结合面向对象方法的多层次干旱与半干旱区域城镇提取方法

在干旱与半干旱区域戈壁及沙漠等高亮地表与城镇连成一片,两者的光谱特征在中等分辨率遥感数据上非常相似;因此,利用基于像素的分类方法很难将城镇准确提取出来。根据两种地物的样本对NDVI、NDBI的分布特征统计分析得出:基于面向对象的分类方法在提取城镇信息方面有较大优势。以典型的干旱区域—黑河流域张掖市及周边地区作为研究区域,将面向对象的方法应用到具有中等分辨率的Landsat-TM数据上,提出了结合面向对象方法的多层次干旱与半干旱区域城镇提取方法。该方法首先使用分层分类的方法得到城镇和荒漠的混合影像,然后使用面向对象的分类方法精确提取城镇信息,其中分割对象过程中引入样本可分离度量化不同尺度的影像分割效果,实现最优尺度分割。结果表明:其目视效果、总体精度(94.51%)和Kappa系数(0.89),均优于支持向量机(SVM)与基于时间序列的分类方法。

A Multi-layers Urban Extraction Method Combining Object-oriented Classification for the Arid and Semi\|arid Region

Urban is one of the important types of land use/cover mapping.The spectral character of barren lands and urban information in arid and semi\|arid region are similarity in the moderate resolution remote sensing images.It is very difficult to extract the urban information using the traditional classification methods based on pixels.For exactly extracting the urban information of the Middle reaches of the Heihe River basin located at the typical arid and semi\|arid region in northwestern China using Landsat TM image,a multi\|layers classification procedure combining object\|oriented classification method for precisely extracting urban information was is proposed in this paper.As one critical step of object\|oriented classification,image segmentation must ensure to get the best segmentation result.In order to choose the optimal segmentation scale,divergence is applied to measure the segmentation result.The urban information extracted using the new method rejects the influence of barren lands for extracting urban information,which is better than traditional supervised classification using SVM classifier and time series classification method.Comparing the urban extraction results using three different methods,the new method works better during more confusing regions and the boundary information is more precise than the other two results.The accuracy estimation result shows that the overall accuracy is about 94.51% and the kappa coefficient is about 0.89 using the new method.