| 顾及分类器参数的全极化SAR图像特征选择与分类 |
| |
| 引用本文： | 袁春琦,徐佳,程圆娥,许康.顾及分类器参数的全极化SAR图像特征选择与分类[J].测绘科学技术学报,2016(5):507-512. |
| |
| 作者姓名： | 袁春琦 徐佳 程圆娥 许康 |
| |
| 作者单位： | 1. 河海大学 地球科学与工程学院,江苏 南京,211100;2. 河海大学 地球科学与工程学院,江苏 南京 211100;江苏省测绘工程院,江苏 南京 210013;3. 江苏省测绘工程院,江苏 南京,210013 |
| |
| 基金项目： | 国家自然科学基金项目(41301449)，江苏省测绘地理信息科研项目(JSCHKY201501)，地理空间信息工程国家测绘地理信息局重点实验室基金项目(201324) |
| |
| 摘 要： | 全极化SAR获取的信息量远多于传统SAR,但信息量的增加并不能确保分类精度的提高,如何有效进行特征选择至关重要。针对自适应特征选择问题,提出一种顾及分类器参数的特征选择和分类方法。该方法以支持向量数为评估依据,结合遗传算法进行特征选择,并同时对分类器参数进行寻优;最后利用优选的特征集和模型参数进行分类。为验证算法的有效性,利用两组全极化数据进行了监督分类实验。实验结果表明,提出方法降低了SVM分类器对自身参数的敏感性,而且能在较少特征个数下具备良好的泛化性能,分类精度优于未经过特征选择和参数优化的方法。
|
| 关 键 词： | 极化SAR 特征选择 支持向量机 分类 参数优化 |
Feature Selection and Classification of Fully Polarimetric SAR Images Considering Classifier Parameters |
| |
| YUAN Chunqi,XU Jia,CHENG Yuane,XU Kang.Feature Selection and Classification of Fully Polarimetric SAR Images Considering Classifier Parameters[J].Journal of Zhengzhou Institute of Surveying and Mapping,2016(5):507-512. |
| |
| Authors: | YUAN Chunqi XU Jia CHENG Yuane XU Kang |
| |
| Abstract: | The polarimetric synthetic aperture radar( PolSAR) images can provide more information than conven-tional SAR images. However, increasing the number of features does not always improve classification accuracy. It is very important to select an optimized feature set from fully polarimetric SAR images. In order to effectively and a-daptively select features, an improved feature selection and classification method considering classifier parameters is proposed in this paper. On the basis of the assessment of support vector number, the effective features are select-ed while the classifier parameters are optimized by using the genetic algorithm( GA) . Then, the optimized feature set and parameters are applied to classify the SAR image. To evaluate the performance and efficiency of the pro-posed method, the experiments have been carried out on two sets of PolSAR data. The classification results demon-strate that the proposed method is more accurate than conventional methods, which can not only reduce the effect caused by classifier parameters, but also obtain high classification accuracy with less number of features. |
| |
| Keywords: | polarimetric synthetic aperture radar( PolSAR) feature selection support vector machine( SVM) classification parameters optimization |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
