摘要: 对尽量少的样本进行人工标注并获得较好的分类性能是图像分类应用的一个关键问题。针对标注样本选择,提出了一种综合样本不确定性度量和代表性度量的主动学习样本选择准则。基于最优标号和次优标号(Best vs.second-best,BvSB)的主动学习方法构建不确定性度量,利用分层聚类(Hierarchical Clustering,HC)方法得到数据集的分层聚类树,然后依据聚类树结构和已标注样本在其中的分布信息定义每个未标注样本的代表性度量。将新方法与随机样本选择以及BvSB主动学习方法进行了比较,对1个光学图像集和1个全极化SAR数据集分类问题的实验结果显示,新方法性能稳定,优于其他两种方法。
[1] Settles B.Active Learning Literature Survey,Computer Science Technical Report 1648[R].University of Wisconsin-Madison,USA,2008:3-4 [2] 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 [3] 李志欣,施智平,李志清,等.融合语义主题的图像自动标注[J].软件学报,2011,22(4):801-812 [4] Carneiro G,Chan A B,Moreno P J,et al.Supervised learning of semantic classes for image annotation and retrieval[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2007,29(3):394-410 [5] Lienou M,Maitre H,Datcu M.Semantic Annotation of Satellite Images Using Latent Dirichlet Allocation[J].IEEE Geoscience and remote sensing letters,2010,7(1):28-32 [6] Tong S,Chang E.Support vector machine active learning for ima-ge retrieval[C]∥Proceedings of ACM Multimedia.Ottawa,Canada,2001 [7] Joshi A J,Porikli F,Papanikolopoulos N.Multi-class activelearning for image classification[C]∥IEEE Conference on Computer Vision and Pattern Recognition (CVPR).2009:2372-2379 [8] Dasgupta S,Hsu D.Hierarchical Sampling for Active Learning[C]∥Proceedings of the 25th international conference on Machine Learning (ICML).2008,307:208-215 [9] Webb A R.Statistical Pattern Recognition (Second Edition)[M].John Wiley & Sons Ltd.,2002 [10] Asuncion A,Newman D J.UCI machine learning repository.http://achive.ics.uci.edu/ml/datasets.html [11] Chang C C,Lin C J.LIBSVM:a library for support vector machine .http://www.csie.ntu.edu.tw/~cjlin/libsvm/ |
No related articles found! |
|