| 基于多尺度纹理和光谱信息的SVM分类研究 |
| |
| 引用本文： | 陈晨,张友静.基于多尺度纹理和光谱信息的SVM分类研究[J].测绘科学,2009,34(1). |
| |
| 作者姓名： | 陈晨 张友静 |
| |
| 作者单位： | 1. 河海大学土木工程学院,南京,210098
2. 水文水资源及水利工程国家重点实验室,南京,210098 |
| |
| 摘 要： | 基于单尺度纹理和光谱信息的地物分类较难取得理想效果,本文结合多尺度纹理与光谱信息,运用SVM分类方法,对IKONOS遥感影像进行分类。结果表明:结合多尺度纹理和光谱信息的SVM高分辨率遥感影像分类,能够更好地描述地物,分类总体精度达到83.9%,与基于光谱信息的最大似然法和基于单尺度纹理和光谱信息的SVM分类方法比较,分类精度分别提高了13.8%和4.9%,该方法有助于提高高分辨率影像的分类正确率。
|
| 关 键 词： | 高分辨率 多尺度 纹理特征 支持向量机 |
The classification of SVM based on the multi-scale texture information and spectral information |
| |
| CHEN Chen,ZHANG You-jing.The classification of SVM based on the multi-scale texture information and spectral information[J].Science of Surveying and Mapping,2009,34(1). |
| |
| Authors: | CHEN Chen ZHANG You-jing |
| |
| Abstract: | It is difficult for classification based on spectral data to obtain satisfactory results.The paper adopted the classification of SVM combined with multi-scale texture data and spectral data to classify the image of IKONOS.It shows that SVM classification combined with multi-scale texture data and spectral data can character the textures precisely.The accuracy of which is 83.9%.And compared with the classification of max-like and the classification of SVM based on single-scale texture data and spectral data,the accuracy has improved 13.8% and 4.9% separately.It is proven that the method can improve the accuracy of high resolution image classification. |
| |
| Keywords: | high spatial resolution multi-scale texture feature SVM |
| 本文献已被 万方数据 等数据库收录！ |
|
