摘要
高分辨率遥感图像纹理信息丰富,将其与光谱信息相结合进行地物分类可提高地物的精度。将改进的局部二值模式(LBP)纹理应用到高分辨率图像的土地覆盖分类中,并与只利用光谱信息和加入传统LBP纹理信息的分类结果相比较。结果表明:改进的LBP具有很好的抗噪性能,并能更有效地表达图像的纹理信息,加入这种纹理信息的图像分类精度明显高于纯光谱分类和加入传统LBP纹理信息的分类。
High - resolution remote sensing images have rich texture information, and combined texture information and image spectral information can improve the recognition accuracy of surface feature. In this paper, a new extended Local Binary Patterns （LBP） texture was applied to the high - resolution images classification in comparison with classifications using spectral data only and using combined spectral data and LBP texture features. The results show that the extended LBP has a good anti - noise performance, and the classification of image including the extended LBP texture can achieve a higher accuracy than the classifications using spectral data alone and using combined spectral data and LBP texture features.
出处
《国土资源遥感》
CSCD
2010年第4期40-45,共6页
Remote Sensing for Land & Resources
关键词
纹理
LBP
分类
Texture
LBP
Classification
作者简介
宋本钦（1984-），男，硕士研究生，主要从事高分辨率遥感信息处理方法研究。
