基于Otsu与海域统计特性的SAR图像海陆分割算法

Sea-Land Segmentation Algorithm of SAR Image Based on Otsu Method and Statistical Characteristic of Sea Area

合成孔径雷达图像海陆分割在海面目标检测、海岸线提取等海洋应用方面具有重要的意义。针对合成孔径雷达图像的特点,本文了提出粗阈值与精确阈值相结合的海陆分割算法,并提供了一套完整的海陆分割方案。首先利用改进的Otsu法计算粗阈值,获得大致海域范围;然后根据海域统计特性近似高斯分布的特点,在分析了高斯分布统计特性的基础上,结合粗阈值分割结果,提出了计算出精确的分割阈值的方法,进而获得准确的海陆分割图像;最后通过去除杂散点进一步完善分割结果。实验结果表明,本文提出的方法具有较高的处理精度和较快的处理速度。

The sea-land segmentation of SAR image is important in the marine applications,such as target detection,coastline extraction,etc.According to the characteristics of SAR image,this paper puts forward a threshold segmentation algorithm combining rough threshold with accurate threshold,and provides a complete set of sea-land segmentation scheme.Firstly,Otsu method is used to calculate a rough threshold,and then the rough sea area can be obtained.Secondly,according to the feature that statistical characteristic of sea area similar to Gauss distribution,an accurate threshold can be calculated on the basis of rough threshold segmentation,then the accurate sea-land segmentation image can be obtained.Finally,the segmentation result can be improved by the operation of removing speckle.The experiment results show that the proposed algorithm has high processing precision and quick processing speed.