从不同森林类型的纹理差异入手，首先利用离散小波变换提取出图像的纹理特征，然后利用面向对象分类方法将纹理信息与原有的光谱信息结合进行分类。对小波变换提取纹理信息的分解层数、滑动窗口及纹理测度等问题进行系统的分析，并找出了有效反映植被纹理差异性的6个纹理特征因子。该方法可用于解决林业遥感中的诸如林种、树种的分类等问题。最后得到的总体分类精度达到92.7 %，与传统的基于像素的分类方法相比效果有所提高。关键词：森林类型分类；高分辨率影像；纹理；小波变换

Considering to the texture differences among foresty types, texture features of forestry were firstly extracted from remote sensing imagery with discrete wavelet transformation, and the imagery was then classified combining objectoriented classification method with the original spectral values and texture features. Six texture measure factors which can reflect the difference efficiently among plant difference are sieved out based on analyzing of wavelet decomposition level, sliding window and texture measure. This method could solve the problems of forestry type classification, tree species classification, etc. The overall classification accuracy reached 92.7 %, and it is much better than traditional method of classification based on pixel values.