基于语义规则的高分辨率影像道路交叉口提取方法

为了准确提取遥感影像上道路交叉口目标,提出一种利用特征语义规则从高分辨率影像上提取道路交叉口的方法。该算法构建交叉口模型时将其视为由同质区域像素集合及区域轮廓边界构成的面对象,提取过程分为两步:1)利用辐射、纹理特征语义匹配提取交叉口候选区域;2)通过几何特征语义匹配筛选候选区域、识别交叉口属性。利用多源遥感影像对算法正确性及合理性进行验证,结果表明:算法能准确、完整地提取道路交叉口,可为影像道路网构建提供辅助信息。

Road Intersections Extracting from Remote Sensing Imagery Based on Semantic Rules

In order to extract the intersection targets from remote sensing images accurately,a new method for road intersection extraction using characterized semantic rules is proposed in this paper.In this model,the intersection is seen as area object which has two components composed of a set of pixels in homogeneous region and of the region's contour boundaries.The method has two steps:firstly,candidate intersection regions are extracted by characterized semantic matching of radiation and texture features;secondly,the optimal candidate intersections are generated and the properties are identified using characterized semantic matching of geometric features.The feasibility of the proposed method is tested using a set of high-resolution remote sensing images.A comparison with several existing methods shows that the proposed method can extract the road intersections accurately and completely,which can provide useful auxiliary information for road network construction.