A non-fuzzy interferometric phase estimation algorithm based on modified Fully Convolutional Network

摘要

Classical phase unwrapping methods suffer from layover regions where phases are discontinuous and layover residues occur. To overcome this weakness, we proposed an interferogram-segmentation-assisted phase estimation method to minimize the influence of layovers. The modified Fully Convolutional Networks (FCN) is first applied to classify interferogram pixels into normal pixels and layover residues. By means of only taking normal pixels as the input of phase filtering and unwrapping steps, the optimized non-fuzzy phase is obtained. Results on simulated and real data verify that the proposed algorithm can effectively avoid the error propagation of residues in layovers, and significantly improve the precision of phase unwrapping. (C) 2019 Elsevier B.V. All rights reserved.