基于随机并行梯度下降算法的InSAR相位解缠方法

相位解缠是干涉合成孔径雷达InSAR数据处理中的一个关键步骤,解缠结果的好坏直接影响最终数字高程模型的精度。介绍了一种基于随机并行梯度下降SPGD算法的解缠方法,该方法对图像中各相位点施加随机并行扰动,通过迭代使得解缠误差代价函数收敛到全局最优值,从而实现相位解缠的目的。模拟和实测数据实验结果表明,相较于最小二乘解缠方法,随机并行梯度下降解缠算法精度更高,且原理简单,易于实现,为相位解缠提供了一个全新的思路。

Phase unwrapping for InSAR based on stochastic parallel gradient descent algorithm

Phase unwrapping is one of the key data processing procedures for interferometric Synthetic Aperture Radar (InSAR), which can directly affect the final precision of the digital elevation model. A phase unwrapping method based on Stochastic Parallel Gradient Descent (SGPD) algorithm is proposed in the paper, which imposes the stochastic parallel perturbation on every point of the phase images and achieves the phase unwrapping through iteration by which the error cost function is converged to its global extremum. Experimental results of simulated and real data indicate that the SPGD based phase unwrapping method is more precise, simple and easy to realize, and provides a total new solution to phase unwrapping.