中文摘要：

采用面向对象的影像分类方法，结合多尺度分割技术，以Quick Bird影像为实验数据，进行农用地的精细自动分类。首先，根据地物大小，选择最优分割尺度，构建多尺度分割等级网；然后，综合利用高分辨率影像的光谱信息、纹理和形状特征，建立各个对象的特征集；最后，通过目视解译建立隶属度函数，实现地物的分层提取。实验表明，该方法能有效区分农作物种类，相对于传统的像素级分类方法，该方法明显提高了高分辨率影像的分类精度，且避免了“椒盐”噪声的产生。

英文摘要：

This paper has explored the method and scheme for classifying different kinds of agricultural land from the high resolution remote sensing images by using multi - scale segmentation technology and rule - based image analysis approaches. Firstly, optimal segmentation scale was examined to construct a multi -scale segmentation level network according to the size of objects. Secondly, on the basis of spectrum, shape, texture and topology characteristics of images, several features of NDVI, shape indices, brightness, mean spectral value of red band, and ratio of near - infrared band, the standard deviation of near - infrared band and homogeneity were selected to classify objects into four agricultural land categories. The results show that these characteristics are effective in identifying agricultural land type and that the precision is higher than that of the traditional maximum likelihood classification.