结合Gabor纹理与几何特征的高分辨率遥感影像城区道路提取方法

Urban Road Extraction of High Resolution Remotely Sensed Imagery with Gabor Texture and Geometrical Features

在充分考虑道路频谱在不同频带、不同方向上能量分布的基础上,提出一种基于Gabor纹理与几何特征相结合的高分辨率遥感影像城区道路提取方法。首先通过Gabor滤波器组得到遥感影像不同频带、不同方向上的Gabor纹理特征,并利用K-means方法对遥感影像进行分割;然后利用形态学方法分割与道路相连的地物,并选取适当的几何特征剔除非道路地物;最后利用形态学方法对道路网进行修整。实验结果表明,该方法可以有效、便捷地从高分辨率遥感影像中提取城区主干道路网。

Through the feature analysis of the frequency curve, the paper proposed a new method which integrated Gabor texture and geometrical features in the urban road extraction from high resolution remotely sensed imagery. In the experiment, the texture features in different frequencies and different directions were obtained by a given bank of Gabor filters, and then K-means clustering method was applied for imagery segmentation. The morphologi- cal methods were utilized to separate the road objects from non-road objects, and the geometrical features were ap- plied to take out the non-road objects. Then the extracted roads were regulated by mathematical morphology. The result showed that the proposed method could effectively extract the urban road information from high resolution re- motely sensed imagery.