息技术学院（大数据学院）硕士生，主要研究方向为数据挖掘。
骆剑承（1970- ），男，博士，中国科学院空天信息创新研究院遥感科学国家重点实验室研究员，中国科学院大学教授，主要研究方向为遥感大数据智能计算。
李德玉（1965- ），男，山西大学计算机与信息技术学院（大数据学院）教授，山西大学计算机与信息技术学院党委书记，山西大学计算智能与中文信息处理教育部重点实验室副主任，主要研究方向为人工智能与数据挖掘。
钱宇华（1976- ），男，山西大学计算机与信息技术学院（大数据学院）教授，山西大学科学技术处处长，山西大学大数据科学与产业研究院负责人，山西大学计算智能与中文信息处理教育部重点实验室副主任，主要研究方向为人工智能与数据挖掘。
白鹤翔（1980- ），男，博士，山西大学计算机与信息技术学院（大数据学院）副教授，主要研究方向为人工智能与数据挖掘。
张超（1989- ），男，博士，山西大学计算机与信息技术学院（大数据学院）副教授，主要研究方向为数据挖掘与粒计算。
基金资助:
国家自然科学基金资助项目(62072291);国家自然科学基金资助项目(42071316);国家自然科学基金资助项目(62072294);国家自然科学基金资助项目(61672332);国家自然科学基金资助项目(41871286);国家自然科学基金资助项目(61806116);山西省重点研发计划项目(201903D421003);山西省重点研发计划项目(201903D421041);山西省教育厅科技成果转化培育项目(2020CG001)

Semi-supervised classification algorithm for hyperspectral remote sensing images fusing spectral measure-based label transfer and tri-training

Feng CAO¹, Wentao LI¹, Jiancheng LUO², Deyu LI¹^,³, Yuhua QIAN¹^,³^,⁴, Hexiang BAI¹, Chao ZHANG¹

¹ School of Computer and Information Technology (School of Big Data), Shanxi University, Taiyuan 030006, China
² State Key Laboratory of Remote Sensing Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, China
³ Key Laboratory of Computational Intelligence and Chinese Information Processing, Ministry of Education, Shanxi University, Taiyuan 030006, China
⁴ Institute of Big Data and Industry, Shanxi University, Taiyuan 030006, China

Online:2023-11-15 Published:2023-11-01
Supported by:
The National Natural Science Foundation of China(62072291);The National Natural Science Foundation of China(42071316);The National Natural Science Foundation of China(62072294);The National Natural Science Foundation of China(61672332);The National Natural Science Foundation of China(41871286);The National Natural Science Foundation of China(61806116);Key Research and Development Plan of Shanxi Province(201903D421003);Key Research and Development Plan of Shanxi Province(201903D421041);Scientific and Technological Transformative Project of Shanxi Provincial Department of Education(2020CG001)

摘要/Abstract

摘要：

针对海量的高光谱遥感图像光谱和丰富的空间信息中可用于分类的有标记样本远少于无标记样本的数据特性，提出了一种融合光谱度量标记迁移和Tri-training的高光谱遥感图像半监督光谱-空间分类算法。该算法提出了一种基于光谱度量的标记迁移方法，通过结合迁移标记和Tri-training预测标记进行扩充样本标记预测，提高了扩充样本标记的准确性。同时，该算法基于空间相关性选择扩充样本，综合运用光谱和空间特征提升图像分类的精度。在两个公开的高光谱遥感图像数据集上进行了实验，结果表明该算法优于基于Tri-training算法的高光谱遥感图像的分类性能。

关键词: 高光谱图像分类, 半监督分类, 纹理特征, 光谱度量, Tri-training算法

Abstract:

Aimed at the problem that a large number of hyperspectral remote sensing images were rich in spectral and spatial information, and the labeled samples available for image classification were far less than unlabeled samples, a semisupervised spectral-spatial classification algorithm was proposed by fusing spectral measure-based label transfer and Tri-training.A spectral measure-based label transfer method was proposed for our algorithm.The transferred labels and predicted labels for Tri-training algorithm were used to predict the labels of expanded unlabeled samples, which can promoted the prediction accuracies of labels for expanded unlabeled samples.Meanwhile, our algorithm selectel expanded samples based on spatial correlation, and used spectral and spatial features to improve the accuracy of image classification.Experimental study was executed on two public hyperspectral remote sensing image datasets, and the results showed that the proposed algorithm outperform tri-training algorithm.

Key words: classification of hyperspectral image, semi-supervised classification, texture feature, spectral measurement, tri-training algorithm

中图分类号:

TP751

曹峰, 李文涛, 骆剑承, 李德玉, 钱宇华, 白鹤翔, 张超. 融合光谱度量标记迁移和Tri-training的高光谱遥感图像半监督分类算法[J]. 大数据, 2023, 9(6): 72-89.

Feng CAO, Wentao LI, Jiancheng LUO, Deyu LI, Yuhua QIAN, Hexiang BAI, Chao ZHANG. Semi-supervised classification algorithm for hyperspectral remote sensing images fusing spectral measure-based label transfer and tri-training[J]. Big Data Research, 2023, 9(6): 72-89.

