主动学习与半监督技术相结合的海冰图像分类
|
摘要
针对海冰遥感图像分类问题中标签样本获取困难、标注成本较高导致海冰分类精度难以提高的问题,提出了一种主动学习与半监督学习相结合的方式用于海冰分类。首先,利用基于不确定性准则和多样性准则进行主动学习方法,选择一批最具信息量的标签样本建立标签样本集;其次,充分利用大量的未标签样本信息,并融合主动学习采样的思想选出部分具有代表性且分布在支持向量周边的半标签样本,建立半监督分类模型;最后,将主动学习方法和直推式支持向量机相结合构建分类模型实现海冰图像分类。实验结果表明,相对于其他方法,该方法在只有少量标签样本的情况下,可以获得更高的分类精度,该方式可有效解决遥感海冰分类问题。
Aiming at the problem of difficulty in improving the classification accuracy of sea ice caused by the difficulty of obtaining labeled samples and high labeling cost in the classification of sea ice remote sensing images,this paper puts forward a method of combining active learning(AL)with semi-supervised learning(SSL)for the classification of sea ice.Firstly,this paper adopts the method of active learning based on the criterion of uncertainty and diversity and chooses a batch of most informative labeled samples to establish a set of labeled samples.Secondly,it makes the best of a great deal of unlabeled sample information,picks out a part of representative unlabeled samples distributed around support vectors(SVs)according to the idea of active learning sampling and builds a semi-supervised classification model.Finally,it combines the method of AL with transductive support vector machine to build a classification model and realize the classification of sea ice images.Experimental results show that the method suggested in this paper can achieve higher classification accuracy than other methods in the case of only a small number of labeled samples.The method can be effectively used to solve the classification of remote sensing sea ice.
Aiming at the problem of difficulty in improving the classification accuracy of sea ice caused by the difficulty of obtaining labeled samples and high labeling cost in the classification of sea ice remote sensing images,this paper puts forward a method of combining active learning(AL)with semi-supervised learning(SSL)for the classification of sea ice.Firstly,this paper adopts the method of active learning based on the criterion of uncertainty and diversity and chooses a batch of most informative labeled samples to establish a set of labeled samples.Secondly,it makes the best of a great deal of unlabeled sample information,picks out a part of representative unlabeled samples distributed around support vectors(SVs)according to the idea of active learning sampling and builds a semi-supervised classification model.Finally,it combines the method of AL with transductive support vector machine to build a classification model and realize the classification of sea ice images.Experimental results show that the method suggested in this paper can achieve higher classification accuracy than other methods in the case of only a small number of labeled samples.The method can be effectively used to solve the classification of remote sensing sea ice.
引文
[1]张晰,张杰,孟俊敏,等.基于极化散射特征的极化合成孔径雷达海冰分类方法研究:以渤海海冰分类为例[J].海洋学报,2013,35(5):95-101.
[2]罗征宇,孙林.基于高光谱影像的渤海海冰遥感监测研究[J].测绘科学,2012,37(1):54-55.
[3]VAPNIK V.The nature of statistical learning theory[M].Berlin:Springer Science &Business Media,2013.
[4]HSU C W,LIN C J.A comparison of methods for multiclass support vector machines[C]∥IEEE Transactions on Neural Networks.2002:415-425.
[5]SETTLES B.Active learning literature survey[J].University of Wisconsinmadison,2009,39(2):127-131.
[6]JOSHI A J,PORIKLI F,PAPANIKOLOPOULOS N.Multi-class active learning for image classification[C]∥Computer Vision and Pattern Recognition,2009.CVPR 2009.IEEE Conference on.IEEE,2009:2372-2379.
[7]ZHU X.Semi-supervised learning literature survey[EB/OL](2008-07-19)[2017-10-19]https://www.docin.com/p-1907314395.html.
[8]WANG Z M,DU B,ZHANG L F,et al.A batch-mode active learning framework by querying discriminative and representative samples for hyperspectral image classification[J].Neurocomputing,2016,179(C):88-100.
