【摘要】 针对高分辨率合成孔径雷达(Synthetic Aperture Radar,SAR)图像中道路目标难以有效提取的问题,提出一种融合马尔可夫随机场(Markov Random Field,MRF)分割与数学形态学处理的高分辨率SAR图像道路提取算法。该算法首先利用直方图均衡化和增强Lee滤波对SAR图像进行预处理,实现道路的边缘增强,抑制相干斑噪声;进而利用基于条件迭代模式(Iterated Conditional Mode,ICM)的MRF对SAR图像中的道路目标进行初分割;再用数学形态学填充空洞,平滑道路边缘;最后,基于道路的几何特征,使用偏心率、矩阵度、复杂度等因子去除虚警,从而提取出道路目标。利用该文算法对两块实际高分辨率SAR图像进行道路目标提取,均可以取得90%以上的正确道路提取率,表明本文算法具有较高的道路提取精度。更多还原

【Abstract】 To solve the problem of extracting the road from high resolution synthetic aperture radar(SAR)images,an algorithm that combines the Markov random field(MRF)segmentation and the mathematical morphology processing was proposed to extract the high precision road targets based on the high resolution SAR images. Firstly,the histogram equalization and the enhanced Lee filter were used to enhance the edge of the road and suppress the speckle noise.Secondly,the primary road segment was realized by MRF based on the iterated conditional mode(ICM).Thirdly,the mathematical morphology was adopted to reoccupy filling empty and smooth road edge.Finally,the geometrical characteristics of the road were adopted to remove the false alarm and get the results of the road extraction,such as the degree of eccentricity,matrix and complexity.The algorithm was used to process two high resolution SAR images to extract the roads.And the results show that more than 90%of the road extraction results is correct in the two experiments.更多还原