Acta Scientiarum Naturalium Universitatis Pekinensis

Previous Articles     Next Articles

Land Cover Classification of Hyperspectral Data Using Composite Kernel Support Vector Machines

SHANG Kun1, LI Peijun2, CHENG Tao3   

  1. 1. Institute of Remote Sensing Applications, Chinese Academy of Sciences, Beijing 100049; 2. Institute of Remote Sensing and Geographic Information System, Peking University, Beijing 100871; 3. Department of Earth and Atmosphere, University of Alberta Edmonton, Alberta AB T6G2E3;
  • Received:2010-01-19 Online:2011-01-20 Published:2011-01-20

基于合成核支持向量机的高光谱土地覆盖分类

尚坤1,李培军2,程涛3   

  1. 1. 中国科学院遥感应用研究所, 北京 100101; 2. 北京大学遥感与地理信息系统研究所, 北京 100871;3. Department of Earth and Atmosphere, University of Alberta Edmonton, Alberta AB T6G2E3;

Abstract: Land cover classification using recently developed composite kernel support vector machines (SVM) and hyperspectral data is proposed. The hyperspectral data are first subdivided into different subsets. SVM method is then used to select optimal parameters for classification of each subset. Finally, different subsets are combined by a composite kernel function, and the best one selected from different parameter combinations is used in final land cover classification using composite kernel SVM. The HYDICE data of Washington DC is used to evaluate and validate the proposed method. The results show that land cover classification of hyperspectral data using composite kernel SVM can obtain higher classification accuracy than the traditional SVM method.

Key words: hyperspectral data, composite kernel, support vector machine, image classification

摘要: 提出一种基于合成核支持向量机的高光谱数据分类方法。该方法首先对高光谱数据进行分组, 对得到的不同数据组分别运用支持向量机方法进行分类参数的优化, 然后组合不同的核函数来综合不同的数据组, 得到最终的分类结果。利用华盛顿地区 HYDICE 高光谱数据对所提出的方法进行评价和验证, 结果表明, 基于合成核支持向量机的高光谱图像分类, 可获得比传统支持向量机更高的分类精度。

关键词: 高光谱, 合成核, 支持向量机, 图像分类

CLC Number: