基于HS样本选择和BvSB反馈的多类图像分类  

Multi-class image classification based on HS sample selection and BvSB feedback

作　　者：金良[1] 曹永锋[1] 苏彩霞[1] 任俊英[1] 

机构地区：[1]贵州师范大学数学与计算机科学学院,贵州贵阳550001

出　　处：《贵州师范大学学报（自然科学版）》2014年第4期56-61,82,共7页Journal of Guizhou Normal University：Natural Sciences

基　　金：国家自然科学基金资助项目(编号:41161065;40901207);贵州师范大学学生科研基金;贵州省科学技术厅;贵州师范大学联合科技基金资助项目(黔科合J字LKS[2013]28号)

摘　　要：针对主动学习由于初始阶段随机选择样本而导致的抽样偏差,将Sanjoy Dasgupta等人提出的分层聚类采样(Hierarchical sampling,HS)引入到主动学习方法中,替代初始阶段随机样本选择,然后在基于支持向量机分类器的图像算法中引入最优标号和次优标号(Best vs second-best,BvSB)的反馈准则,提出了基于HS和BvSB(HS+BvSB)的多类图像分类方法。分别在两组标准测试数据集上进行分类实验,比较HS+BvSB方法与随机选择样本+BvSB方法的学习性能,结果表明,随着初始选择样本数目的增多,提出的Hs+BvSB方法具有更优的性能。For the bias sampling due to primary randomly selection of sample in active learning,stratified cluster sampling （HS） put forward by Sanjoy Dasgupta is introduced into the active learning method in this article,then the feedback criterion of optimal label and subprime label （Best vs second-Best BvSB） is introduced in the image algorithm based on support vector machine （SVM） classifier.The proposed method is tested on two groups of data set.The comparison with the learning performance of active learning whose rimary samples are randomly selected is given.The results show that the classification performance of this new method is better with the increase number of sample using HS method.

关 键 词：主动学习 支持向量机 图像分类 抽样偏差 最优标号 次优标号 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]