基于图像分块和线段投票的遥感道路边缘线提取

Extraction of road edge lines from remote sensing image based on image blocking and line segment voting

从遥感图像中提取边缘线是一个经典的课题,不同的边缘提取算法适用于不同类型的图像。实际图像中道路的几何形状不甚规则,因受到建筑和树木遮挡导致对比度较低且噪声影响较严重,道路边缘线会发生断裂,故高分辨率遥感图像道路边缘线提取一直是一个研究热点。针对现有方法很难提取出清晰连续的道路边缘线问题,提出一种遥感图像道路边缘线提取新方法:首先通过方向模板检测边缘点,搜索出分块图像中的子线段;然后延伸子线段并进行投票,连接处于弯曲边缘线的直线段,将长度大于特定阈值的边缘线作为输出结果;最后去除毛刺和分叉,取8个方向道路的并集作为最终道路网。实验结果表明,该方法能够从高分辨率遥感图像中较好地提取带有一定曲率、对比度较低、噪声影响严重的道路边缘线。

Edge lines extraction from remote sensing images is a classic problem, and different edge extraction algorithms are applicable to different types of images. The road shape is not very regular, the contrast is low and the impact of noise is serious in actual remote sensing image because the road might be blocked by buildings and trees, and the road edge lines are likely to be broken; therefore road edge lines extraction from high-resolution remote sensing image is always a hot research topic. In this paper,the authors propose a new method for extraction of the road lines from remote sensing image so as to solve the problem that it is difficult for the methods available to extract clear and continuous road edge lines. Firstly, the direction templates are introduced to detect the edge points and search for the sub-segments in block image;then the sub-segments are extended and the line segment voting is taken to connect straight line segments in the curved edge lines, and the edge lines whose length is greater than a given threshold are output; finally, the spur and bifurcation are removed and the union of edge lines in eight directions is taken as the final road network. Experiment results show that the method proposed in this paper can be used to extract the road edge lines which have a certain curvature and low contrast and are affected by noise seriously from high-resolution remote sensing images.