摘要
根据合成孔径雷达(SAR)成像机理和图像特征,针对SAR图像中的起伏地形,提出了一种基于有向多尺度模型分割算法。该方法利用了SAR图像不同方向不同尺度间的统计相依性,而且考虑了SAR图像的空间信息。由于是基于有向多尺度模型少量的特征数据,利用最大期望(EM)算法可以快速估计参数,同时利用数学形态学算法进行分割修正,实现了高精度的SAR图像起伏地形的快速分割。实测SAR图像分割实验结果证明,对比其他分割算法,该方法对起伏地形的分割效果更为精确。
A segmentation method based on directional multi-scale for the rugged topographical in SAR image is presented in this paper according to the principle of SAR （ Synthetic Aperture Radar） imaging and the characteristics of image. The improved directed multi-scale algorithm is used in this method to describe the statistical dependence of different between direction and scale. Starting from characteristics of SAR image, the advantages of multi-scale model and expectation-maximization algorithm has been combined in this method, the spatial information of SAR image is also considered. The parameters are estimated fast and effectively by expectation-maximization algorithm , and a SAR image segmentation processing has been realized by mathematical morphology modifying the segmentation result. The better experimental segmentation result of the rugged topographical in real SAR image has been obtained by using this method, which shows that this method is effective and available.
出处
《现代雷达》
CSCD
北大核心
2011年第6期65-67,86,共4页
Modern Radar
基金
江苏省"六大人才高峰高层次人才资助项目"(08-E-008)
关键词
合成孔径雷达
图像分割
最大期望算法
有向多尺度模型
数学形态学
Synthetic Aperture Radar （SAR）
image segmentation
expectation maximization algorithm
directed multi-scale method
mathematical morphology
