基于空间特征的SAR图像分割方法研究

Research on Spatial Matrix Based Method for SAR Image Segmenta tion

空间矩阵描述了不同区域间的相邻概率。SAR图像具有不同地域呈现不同纹理的特征,并且区域间有较大的对比度,因此SAR图像的空间矩阵具有同类区域间相邻概率最大,异类区域间相邻概率较小的特点。利用这一特点,该文提出了一种基于空间矩阵的图像分割的方法。该方法将空间矩阵作为适应度,用遗传算法可以使分割阈值收敛到最优,并在阈值搜索的过程中用孤立点检测的方法消除相干斑噪声的影响。仿真结果表明,这是一种有效的SAR图像区域分割方法。

Spatial matrix describes the probabilities that one area is the neighbor of other areas.Different areas in SAR images have different textural features and have big contrast.Therefore the spatial matrix of a SAR image has the character that the probability of the areas of same kind is the largest and the probability of the areas of different kinds is the smallest.By using this character,this paper proposes an image segmentation method based on spatial matrix.The method proposed utilizes Genetic Algorithms to search optimal threshold where spatial matrix is used as fitness.Moreover in the method,the outlier detection is used to eliminate the effect of correlative noises.Simulation results show that this method is effective for SAR image segmentation.