基于高分二号的面向对象城市水体信息提取

Object-oriented Urban Water Information Extraction Based on GF-2

根据国产高分二号卫星影像具有光谱信息丰富、空间分辨率高的特点,利用面向对象的方法提取了城市水体信息。首先借助ESP尺度分割工具,确定最优分割尺度;再统计地物的光谱信息、形状因子、纹理特征等,构建城市水体提取知识库。针对城市水体和建筑物阴影混淆严重的问题,提出了一种利用能区分城市水体和阴影的ISTD指数提取城市水体的方法。结果表明,该方法的总体精度为89.55%,Kappa系数为0.86,比传统的基于像元的方法精度提高了16%。该方法可有效解决水体和建筑物阴影混分问题,在城市水塘等小面积水体提取方面具有明显优势。

Based on the plentiful spectral information and high spatial resolution images of GF-2, we used the object-oriented method to extract the urban water information. Firstly, we used the ESP scale segmentation tool to determine the optimal segmentation scale. And then, we established the knowledge base of urban water by using the spectral information, shape factor and texture feature of statistical features. Aiming at the serious confusion of urban water and buildings, we proposed an ISTD index to distinguish urban water and shadows. The results show that the total accuracy of urban water extraction is 89.55% and the Kappa coefficient is 0.86, which is 16% higher than that of traditional pixel-based extraction.The method can effectively solve the problem of water and building shadow blending, and has obvious advantages in the extraction of small area water in urban pond.