Abstract: In order to overcome the defects that extracts rivers sample with spectrum information, using high-resolution remote sensing satellite imagery of the outstanding features of high resolution, a method of comprehensive satellite image feature is put forward, including spectrum, texture, geometric features and so on. It is a method that extracts river with joint features. The rivers spectral features, texture features and geometric shape features are described respectively. By selecting characteristic parameters, structure integrated characteristic matrix, this paper reuses the average clustering segmentation, and eventually gets river target. The real high-resolution remote sensing images Worldview1 experiments validate the preciseness and rapidity of the methods.

Key words: rivers extraction, feature extraction, K-mean, Worldview1 image, multi-feature fusion

摘要： 为了克服单纯采用光谱信息提取河流的缺陷，利用高分辨率遥感影像突出的高分辨率的特性提出一种综合影像中光谱、纹理、几何特性等多特征联合提取河流的方法。该方法分别对河流水体的光谱特征、纹理特征及河流几何形状进行描述，选取特征参数，构造综合特征矩阵，利用均值聚类分割最终得到河流目标。通过对真实高分辨率遥感影像Worldview1影像进行的实验验证了该方法的高精准性及快速性。

关键词: 河流提取, 特征提取, K-mean, Worldview1影像, 多特征融合