[9]JOACHIMS T.Transductive inference for text classification using support vector machines[C]∥Proceeding of the sixteenth international conference on machine learning.San Francisco:Morgan Kaufmann,1999.200-209.
[10]ANGLUIN D.Queries and concept learning[J].Machine Learning,1988,2(4):319-342.
[11]LI M,SETHI I K.Confidence-based active learning[J].IEEE Transactions on Pattern Analysis &Machine Intelligence,2006,28(8):1251-61.
[12]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.
[13]HAN Y L,REN J,HONG Z,et al.Active learning algorithms for the classification of hyperspectral sea ice images[J].Mathematical Problems in Engineering,2015.
[14]BRUZZONE L,CHI M,MARCONCINI M.A novel transductive SVM for semisupervised classification of remote sensing images[J].IEEE Transactions on Geoscience &Remote Sensing,2006,44(11):3363-3373.
[15]GOMEZ-CHOVA L,CAMPS-VALLS G,MUNOZ-MARI J,et al.Semisupervised image classification with laplacian support vector machines[J].IEEE Geoscience &Remote Sensing Letters,2008,5(3):336-340.
[16]PERSELLO C,BRUZZONE L.Active and semisupervised learning for the classification of remote sensing images[J].IEEE Transactions on Geoscience &Remote Sensing,2014,52(11):6937-6956.
[17]BLINKER K.Incorporating diversity in active learning with support vector machines[C]∥Machine Learning,Proceedings of the Twentieth International Conference.DBLP,2003:59-66.
[2]罗征宇,孙林.基于高光谱影像的渤海海冰遥感监测研究[J].测绘科学,2012,37(1):54-55.
[3]VAPNIK V.The nature of statistical learning theory[M].Berlin:Springer Science &Business Media,2013.
[4]HSU C W,LIN C J.A comparison of methods for multiclass support vector machines[C]∥IEEE Transactions on Neural Networks.2002:415-425.
[5]SETTLES B.Active learning literature survey[J].University of Wisconsinmadison,2009,39(2):127-131.
[6]JOSHI A J,PORIKLI F,PAPANIKOLOPOULOS N.Multi-class active learning for image classification[C]∥Computer Vision and Pattern Recognition,2009.CVPR 2009.IEEE Conference on.IEEE,2009:2372-2379.
[7]ZHU X.Semi-supervised learning literature survey[EB/OL](2008-07-19)[2017-10-19]https://www.docin.com/p-1907314395.html.
[8]WANG Z M,DU B,ZHANG L F,et al.A batch-mode active learning framework by querying discriminative and representative samples for hyperspectral image classification[J].Neurocomputing,2016,179(C):88-100.
[9]JOACHIMS T.Transductive inference for text classification using support vector machines[C]∥Proceeding of the sixteenth international conference on machine learning.San Francisco:Morgan Kaufmann,1999.200-209.
[10]ANGLUIN D.Queries and concept learning[J].Machine Learning,1988,2(4):319-342.
[11]LI M,SETHI I K.Confidence-based active learning[J].IEEE Transactions on Pattern Analysis &Machine Intelligence,2006,28(8):1251-61.
[12]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.
[13]HAN Y L,REN J,HONG Z,et al.Active learning algorithms for the classification of hyperspectral sea ice images[J].Mathematical Problems in Engineering,2015.
[14]BRUZZONE L,CHI M,MARCONCINI M.A novel transductive SVM for semisupervised classification of remote sensing images[J].IEEE Transactions on Geoscience &Remote Sensing,2006,44(11):3363-3373.
[15]GOMEZ-CHOVA L,CAMPS-VALLS G,MUNOZ-MARI J,et al.Semisupervised image classification with laplacian support vector machines[J].IEEE Geoscience &Remote Sensing Letters,2008,5(3):336-340.
[16]PERSELLO C,BRUZZONE L.Active and semisupervised learning for the classification of remote sensing images[J].IEEE Transactions on Geoscience &Remote Sensing,2014,52(11):6937-6956.
[17]BLINKER K.Incorporating diversity in active learning with support vector machines[C]∥Machine Learning,Proceedings of the Twentieth International Conference.DBLP,2003:59-66.