基于DLSVM算法的高分辨率遥感图像分类研究  

Research on the Classification of the High Resolution Remote Sensing Images Based on DLSVM

作　　者：舒振宇[1,2] 王典洪[2] 周城[1] 海涛洋 

机构地区：[1]中南民族大学电信学院,武汉430074 [2]中国地质大学地球物理与空间信息学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2015年第4期78-84,共7页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：湖北省自然科学基金资助项目(BZY14019)

摘　　要：为了进一步提高高分辨率遥感图像的分类精度及效率,融合支持向量机SVM及局部支持向量机KNNSVM算法,借助主动学习相关理论,提出了基于距离的局部支持向量机算法(DLSVM).该算法通过对未标记样本和超平面之间的距离与预先设定的距离阈值相比较,判断是否需要进一步建立局部支持向量机KNNSVM来确定样本的类标.对实际的高分辨率遥感图像分类的实验结果显示:在合适的距离阈值与K值的设置下,该算法能够提高支持向量机SVM的分类精度,同时大大降低KNNSVM算法的时间消耗.In order to further improve the classification accuracy and efficiency of the high resolution remote sensing images,a local support vector machine algorithm based on distance（ DLSVM） with combining the algorithms of the Support Vector Machine（ SVM） and the Local Support Vector Machine（ KNNSVM） under the assistance of the Active Learning theory is proposed. Through comparing the distance from the unlabeled samples and the hyper plane with the threshold,the label of the unlabeled samples are determined by building a KNNSVM classifier or not. Experimental results on the real high resolution remote sensing images classification show that the proposed algorithm can improve the accuracy of the SVM and decrease time consuming of the algorithm KNNSVM greatly with the right threshold and K.

关 键 词：高分辨率遥感图像分类 支持向量机 局部支持向量机 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]