图/表 14

图1

图2

图3

图4

表1

图5

表2

表3

表4

图6

图7

表5

图8

图9

参考文献 23

[1]	骆剑承, 胡晓东, 吴田军 ,等. 高分遥感驱动的精准土地利用与土地覆盖变化信息智能计算模型与方法研究[J]. 遥感学报, 2021,25(7): 1351-1373.
	LUO J C , HU X D , WU T J ,et al. Research on intelligent calculation model and method of precision land use/cover change information driven by high-resolution remote sensing[J]. Journal of Remote Sensing, 2021,25(7): 1351-1373.
[2]	王增茂, 杜博, 张良培 ,等. 基于纹理特征和形态学特征融合的高光谱影像分类法[J]. 光子学报, 2014,43(8).
	WANG Z M , DU B , ZHANG L P ,et al. Based on texture feature and extend morphological profile fusion for hyperspectral image classification[J]. Acta PhotonicaSinica, 2014,43(8).
[3]	王杰, 张松岩, 梁吉业 . 融合一致性正则与流形正则的半监督深度学习算法[J]. 大数据, 2022,8(3): 103-114.
	WANG J , ZHANG S Y , LIANG J Y . A semi-supervised deep learning algorithm combining consistency regularization and manifold regularization[J]. Big Data Research, 2022,8(3): 103-114.
[4]	SHAHSHAHANI B M , LANDGREBE D A . The effect of unlabeled samples in reducing the small sample size problem and mitigating the Hughes phenomenon[J]. IEEE Transactions on Geoscience and Remote Sensing, 1994,32(5): 1087-1095.
[5]	焦洪赞, 王少宇, 彭正洪 . 基于条件随机场的光谱相似性匹配高光谱遥感影像聚类方法[J]. 武汉大学学报(工学版), 2016,49(6): 937-943,948.
	JIAO H Z , WANG S Y , PENG Z H . A spectral similarity matching classifier based on conditional random field for hyperspectral remote sensing imagery[J]. Engineering Journal of Wuhan University, 2016,49(6): 937-943,948.
[6]	程志会, 谢福鼎 . 基于空间特征与纹理信息的高光谱图像半监督分类[J]. 测绘通报, 2016(12): 56-59,73.
	CHENG Z H , XIE F D . Semi-supervised classification for hyperspectral image based on spatial features and texture information[J]. Bulletin of Surveying and Mapping, 2016(12): 56-59,73.
[7]	何浩, 沈永林, 刘修国 ,等. 空间-光谱约束的图半监督高光谱影像分类算法[J]. 国土资源遥感, 2016,28(3): 31-36.
	HE H , SHEN Y L , LIU X G ,et al. Spatialspectral constrained graph-based semisupervised classification for hyperspectral image[J]. Remote Sensing for Natural Resources, 2016,28(3): 31-36.
[8]	王俊淑, 江南, 张国明 ,等. 高光谱遥感图像DE-self-training半监督分类算法[J]. 农业机械学报, 2015,46(5): 239-244.
	WANG J S , JIANG N , ZHANG G M ,et al. Semi-supervised classification algorithm for hyperspectral remote sensing image based on DE-self-training[J]. Transactions of the Chinese Society for Agricultural Machinery, 2015,46(5): 239-244.
[9]	王春阳 . 基于信息熵的自训练半监督高光谱遥感影像分类研究[D]. 焦作:河南理工大学, 2015.
	WANG C Y . Research on self-training semi-supervised hyperspectral remote sensing image classification based on information entropy[D]. Jiaozuo:Henan Polytechnic University, 2015.
[10]	ROMASZEWSKI M , GLOMB P , CHLOEW M . Semi-supervised hyperspectral classification from a small number of training samples using a co-training approach[J]. ISPRS:Journal of Photogrammetry ＆ Remote Sensing, 2016,121: 60-76.
[11]	王立国, 杨月霜, 刘丹凤 . 基于改进三重训练算法的高光谱图像半监督分类[J]. 哈尔滨工程大学学报, 2016,37(6): 849-854.
	WANG L G , YANG Y S , LIU D F . Semisupervised classification for hyperspectral image based on improved tri-training method[J]. Journal of Harbin Engineering University, 2016,37(6): 849-854.
[12]	宋国娇 . 高光谱图像降维及半监督分类算法研究[D]. 哈尔滨:哈尔滨工程大学, 2017.
	SONG G J . Research on dimensionality reduction and semi-supervised classification algorithm of hyperspectral images[D]. Harbin:Harbin Engineering University, 2017.
[13]	ZHOU Z H , LI M . Tri-training:exploiting unlabeled data using three classifiers[J]. IEEE Transactions on Knowledge and Data Engineering, 2005,17(11): 1529-1541.
[14]	GOLDMAN S , YAN Z . Enhancing supervised learning with unlabeled data[D]. Washington:Washington University, 2001.
[15]	胡梦林 . 基于混沌杜鹃搜索算法的高光谱影像波段选择和半监督分类[D]. 武汉:武汉大学, 2018.
	HU M L . Band selection and semi-supervised classification of hyperspectral images based on chaotic cuckoo search algorithm[D]. Wuhan:Wuhan University, 2018.
[16]	HALL M A . Feature selection for discrete and numeric class machine learning[C]// Proceedings of the 17th International Conference on Machine Learning.[S.l.]: Morgan Kaufmann Publishers Inc, 2000: 584-596.
[17]	HARALICK R M . Statistical and structural approaches to texture[J]. Proceedings of the IEEE, 1979,67(5): 786-804.
[18]	王立国, 马骏宇, 李阳 . 联合多种空间信息的高光谱半监督分类方法[J]. 哈尔滨工程大学学报, 2021,42(2): 280-285.
	WANG L G , MA J Y , LI Y . Hyperspectral semi-supervised classification algorithm considering multiple spatial information[J]. Journal of Harbin Engineering University, 2021,42(2): 280-285.
[19]	关世豪, 杨桄, 卢珊 ,等. 基于改进阶梯网络的高光谱半监督分类算法[J]. 激光与光电子学进展, 2021,58(24): 480-490.
	GUAN S H , YANG G , LU S ,et al. Hyperspectral semi-supervised classification algorithm based on improved ladder network[J]. Laser ＆Optoelectronics Progress, 2021,58(24): 480-490.
[20]	李彩虹, 赵祎霏 . 一种高光谱图像的半监督分类方法[J]. 测绘通报, 2018(2): 46-49.
	LI C H , ZHAO Y F . A semi-supervised classification method for hyperspectral images[J]. Bulletin of Surveying and Mapping, 2018(2): 46-49.
[21]	崔颖, 王雪婷, 陆忠军 ,等. 改进M-training算法的高光谱图像分类[J]. 哈尔滨工程大学学报, 2018,39(10): 1688-1694.
	CUI Y , WANG X T , LU Z J ,et al. Hyperspectral image classification based on improved M-training algorithm[J]. Journal of Harbin Engineering University, 2018,39(10): 1688-1694.
[22]	宋国娇 . 高光谱图像降维及半监督分类算法研究[D]. 哈尔滨:哈尔滨工程大学, 2017.
	SONG G J . Research on dimensionality reduction and semi-supervised classification algorithm of hyperspectral images[D]. Harbin:Harbin Engineering University, 2017.
[23]	罗甫林 . 高光谱图像稀疏流形学习方法研究[J]. 测绘学报, 2017,46(3): 400.
	LUO F L . Sparse manifold learning for hyperspectral imagery[J]. Acta Geodaetica et CartographicaSinica, 2017,46(3): 400.

