一种基于最大后验框架的聚类分析多基线干涉SAR高度重建算法

A Novel Cluster-Analysis Algorithm Based on MAP Framework for Multi-baseline In SAR Height Reconstruction

多基线干涉SAR能有效减小由目标高度急剧变化和较大噪声干扰带来的不利影响,可以获取比单基线干涉SAR更精确的地表数字高程模型(DEM)。传统的基于最大似然估计(ML)的多基线高度重建算法在通道数目较少情况下重建结果不佳,基于最大后验估计(MAP)的多基线高度重建算法存在运行时间较长的问题,针对以上问题,该文提出了一种基于最大后验框架的聚类分析高度重建算法(CABMAP)。该算法首先利用了ML估计法得到粗略的DEM,以此为基础在每次迭代过程中利用聚类分析(CA)判断出邻域内的噪声像素,并通过计算后验概率完成重建,此外采用了一种改进措施提高精度。这样,既保留了ML估计法运行速度快的特征,又具有MAP估计法精度高的优点。经实验验证,该算法精度较好且运行效率较高。

The multi-baseline In SAR can effectively reduce the adverse effect caused by the abrupt change of the target and the large noise disturbance and can obtain the Digital Elevation Model（DEM） that is more accurate than the single baseline In SAR. Traditional multi-baseline height reconstruction algorithm based on Maximum Likelihood（ML） estimation is poorly reconstructed in the case of fewer channels, and the height reconstruction algorithm based on Maximum A Posteriori estimation（MAP） has a long runtime defect; to solve this problem, this study proposes the cluster analysis based on maximum a posteriori algorithm. This algorithm uses the ML estimation to obtain a rough DEM. Based on this result, the noise pixels in the neighborhood in each iteration process are determined by cluster analysis. Finally, through the calculation of posterior probability to complete the reconstruction, an optimized method is adopted to improve the accuracy.Experiments reveal that the algorithm retains the speed of the ML method as well as the high precision of the MAP estimation, thus maintaining accuracy and high operating efficiency.