Abstract: For the coherent speckles in SAR imagery，the processing of object extraction becomes more difficult.Using the traditional target extraction arithmetic for optical imagery，we can not get ideal result on SAR imagery.Water object in SAR imagery represents as some continuous region with low lightness and strong speckle noise.The traditional method to extract low lightness object can be implement by segment the imagery to binary imagery using a specify threshold.As the strong effect of the coherent speckle，how to gain an appropriate threshold is difficult.In this paper，a sequential nonlinear filtering-based method is proposed to extract the water object in SAR imagery automatically.The experimental results using real SAR imagery show that this method can extract water objects in SAR imagery rapidly and accurately.

Key words: synthetic aperture radar imagery, water target, automatic recognition, nonlinear filter, mathematical morphology

摘要： 合成孔径雷达（SAR）影像由于其特有的相干斑点噪声，加大了其目标边界特征提取的难度。采用一般针对光学影像的目标提取区域分割算子，在SAR影像的自动目标识别处理中往往不能取得理想的效果。水体目标在SAR影像中表现为整体亮度很低且成面状连续分布的像素区域。目前大多采用阈值分割转为二值影像的方法提取低亮度目标，但由于强烈的相干噪声影响，如何自动获得一个合适的阈值是一个比较困难的问题。一般多采用基于假设检验，多次迭代获得阈值。针对水体目标的亮度及形状分布特征，不作二值分割处理，而采用序列非线性滤波处理，提出一种针对SAR影像水体目标的快速提取方法。经实际数据检验，算法可以快速有效地实现SAR影像水体目标的自动提取识别。

关键词: 合成孔径雷达影像, 水体目标, 自动识别, 非线性滤波, 灰度形态学