基于全局Maxflow邻域生长算法的SAR图像目标分割

Target Segmentation for SAR Images Based on Global Maxflow Neighbor Region Grow Algorithm

提出一种改进的GrabCut算法,利用迭代图割(GrabCut)算法对彩色图像进行分割的特点,在不需要交互式操作的情况下,将SAR图像转化为24 bit BMP的位图进行处理,再结合邻域生长算法实现对SAR图像目标的自动分割。通过对MSTAR和NASA喷气推进实验室的SAR图像进行目标分割实验,结果表明,该算法分割精度高于目前通用的其他算法。

The characteristics of color image segmentation are adapted by the revising GrabCut algorithm,so SAR image is transformed into the format of 24 bit BMP in the case without any interaction,and furthermore the target segmentation of SAR image is implemented by neighbor region growing algorithm.By the test of the SAR image of MSTAR and NASA＇ jet propulsion laboratory,experimental results show that the proposed method outperforms the other algorithms for target obtaining at present.