主动学习策略融合算法在高光谱图像分类中的应用

doi:10.11959/j.issn.1000-436x.2018067

通信学报 ›› 2018, Vol. 39 ›› Issue (4): 91-99.doi: 10.11959/j.issn.1000-436x.2018067

主动学习策略融合算法在高光谱图像分类中的应用

崔颖¹,徐凯¹,陆忠军²,刘述彬²,王立国¹

¹ 哈尔滨工程大学信息与通信工程学院，黑龙江哈尔滨 150001
² 黑龙江省农业科学院遥感技术中心，黑龙江哈尔滨 150086

修回日期:2018-01-03 出版日期:2018-04-01 发布日期:2018-04-29
作者简介:崔颖（1979-），女，黑龙江哈尔滨人，博士，哈尔滨工程大学副教授，主要研究方向为遥感图像处理、智能信号处理、无线传感器网络。|徐凯（1992-），男，浙江绍兴人，哈尔滨工程大学硕士生，主要研究方向为遥感图像处理、机器学习。|陆忠军（1975-），男，黑龙江哈尔滨人，黑龙江省农业科学院遥感技术中心副研究员，主要研究方向为农业遥感图像分析。|刘述彬（1963-），男，黑龙江哈尔滨人，黑龙江省农业科学院遥感技术中心研究员，主要研究方向为农业遥感图像建模、数据分析。|王立国（1974-），男，黑龙江哈尔滨人，哈尔滨工程大学教授，主要研究方向为图像/信号处理技术、机器学习与模式识别理论。
基金资助:
国家自然科学基金资助项目(61675051);教育部博士点基金资助项目(20132304110007)

Combination strategy of active learning for hyperspectral images classification

Ying CUI¹,Kai XU¹,Zhongjun LU²,Shubin LIU²,Liguo WANG¹

¹ College of Information and Communications Engineering,Harbin Engineering University,Harbin 150001,China
² Remote Sensing Technology Center of Heilongjiang Academy of Agricultural Sciences,Harbin 150086,China

Revised:2018-01-03 Online:2018-04-01 Published:2018-04-29
Supported by:
The National Natural Science Foundation of China(61675051);Education Ministry Doctoral Research Foundation of China(20132304110007)

摘要/Abstract

摘要：

针对传统主动学习单一策略算法在挑选最有价值未标记样本时出现的抖动和不稳定的现象，引入集成学习（ensemble learning）分类器的加权组合思想，提出一种基于组合策略的联合挑选（ESAL）方法，将模型的组合衍生至策略的组合，从而实现单一模型多策略的融合，获得更高的稳定性。通过对高光谱遥感图像分类结果的分析可以看出，在获得相同精度阈值时，ESAL 算法相对于单一策略算法最高可节省成本 25.4%，抖动频率减少至原来的16.67%，抖动明显改善，体现出ESAL算法良好的稳定性。

关键词: 主动学习, 集成学习, 高光谱图像, 策略组合

Abstract:

In order to improve the phenomena of jitter and instability of the traditional active learning single strategy algorithm in selecting the most valuable unlabeled samples.The idea of weighted combination of ensemble learning classifier and proposes a joint selection based on the combination strategy method (ESAL,ensemble strategy active learning) was introduced,the combination of the model was extended to the combination of the strategy so as to achieve the fusion of multiple strategies in a single model and achieve higher stability.By analyzing the classification results of hyperspectral remote sensing images,the ESAL algorithm can save 25.4% of the cost compared with the single strategy algorithm and reduce the jitter frequency to 16.67% when the same accuracy threshold is obtained,and the jitter is obviously improved.ESAL algorithm is out of good stability.

Key words: active learning, ensemble learning, hyperspectral image, strategy combination

中图分类号:

TP302

崔颖,徐凯,陆忠军,刘述彬,王立国. 主动学习策略融合算法在高光谱图像分类中的应用[J]. 通信学报, 2018, 39(4): 91-99.

Ying CUI,Kai XU,Zhongjun LU,Shubin LIU,Liguo WANG. Combination strategy of active learning for hyperspectral images classification[J]. Journal on Communications, 2018, 39(4): 91-99.

