条纹复杂且密集的低信噪比干涉图展开问题研究

Phase unwrapping of the serious noisy interferograms with complicated and dense fringes is hot and difficult problems in INSAR technology application, and will be studied in this project. Firstly, robust and effective phase gradient evaluation algorithms will be constructed to solve the problem of underestimate of phase gradient of interferograms, caused by phase noise, phase residues and non-stationary phenomenon of INSAR images. Secondly, transform phase unwraping into phase estmate problems by buliding high-order phase unwrapping system model under the frame of the adaptive unscented Kalman filter (AUKF), then the phase unwrapping algorithms based on the optima information fusion under the frame of high-order AUKF will be proposed to unwrap the serious noisy interferograms with complicated and dense fringes by combining an AUKF with an onmi-directional local phase gradient estimation technology and a path-following search strategy. Thirdly, using Kalman particle filtering (KPF), cubature particle filtering (CPF) algorithm to overcome the problems that phase unwrapping precision is constrained by noise statistics and nonlinearity of the model, then the effective region-growing phase unwrapping algorithms under the frame of Bayesian will be proposed to further improve the accuracy and the efficiency of phase unwrapping by integrating fast hierarchical path-following search strategy and a region-growing strategy, which will be helpful to solve the difficult problem of balancing the accuracy and the efficiency of phase unwrapping. Finally, the proposed methods will be performed with simulation and real data.

本项目针对INSAR技术应用研究中的热点与难点问题—条纹复杂且密集的低信噪比干涉图的展开问题进行研究。研究高效和稳健的干涉图相位梯度估计算法，解决由干涉相位噪声、相位残差点以及干涉图非平稳现象等因素导致的相位梯度欠估计问题；通过建立自适应无味卡尔曼滤波（AUKF）框架下的高阶相位展开系统模型，把相位展开问题转化为状态估计问题，结合全方位局部相位梯度估计技术以及路径搜索跟踪策略，提出高阶AUKF框架下基于最佳信息融合的相位展开算法，解决条纹复杂且密集的低信噪比干涉图的展开问题；利用卡尔曼粒子滤波、容积卡尔曼粒子滤波等算法克服问题非线性与模型噪声统计特性对相位展开精度的制约问题，结合分层快速路径搜索跟踪策略和区域增长策略，提出贝叶斯框架下的高效区域增长相位展开方法，进一步提高相位展开精度及效率，解决干涉相位展开中精度与效率难以兼顾的问题。最后，完成模拟与实测干涉图相位评估实验，验证算法有效性。

本项目针对InSAR技术应用研究中的热点与难点问题—条纹复杂且密集的低信噪比干涉图的展开问题进行研究。研究稳健和高效的干涉图相位梯度估计算法，解决由干涉相位噪声、相位残差点以及干涉图非平稳现象等因素导致的相位梯度欠估计问题，有利于提高干涉图相位解缠技术的精度；针对现有方法稳健性和适应性不够，难以有效解决条纹复杂且密集的缠绕干涉图的展开问题，把相位展开问题转化为无迹卡尔曼滤波（UKF）框架下的状态估计问题，提出UKF框架下基于最佳信息融合的相位展开算法，解决条纹复杂且密集的低信噪比干涉图的展开问题；利用无迹卡尔曼粒子滤波、容积卡尔曼粒子滤波、容积信息粒子滤波等算法克服问题非线性与模型噪声统计特性对相位展开精度的制约问题，结合快速路径搜索跟踪策略，提出贝叶斯框架下的粒子滤波相位展开算法，进一步提高相位展开精度；此外，对利用深度学习神经网络技术来解决干涉图相位解缠的问题进行了研究，并在模拟与实测干涉图解缠实验中获得较好的结果；最后，完成模拟与实测干涉图相位解缠实验，验证算法有效性。本项目研究成果，有望在InSAR技术应用、 数字全息成像以及光学干涉测量技术应用等方面获得良好的应用前景。