中文摘要：

针对遥感影像中植被信息的波谱特征，提出了整体一局部植被信息多尺度迭代转换提取模型。首先在基于植被指数的基础上对影像进行分割，并通过样本的自动选择，对影像进行大尺度分类；然后对分类结果进行缓冲区分析，建立局部区域对象，再进行小尺度的局部分割与分类；最后通过迭代，重复整体一局部的过程，使得植被与非植被信息的边界得到最优化分离，从而提高了植被信息提取的精度。选取江汉平原地区的LANDSAT ETM＋影像进行实验，并与常规方法得到的结果进行了对比，实验证明，多尺度迭代提取方法可以有效地提高植被信息提取的精度。

英文摘要：

According to the spectral characteristics of vegetation, this paper proposed a new multi-scale iterative information extraction model for vegetation classification from remote sensing data. For the multi-scale iterative vegetation information extraction model, the first step is NDVI calculating, and then, segmented the whole remote sensing image based on the value of NDVI by histogram threshold segmentation method on the whole level. Combined with the spectral value of the image, this paper chose the sample point automatically by histogram and used maximum likelihood classification method to extract the vegetation information. Implemented the local level processing would build new local units by using buffer zone, and would do segmentation and classification on the local level. The last step was to do the iterative process. To repeat the segmentation, buffering, classification step by step, would sketch the outline of the vegetation information more precisely. When the result of the iteration was the same as the former one, it could get the final result, and then stopped the iteration and combined all the feature units to get the result of vegetation classification. In the end, this paper used the LANDSAT ETM ＋ data to do the experiment. The results show that the method can extract vegetation information more precisely than the traditional methods.