主动学习在基于内容的遥感影像检索中的应用  

APPLICATION OF ACTIVE LEARNING IN CONTENT-BASED REMOTE SENSING IMAGE RETRIEVAL

作　　者：李政[1] 

机构地区：[1]山东艺术学院,山东济南250014

出　　处：《计算机应用与软件》2010年第10期110-113,共4页Computer Applications and Software

基　　金：山东省高等学校科技计划项目(J09LG51)

摘　　要：提出一种集成主动学习和支持向量机的学习算法并应用于基于内容的图像检索。首先在相关反馈过程中结合样本不确定性与减小样本间冗余度的采样策略选择样本进行类别标记组成样本集训练支持向量机分类器,然后利用得到的分类器进行图像检索,直至用户满意为止。基于遥感影像的实验表明,算法能够提高基于内容的图像检索的效率和性能。The thesis proposes a learning algorithm that combines both active learning and support vector machine to apply to content-based image retrieval.At first,during the relevance feedback process the sampling strategy which integrates sample prediction uncertainty with sample redundancy reduction is used in the decision of sample classification labeling to consist sample sets to train the support vector machine classifier.Then content-based image retrieval is carried out by the trained classifier to strive for user satisfaction.Experiments with remote sensing images demonstrate that the proposed algorithm can improve the efficiency and performance of the content-based image retrieval.

关 键 词：基于内容的图像检索 支持向量机 主动学习 变型空间 相关反馈 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]