2-D Phase Unwrapping Based on Fringe Line Detection

Abstract:
Phase unwrapping is crucial in interferometric synthetic aperture radar (InSAR) data processing. Earlier efforts aimed to perform phase unwrapping by detecting fringe lines under a phase continuity assumption. However, noise can distort or create fringe lines, complicating their correct identification. Capitalizing on the robustness of deep learning, a convolutional neural network (CNN) is used for precise fringe line detection in this study, followed by specific post-processing for phase unwrapping. Experiments demonstrate this method's ability to accurately identify fringe lines even in high-noise conditions.