光谱与形状特征相结合的道路提取方法研究

Research on Extracting Road Based on Its Spectral Feature and Shape Feature

针对从遥感影像上提取道路,目前已经有很多研究,但尚有许多问题有待解决。如果仅从光谱特征分类入手提取道路,会存在“异物同谱”问题。为解决这个问题,提出了用光谱特征与形状特征相结合的方法提取道路。提出了3个形状指数描述道路的形状特征。以SPOT卫星影像为例,将高空间分辨率的全色图像与多光谱图像融合,首先基于道路光谱特征进行图像分类,然后再应用道路的形状特征去除分类图像中的“异物”目标。研究结果表明该方法能有效地从遥感影像中提取主干道路。

There are still many problems needed to resolve about extracting road from remote sensing image, although a lot of efforts have been put in the area. The phenomenon of different objects with same spectrum is one of the problems, which often decreases the accuracy of classification based on spectral feature when we extract roads. In order to avoid this problem, the paper proposes the method that spectral feature and shape feature are all used to identify the road. The shape feature of the road is described by three shape parameters in this paper. The experimentation uses the image coming from the fusion of multispectral and panchromatic data. After the image is classified based on spectral feature, the objects that have same spectrum with road are removed in the classified image because their shape parameters are not like road＇s shape parameters. According to the result of experimentation, the proposed approach can extract the main roads in the image effectively.