| 遥感图像土地覆盖分类中多源特征数据选择研究 |
| |
| 引用本文： | 张元,陈亮,王文种,王军战.遥感图像土地覆盖分类中多源特征数据选择研究[J].测绘科学,2009,34(2). |
| |
| 作者姓名： | 张元 陈亮 王文种 王军战 |
| |
| 作者单位： | 1. 河海大学水文水资源及水利工程国家重点实验室,南京,210098
2. 河海大学土木工程学院,南京,210098 |
| |
| 摘 要： | 多源特征数据可以提高遥感图像的分类精度,选择合适的特征数据十分重要。利用基尼指数对多尺度纹理信息、主成分变换前三分量、地形数据等特征进行选择,选出最佳特征子集。利用支持向量机、神经网络分类法、最大似然法分别对全部特征数据和最佳特征子集结合多光谱数据进行分类。实验结果表明:基尼指数可以有效地对多源特征数据进行选择,特征选择可以提高分类器效率,提高分类精度。
|
| 关 键 词： | 基尼指数 多源特征数据 特征选择 分类 |
Multi-source feature data selection for land cover classification using remote sensing image |
| |
| ZHANG Yuan,CHEN Liang,WANG Wen-zhong,WANG Jun-zhan.Multi-source feature data selection for land cover classification using remote sensing image[J].Science of Surveying and Mapping,2009,34(2). |
| |
| Authors: | ZHANG Yuan CHEN Liang WANG Wen-zhong WANG Jun-zhan |
| |
| Abstract: | Multi-source feature data can be used to improve the accuracy of land cover classification. However, selecting suitable features is an important step. Gini index was applied to select features from the feature set including multi-scale texture, the components of principal component analysis, and terrain data. The selected features and multi-band data were classified into six classes by support vector machine, neural network classifier and maximum likelihood classifier. The results showed that Gini index could select features successfully, and the classification accuracies were improved while the run time was reduced after feature selection. |
| |
| Keywords: | Gini index multi-source feature data feature selection land cover classification |
| 本文献已被 万方数据 等数据库收录！ |
|
