文献类型：会议论文

英文题名：Feature selection method of support vector machine for polarimeric sar landcover classification

作者：Feng, Qi[1] Chen, Er-Xue[1] Li, Zengyuan[1] Guo, Ying[1] Li, Wenmei[1]

第一作者：Feng, Qi

机构：[1] Research Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, Beijing 100091, China

会议论文集：Proceedings of Dragon 2 Final Results and Dragon 3 Kick-Off Symposium

会议日期：June 25, 2012 - June 29, 2012

会议地点：Beijing, China

语种：英文

外文关键词：Feature extraction - Knowledge based systems - Support vector machines

年份：2013

摘要：In order to improve multi-temporal polarimetric SAR data land cover classification accuracy, we investigated one optimal feature selection method. It uses the exhaustive algorithm to get the most comprehensive feature combination, and its feature selection criterion is based on some independent testing samples. Land cover classification experiment using mutil-temporal PolSAR dataset was carried out to validate the performance of the feature selection method, and its classification accuracy was compared with the knowledge based feature selection method and the classification without any feature selection. The results indicate that, the higher accuracy is obtained by the optimal feature selection method than by the other two methods. The accuracies of these three methods are 84.96%, 75.74% and 65.68% respectively.