基于多特征的GF-2影像城市植被分类方法

Urban Vegetation Classification Based on Multi Feature GF-2 Image

城市植被是改善城市生态环境,提高城市生态质量的重要保障。针对城市植被研究中国产卫星使用较少及传统分类技术无法应用高分辨率影像的纹理特征问题。本文通过提取光谱特征、纹理特征,结合随机森林算法的方法对单时相高分二号影像进行分类。相比于只用光谱特征,分类的结果比单一光谱特征分类的效果较好,总体分类精度为85.11%,Kappa系数为0.80,比单一光谱特征分类总体精度提高了4.19%,Kappa系数提高了0.6。基于多特征的分类方法提高了分类的精度,为获取城市植被的空间分布特征提供了一种新的方法。

Urban vegetation is an important guarantee to improve urban ecological environment and urban ecological quality.In order to solve the problem that Chinese satellites are rarely used in urban vegetation research and the traditional classification technology can not apply the texture features of high-resolution images.In this paper,by extracting spectral features,texture features,combined with random forest algorithm,we classify single temporal high-resolution image No.2.Compared with using only spectral features,the result of classification is better than that of single spectral feature classification,the overall classification accuracy is 85.11%,and the Kappa coefficient is 0.80,which is 4.19%higher than that of single spectral feature classification,and the kappa coefficient is 0.6 higher than that of single spectral feature classification.The classification method based on multi features improves the accuracy of classification and provides a new method for obtaining the spatial distribution characteristics of urban vegetation.