论文部分内容阅读
本文提出了一种基于证据理论的融合像素信息和上下文信息的多极化合成孔径雷达(SAR)图像分割方法。D-S证据理论是一种不确定性推理方法。基于D-S的SAR影像分割方法将像素信息和上下文信息看作两类证据,先对高度平滑后的影像作初始的过分割,然后基于D-S理论对初始分割图斑的边界进行迭代修正,最后再融合两类证据对初始分割的图斑进行合并。本文方法比传统基于像素信息的方法具有更强的抗噪性能。横断山脉的双极化SAR影像分割实验表明,该方法对噪声敏感度相对较低,可以获得较可靠的分割结果。
In this paper, a multipolarized synthetic aperture radar (SAR) image segmentation method based on evidence theory and fusion of pixel information and context information is proposed. D-S evidence theory is a method of uncertainty reasoning. The SAR image segmentation method based on DS considers the pixel information and the contextual information as two types of evidence. Firstly, the highly smoothed image is initially over-segmented, then the boundary of the initial segmentation patch is iteratively modified based on the DS theory, and finally fused The two types of evidence merge the initially segmented speckles. The proposed method has stronger anti-noise performance than the traditional pixel-based method. Experiments on the dual-polarized SAR image segmentation of the Hengduan Mountains show that this method is relatively sensitive to noise and can obtain more reliable segmentation results.