Study On Single/Multi-Baseline Phase Unwrapping

Phase unwrapping (PU), an important image processing technique, has been applied in many fields of interferometric measurement, such as, synthetic aperture radar interferometry (InSAR), magnetic resonance imaging (MRI), optical interferome-try, and so on. This dissertation, under the context of InSAR, makes a study on phase unwrapping technique with the combination of some key problems in the engineering application.This dissertation falls into two parts. The research on single-baseline phase un-wrapping technique is described in the first part, where some basic concepts of single-baseline phase unwrapping are introduced firstly, and then some technical problems in it are described with solutions proposed. Single-baseline phase unwrapping has been developed greatly in the recent decades, but it has the "nature disadvantage" in the recovery of the absolute phase in complicated areas (such as, valleys and steep mountains). For this reason, the multi-baseline phase unwrapping technique is intro-duced in the second part, where the basic principle of multi-baseline phase unwrap-ping is described firstly. In addition, some multi-baseline phase unwrapping algorithms are proposed.The main content of this dissertation is summarized as follows.1. In chapter2, an Outlier-Detection based single-baseline phase unwrapping method is proposed. It is known that the L0-norm phase unwrapping method is sta-tistically the best strategy, and its phase unwrapping solution is the most desired in practice. Unfortunately, Chen and Zebker have investigated that the L0-norm is an NP-hard problem, i.e., which is very difficult to find an exact solution in a polynomial time. In order to effectively solve this problem, PU process is abstracted as a process of solving an inconsistent equation system firstly, and then the equation correspond-ing to branch cuts of the L0-norm in the inconsistent equation system are considered as the outliers. Furthermore, after the removal of the outliers by the density-based Outlier-Detection technique in data mining, the approximation phase unwrapping so-lution of the L0-norm can be achieved by solving the remain equations. The significant difference from other conventional approximation algorithms is that the approximation phase unwrapping solution of the L0-norm is not obtained from branch cuts of the L0-norm directly but from the mother-set of branch cuts of the L0-norm. That is to say, the Outlier-Detection based single-baseline phase unwrapping method can obtain the L0-norm phase unwrapping solution without solving an NP-hard problem directly. A set of experimental results shows that the proposed method is effective in phase un-wrapping.2. In chapter3, two Residues Cluster-based tiling strategies are presented. So far, many phase unwrapping methods have been proposed. These methods are ad-vanced in different aspects, such as the accuracy of the solution or the speed of the algorithm. However, the limitation of computer’s memory size is ignored in the design of most of these methods. With the rapid development of the InSAR technology, the interferograms are becoming larger and larger. When the size of the interferogram ex-ceeds the limitation of computing capability, adopting divide-and-conquer policy dur-ing the phase unwrapping process is unavoidable in practice. Under this condition, whether the phase unwrapping result of each tile is consistent with that of the whole image or not becomes a new challenge for phase unwrapping. In this dissertation, a tiling strategy according to the clustering characteristic of residue distribution is pro-posed to assist the L0-norm large-scale phase unwrapping, which can approximately ensure the consistency between local and global phase unwrapping solutions of the L0-norm. Then, the other tiling strategy for the L1-norm large-scale phase unwrap-ping is proposed, which is the extension and improvement of the previous tiling strat-egy. This tiling strategy can exactly ensure the consistency between local and global phase unwrapping results of the L1-norm. Both theoretical analysis and experiments demonstrate that these two tiling strategies are effective for the large-scale phase un-wrapping.3. In chapter4, an L1-norm based multi-baseline phase unwrapping method is given. Single-baseline phase unwrapping is an inverse problem, whose signifi-cant character is that there are an infinite number of different solutions. In order to find the unique solution to the phase unwrapping problem, the phase continuity as-sumption is proposed. In essence, this assumption demands the measured regions have the space continuity. However, this assumption does not hold in any actual terrains, such as, valleys and steep mountains. In order to solve this problem, the multi-baseline phase unwrapping technique is proposed. By taking conventional L1-norm based single-baseline phase unwrapping method as a base, and combining the relation among interferograms obtained by multi-baseline InSAR, an L1-norm based multi-baseline phase unwrapping method is proposed in this dissertation. Simulated experiments demonstrate that this method is of great accuracy on phase unwrapping results and it can be applied to measure the complex terrains.4. In chapter5, the Cluster-Analysis-based multi-baseline phase unwrapping method is put forward. Multi-baseline phase unwrapping technique can break the lim-itation of phase continuity assumption, but it has a disadvantage that it is not robust to noise. Moreover, because multi-baseline phase unwrapping technology processes more than one interferograms, great pressure of execution speed and memory re-quirement exits in the multi-baseline phase unwrapping processing. This dissertation first makes an analysis of the combined information of multi InSAR interferograms, and then presents the Cluster-Analysis-based multi-baseline phase unwrapping method. The pixels with the same ambiguity vectors are clustered firstly by this method and the phase of pixels are unwrapped group by group. Simulation results show the pro-posed method has advantage in efficiency and noise robustness over the conventional methods.