基于混合模型的SAR影像海陆分割算法

Sea-land segmentation algorithm for SAR images based on mixture models

合成孔径雷达(SAR)影像的海陆分割是诸如海洋目标检测和识别等基于海洋区域SAR影像解译的基础和关键环节之一。为解决复杂背景下遥感影像海陆分割问题,提出一种基于混合概率模型的海陆分割算法。首先利用Harris角点检测算法检测出影像中包含角点的图像子块,进而通过均值漂移(MS,mean-shift)算法对图像子块进行聚类分析得到陆地区域的像素样本;然后利用陆地的像素样本,通过最大期望(EM,expectation maximization)迭代算法拟合出混合模型概率密度分布的相关参数;最后通过混合概率模型检测出陆地前景区域,得到海陆分割结果。实验证明,本文方法能够对包含海陆的异质遥感影像实现有效的海陆分割。

Sea-land segmentation of synthetic aperture radar （SAR） image is one of the key stages for SAR image applications such as sea target detection and recognition,which are operated only in sea re- gions. A mixture probability models based algorithm is explored to solve the sea-land segmentation prob- lems in complex SAR image. First, Harris comer detector is employed to detect image patches containing corner points. Furthermore,the image patches are analyzed by mean-shift clustering algorithm, and the pixel samples of land regions are obtained. Second,according to pixel samples from land regions,the pa- rameters of probability density function in mixture models are fitted by expectation maximization （EM） algorithm. Finally,land foreground regions are segmented by mixture probability models. Experimental results demonstrate that the proposed algorithm has excellent performance to deal with heterogeneous SAR image.