MSTAR图像分割算法研究

Research on the Segmentation Algorithm of MSTAR Image

由于SAR图像中具有极其严重的乘性斑点噪声 ,不可能用纯粹的传统的基于强度的分割方法来获取目标。分析了马尔可夫随机场在各向同性的连续模型 ,提出了一种新的分割方法 ,采用该方法不仅分割出了目标而且较好地保持了其空间结构。通过形态学中的膨胀和腐蚀运算得到了感兴趣的目标区域 ,并给出了它的轮廓图。利用实际的运动和静止目标获取与识别 (movingandstationarytargetacquisitionandrecognition ,MSTAR)合成孔径雷达 (syntheticapertureradar,SAR) ,图像数据得到了较好的实验效果。

SAR imagery is extremely speckled, which makes traditional, purely intensity based segmentation techniques fail, the result exists much of speckle noise. This paper analyzes Markov random field (MRF) isotropic continuous model, and presents a new segmentation algorithm, which can obtain the target and keep the structure good. Applying the dilate and erode operation in morphology to SAR image, the target region of interest, and the target outline are obtained. A series of preferable experimental result about MSTAR SAR image data are given.