Discontinuity-guided Two-dimensional Phase Unwrapping Method for Large Gradient SAR Interferograms

In this paper, an InSAR phase discontinuity estimation network for areas with substantial elevation change ( ASEC-PDENet ) is developed based on the classic medical segmentation model U-Net. The ASEC-PDENet predicts the position of the phase discontinuity points in the interferogram, and then the phase discontinuity confidence score is used as a priori information of PUMA to obtain the unwrapped phase.

Problem Analysis

Fig.1
Fig.2

Proposed Method

Dataset Simulation Strategy

First we use the public dem dataset for registration and difference. The difference map is then back-calculated into differential phase. Finally, large gradient phase and random noise are added to it.
Fig.3
Fig.4

Network Structure

The main advantages of ASEC-PDENet are as follows : (1) The interactive compression module (ICM) is used to reduce the loss of down-sampling detail information and ensure the ability of the model to segment discrete phase discontinuous information ; (2) The cross fusion module (CFM) is used to fuse high-level and low-level features. This module highlight the phase gradient information features and position information, while suppressing noise.
Fig.5

Results

Ablation Experiments
Fig.6

Compare with Other Models
Fig.7

Performance Evaluation With Simulated Datasets
Fig.8

Performance Evaluation With InSAR Datasets
Fig.9
Fig.10

Discontinuity-guided Two-dimensional Phase Unwrapping Method for Large Gradient SAR Interferograms
http://example.com/2023/12/17/ASEC-PDENet-PU/
Author
Silevoem
Posted on
December 17, 2023
Licensed under