在SAR图像分割中,尤其是车辆目标的SAR 图像分割中,一般需要得到目标和阴影两个区域的分割结果。文中为 了解决车辆目标的SAR图像多区域分割,提出了一种分层多区域CV模型,该模型结合了一种新的惩罚项,并且同时使用 水平集函数的阶跃初始化,使模型具有了良好的水平集演化的属性。同时,模型对噪声的敏感性下降,使模型适用于未预 处理的SAR图像。最后,对比Chan-Vese多区域分割模型,将分层多区域CV模型应用于未预处理的SAR图像,实验结果 验证了模型的有效性。

In the SAR image segmentation, especially vehicle targets in SAR images, which generally need to get segmentation results about two regions that objective and shadow. In this paper, in order to solve multiregion segmentation of vehicle targets in SAR images, a demixing multiregion CV model was presented, which combines a new penalty term. At the same time, level set function uses the step initialization. As a consequence, this model has some good properties of level set evolution and lower noise sensitivity, and can be applied to SAR image that without preprocessing. Finally, contrast to the multiregion model of chan-vese, demixing multiregion CV model is applied to SAR image that without preprocessing, and experimental results verify the validity of the model.