Multilevel Perception Sequential Phase Optimization for Distributed Scatterer InSAR Steep Deformation Monitoring

Phase optimization is a key part in distributed scatterer (DS) interferometric synthetic aperture radar (SAR) technique. It aims to weaken the influence of decoherence factors on DS and retrieve the optimal phase series from the SAR stack. However, popular approaches have not considered the regional surface deformation difference. Resulting in statistically homogenous pixels (SHPs) are often limited to a fixed-size information block (IB) to participate in the optimization. This limitation might render it challenging to strike an optimal balance between phase noise robustness and interferometric fringe integrity. To solve the above problem, a multilevel perception sequential phase optimization algorithm (MPSEO) is proposed in this paper. MPSEO broke through the dimensional limitation of IBs by spatially extending the original SHPs, and adopted relative entropy to perform similarity perception fusion on each homogeneous IB to enhance the phase signal-to-noise ratio. Meanwhile, the Laplacian-like operator was employed to adaptively perceive the size of IB based on the intensity of surface deformation, while also ensuring the protection of phase in deformation zones. Finally, a sequential compression strategy was applied to the SAR stack for phase estimation. The flexible SHP refactoring mode in proposed MPSEO can produce more robust complex coherence matrix, further boost optimization performance. Experimental results show that the proposed MPSEO has excellent noise reduction and phase information recovery effect compared with traditional approaches. Moreover, it can also extract 41% stable DS pixels more than popular approach, showing the global superiority.