图/表 9

图1

图2

图3

图4

表1

图5

表2

表3

图6

参考文献 18

[1]	WEI D M M C . Active learning via multi-view and local proximity co-regularization for hyperspectral image classification[J]. IEEE Journal of Selected Topics in Signal Processing, 2011,5(3): 618-628.
[2]	MITRA P , UMA SHANKAR B , PAL S K . Segmentation of multispectral remote sensing images using active support vector machines[J]. Pattern Recognition Letters, 2004,25(9): 1067-1074.
[3]	JOSHI A J , PORIKLI F , PAPANIKOLOPOULOS N . Multi-class active learning for image classification[C]// 2009 IEEE Symposium on The Computer Vision and Pattern Recognition. 2009: 2372-2379.
[4]	TUIA D , VOLPI M , COPA L ,et al. A Survey of active learning algorithms for supervised remote sensing image classification[J]. IEEE Journal of Selected Topics in Signal Processing, 2011,5(3): 606-617.
[5]	李宠, 谷琼, 蔡之华 . 改进的主动学习算法及在高光谱分类中的应用[J]. 华中科技大学学报, 2013,41(11): 274-278.
	LI C , GU Q , CAI Z H . Improved active learning algorithm and its application in hyperspectral classification[J]. Journal of Huazhong University of Science and Technology, 2013,41(11): 274-278.
[6]	LI M , WANG R , TANG K . Combining semi-supervised and active learning for hyperspectral image classification[C]// 2013 IEEE Symposium on The Computational Intelligence and Data Mining (CIDM). 2013: 89-94.
[7]	WAN L , TANG K , LI M ,et al. Collaborative active and semisupervised learning for hyperspectral remote sensing image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015,53(5): 2384-2396.
[8]	WANG Z , DU B , ZHANG L ,et al. A novel semisupervised active-learning algorithm for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017,55(6): 3071-3083.
[9]	王立国, 李阳 . 融合主动学习的高光谱图像半监督分类[J]. 哈尔滨工程大学学报, 2017,38(8): 1322-1327.
	WANG L G , LI Y . Semi-supervised classification for hyperspectral image collaborating with active learning algorithm[J]. Journal of Harbin Engineering University, 2017,38(8): 1322-1327.
[10]	SAMIAPPAN S , MOORHEAD R J . Semi-supervised co-training and active learning framework for hyperspectral image classification[C]// the Geoscience and Remote Sensing Symposium (IGARSS), 2015: 401-404.
[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 . Semi-supervised classification for hyperspectral image based on improved tri-training method[J]. Journal of Harbin Engineering University, 2016,37(6): 849-854.
[12]	赵建华, 刘宁 . 结合主动学习策略的半监督分类算法[J]. 计算机应用研究, 2015,32(8): 2295-2298.
	ZHAO J H , LIU N . Semi-supervised classification algorithm based on active learning strategies[J]. Application Research of Computers, 2015,32(8): 2295-2298.
[13]	ZHANG Z , PASOLLI E , CRAWFORD M M ,et al. An active learning framework for hyperspectral image classification using hierarchical segmentation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016,9(2): 640-654.
[14]	陈荣, 曹永锋, 孙洪 . 基于主动学习和半监督学习的多类图像分类[J]. 自动化学报, 2011,37(8): 954-962.
	CHEN R , CAO Y F , SUN H . Multi-class image classification with active learning and semi-supervised learning[J]. Acta Automatica Sinica, 2011,37(8): 954-962.
[15]	韩松来, 张辉, 周华平 . 基于关联度函数的决策树分类算法[J]. 计算机应用, 2005,25(11): 2655-2657.
	HAN S L , ZHANG H , ZHOU H P . Association function algorithm for decision tree[J]. Journal of Computer Applications, 2005,25(11): 2655-2657.
[16]	QUINLAN J R . Induction of decision trees[J]. Machine Learning, 1986,1(1): 81-106.
[17]	LIU Y , YAO X . Ensemble learning via negative correlation[J]. Neural Networks, 1999,12(10): 1399-1404.
[18]	DIETTERICH T G . Ensemble learning[J]. The Handbook of Brain Theory and Neural Networks, 2002,2: 110-125.

数据集	单基策略/次	ESAL/次
Indian Pines	12	2
KSC	6	1

OA		数据集标记成本数量/个
OA	aEQB	BvSB	ESAL
0.73	145	145	125
0.74	195	195	165
0.75	205	215	185
0.76	235	235	215
0.77	255	245	235
0.78	295	265	255
0.79	345	305	295
0.80	405	335	335
0.81	445	405	385
0.82	515	485	435
0.83	595	595	545
0.84	685	745	665
0.85	825	905	795

OA		数据集标记成本数量/个
OA	aEQB	BvSB	ESAL
0.81	5	5	5
0.82	15	15	15
0.83	15	15	15
0.84	25	25	15
0.85	35	25	25
0.86	45	35	25
0.87	55	45	45
0.88	75	55	55
0.89	95	85	65
0.90	115	95	85
0.91	145	125	105
0.92	175	195	145
0.93	315	285	235

主动学习策略融合算法在高光谱图像分类中的应用

Combination strategy of active learning for hyperspectral images classification

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 18

相关文章 11

Metrics

推荐阅读 0

[1]	戴千一, 张斌, 郭松, 徐开勇. 基于多分类器集成的区块链网络层异常流量检测方法[J]. 通信学报, 2023, 44(3): 66-80.
[2]	高红民, 曹雪莹, 陈忠昊, 花再军, 李臣明, 陈月. 基于多尺度近端特征拼接网络的高光谱图像分类方法[J]. 通信学报, 2021, 42(2): 92-102.
[3]	何遵文, 侯帅, 张万成, 张焱. 通信特定辐射源识别的多特征融合分类方法[J]. 通信学报, 2021, 42(2): 103-112.
[4]	孙晓川,李志刚,张明辉,桂冠. 基于集成学习的广域光骨干网多信道传输质量预测方法[J]. 通信学报, 2020, 41(9): 1-7.
[5]	高红民,曹雪莹,杨耀,花再军,李臣明. 基于CNN的双边融合网络在高光谱图像分类中的应用[J]. 通信学报, 2020, 41(11): 132-140.
[6]	李涛,郭渊博,琚安康. 融合对抗主动学习的网络安全知识三元组抽取[J]. 通信学报, 2020, 41(10): 80-91.
[7]	刘敬，谷利泽，钮心忻，杨义先. 基于单分类支持向量机和主动学习的网络异常检测研究[J]. 通信学报, 2015, 36(11): 136-146.
[8]	余晓东,雷英杰,宋亚飞,岳韶华,胡军红. 基于直觉模糊核匹配追踪集成的目标识别方法[J]. 通信学报, 2015, 36(10): 165-171.
[9]	吴健,盛胜利,赵朋朋,崔志明. 最小差异采样的主动学习图像分类方法[J]. 通信学报, 2014, 35(1): 107-114.
[10]	王一鹏,云晓春,张永铮,李书豪. 基于主动学习和SVM方法的网络协议识别技术[J]. 通信学报, 2013, 34(10): 135-142.
[11]	王一鹏1,2,3，云晓春1,3，张永铮3，李书豪3. 基于主动学习和SVM方法的网络协议识别技术[J]. 通信学报, 2013, 34(10): 16-142.