Study On Multitemporal Small BAseline Subset(SBAS) DInSAR Technique

Spaceborne Differential Synthetic Aperture Radar Interferometry(DInSAR) is a well-established microwave remote sensing methodology which has been developed in recent 30 years. As an observation tool from space, it provides the possibility to investigate large-scale earth surface deformation phenomena with high precision and at low cost. However, the application of conventional DIn SAR technique is somewhat limited when the decorrelation noise and atmospheric delay variations are severe enough to deteriorate the measurement accuracy.As a typical multitemporal DInSAR technique developed in last decades, the Small BAseline Subset(SBAS) technique can achieve highly precise temporal evolution of the displacements relevant to the investigated area by effectively overcoming the key limitation of conventional DInSAR. Accordingly, it is exploited in a wide range of different earth surface deformation surveillance scenarios, offering very relevant measurements for early warning and evaluation of geologic hazard, geophysical modeling, government programming and decision-making, as well as military highly precise mapping. This thesis, aimed at improving the observation accuracy and robustness of SBAS technique in low coherence area, studied and proposed solutions for several key issues relevant to SBAS processing chain. Moreover, the effectiveness of the developed approaches is tested by carrying out a number of real SAR data experiments. The main work of this thesis is organized as follows:Chapter II introduces the theory of DInSAR and SBAS techniques. We firstly analyzd the components of interferometric phase and described the principle of conventional DInSAR technique. Then, the statistic distribution model relevant to interferometric phase is established and several decorrelation sources are distinguished. Finally, the rationale of SBAS technique and its processing chain are fully described.Chapter III is dedicated to the presentation of the rationale of the multitemporal interferogram noise-filtering. Firstly, the statistical characteristic of distributed targets is introduced, concluding that the phase of such targets does not satisfy time-consistence property. Subsequently, a multitemporal noise-filtering algorithm based on statistic estimation is discussed. Algouth the algorithm can effectively recover the fringe in low coherence area and improve the overall coherence of interferometric sequence, by jointly applying temporal redundant information among interferometric sequence, it suffers from the drawback of its high computational burden, limited application range and causal information exploitation. To overcome these problems, we proposed a new algorithm on the basis of directional statistics. This new approach, which does not rely on any statistic hypothesis relevant to interferometric phase, can directly be applied to a sequence of conventional multilook interferograms, without the need to implement a time consuming pixel-by-pixel statistic test procedure aimed at distinguishing point-like and distributed targets, thus, allowing to significantly reduce the overall computation time. Moreover, by taking profit from using only small baseline interferograms as input, the new filtering method can reduce the phase estimation errors due to decorrelation noise. The performance of the developed method is confirmed by the overall coherence improvement of filtered sequence of interferograms relevant to two real SAR dataset.In chapter IV we studied the issue of interferogram selection in the framework of multitemporal DInSAR technique. As the precondition of multitemporal DInSAR technique to monitor deformation phenomenon, the way to select proper interferograms from obtained SAR dataset will directly affect the accuracy of achieved deformation products. Firstly, the nowadays typical interferogram selection methods as well as their impact on deformation investigation accuracy are discussed. Subsequently, by introducing the cost function that can define the overall coherence of interferometric sequence in various coherence condition, we proposed a method to select the interferometric SAR data pairs based on Simulated Annealing(SA) which is fully compatible with the SBAS technique. The developed methodology defines the interferograms by searching the triangulation that minimizes the decorrelation noise in the temporal/perpendicular baseline plane. The real data experimental results demonstrate that the selected interferometric sequence by exploiting the proposed approach have higher overall coherence which can achieve better phase unwrapping performances.In chapter V, the multitemporal Phase Unwrapping(Ph U) algorithms based on Minimum Cost Flow(MCF) technique are studied. Firstly, we discussed the theory and principle of basic MCF, extended MCF as well as divide-and-conquer MCF Ph U algorithms, respectively. With respect to the issue that high quality Ph U results on spatial dense grids can be hardly achieved in low coherence area exploiting currently available Ph U algorithms, we proposed a region-growing Ph U approach based on MCF technique that can propagate the Ph U results from high coherence region to low coherence one. The real data experimental results demonstrate that the proposed method can effectively improve the spatial density of high quality Ph U pixels.In chapter VI we integrated the aforementioned new solutions into the enhanced SBAS processing chain, and carried out experiments on the set of 49 SAR images collected from 2003 to 2010 by the ENVISAT satellite over the Shanghai city, China. Comparing with the results obtained by currently available SBAS technique, the new processing chain can achieve deformation products with much denser grid. Accordingly, by taking profit from spaceborne DIn SAR measurements, it provides the opportunity to obtain more reliable information for the research, analysis, understanding and modeling of the deformation phenomena relevant to the investigated area.