评价指标	SMT-RSK	SMT-RF	SMT-SVM	SMT-KNN	T-RSK	T-RF	T-SVM	T-KNN
OA	90.81%	88.13%	85.24%	86.40%	81.83%	79.83%	75.44%	77.12%
AA	92.22%	89.58%	86.88%	88.19%	83.25%	81.94%	77.55%	78.49%
Kappa	0.8691	0.8277	0.7892	0.8036	0.7611	0.6888	0.6544	0.6805

评价指标	SMT-RSK	3CLL	CMSI	SFCM	MT	LGC
OA	90.81%	95.64%	85.71%	88.00%	85.27%	87.32%
AA	92.22%	90.17%	88.88%	90.14%	87.41%	—
Kappa	0.8691	—	0.8318	0.883	0.8255	0.8410

评价指标	SMT-RSK	SMT-RF	SMT-SVM	SMT-KNN	T-RSK	T-RF	T-SVM	T-KNN
OA	92.47%	89.88%	87.55%	88.31%	83.59%	81.93%	76.81%	79.56%
AA	93.92%	91.14%	88.96%	90.23%	85.27%	83.11%	77.99%	81.94%
Kappa	0.8874	0.8362	0.8042	0.8186	0.7815	0.6954	0.6713	0.7001

评价指标	SMT-RSK	CMSI	FWA	SDME	SFCM	MT
OA	92.47%	87.68%	87.44%	89.4%	91.35%	66.89%
AA	93.92%	92.14%	—	86.6%	—	73.76%
Kappa	0.8874	0.8283	0.83	—	0.8974	0.5802

数据集		SM			T-RSK			SMT-RSK
数据集	OA	AA	Kappa	OA	AA	Kappa	OA	AA	Kappa
India Pines	71.42%	75.33%	0.6919	81.83%	83.25%	0.7611	90.81%	92.22%	0.8691
Pavia University	75.67%	78.41%	0.7178	83.59%	85.27%	0.7815	92.47%	93.92%	0.8874

融合光谱度量标记迁移和Tri-training的高光谱遥感图像半监督分类算法

Semi-supervised classification algorithm for hyperspectral remote sensing images fusing spectral measure-based label transfer and tri-training

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 23

相关文章 1

Metrics

推荐阅读 0