Abstract: As an important ground object information, road plays an irreplaceable role in urban planning and so on. Because Synthetic Aperture Radar（SAR） has characteristic of all weather, a large number of road edge algorithms which based on SAR image emerge spontaneously. A multi-window road edge detection algorithm is proposed to solve the problems such as high false detection rate and poor integrity of road edges causing by speckle noise. Based on the weighted local entropy, this paper evaluates the probability of SAR images of pixel points which falling on the road and determines the weight of each pixel in multi-window fusion. Finally, experiments on SAR chip images in variant regions are carried out, and the results show that the integrity of the road edges and the suppression effect of noise are improved after weighted fusion.

Key words: SAR image, edge detection, road detection, weighted local entropy, multi-windows fusion method

摘要： 道路作为一种重要地物信息，在城市规划等领域中起着不可替代作用。合成孔径雷达（SAR）具有全天候等成像特点，因此基于SAR图像已有许多道路边缘检测算法。提出一种多窗口道路边缘检测算法，来解决相干斑噪声引起的道路边缘误检率高完整性差等问题。该算法首先以加权局部熵的大小为基础，评估SAR图像中像素点落在道路上的概率，然后以该概率为依据，确定每个像素点多窗口融合的权值。最后，对不同大小窗口的边缘检测结果进行加权融合。通过对不同区域的SAR图像切片进行实验，结果表明加权融合后得到的道路边缘的完整性及对噪声的抑制效果均有所提高。

关键词: SAR图像, 边缘检测, 道路提取, 加权局部熵, 多窗口融合