Abstract:Interferometric Synthetic Aperture Radar (InSAR) enables efficient retrieval of surface elevation and finds extensive applications in terrain mapping. Dual/Multi-channel InSAR techniques utilize the differences in elevation ambiguity of different InSAR channels (baselines、frequencies) to achieve phase unwrapping (PU). This enables the effective application of InSAR in regions with abrupt terrain changes. In response to the growing demand for efficient and precise PU, this paper leverages deep learning and proposes a dual/multi-channel joint interferometric PU network: (Multi-Channel-Joint) MCJ-UNet. It effectively combines the multi-channel phase characteristics and their mutual constraint relationships. The proposed network is constructed based on the dual-channel (dual-frequency、dual-baseline) InSAR observation configuration. And it can also be extended to the case of multi-channel InSAR. The core ideas of the proposed method can be summarized into three aspects. Firstly, it transforms the elevation ambiguity estimation problem in PU into a semantic segmentation, and the UNet network is employed to accomplish the segmentation processing. Secondly, the Squeeze-and-Excitation (SE) module is introduced to dynamically adjust the information weights, enhancing the network’s perception of the required information across different channels. Thirdly, a phase residual optimization loss function is employed under the context of multi-channel joint constraints to achieve network tuning. In addition, to mitigate the impact of edge details errors in semantic segmentation results on the PU performance, a self-correcting approach of PU errors based on multi-channel joint constraints is proposed. The proposed MCJ-UNet is verified by computer simulations based on simulated and real terrains, and experiments based on TerraSAR-X real data.

Phase Unwrapping of Speckle Interferometry Based on UCNet

Practical Phase Unwrapping of Interferometric Fringes Based on Unscented Kalman Filter Technique.

Single-frame Noisy Interferogram Phase Retrieval Using an End-to-end Deep Learning Network with Physical Information Constraints

A Hybrid Phase-Unwrapping Method For Optical Interferometry Based On New Parameter Map And Local Plane Approximation

PU-M-Net for Phase Unwrapping with Speckle Reduction and Structure Protection in ESPI

Efficient and Robust Phase Unwrapping Method Based on SFNet

Depth_ ASPP_ UNet++: A novel 2D ESPI phase unwrapping based on convolutional neural network

Phase Unwrapping Method for Point Diffraction Interferometer Based on Residual Auto Encoder Neural Network

UN-PUNet for Phase Unwrapping from a Single Uneven and Noisy ESPI Phase Pattern.

Convolution Network for Phase Unwrapping of Interferometric Fiber Sensing Signals

Multi task deep learning phase unwrapping method based on semantic segmentation

Joint Phase Unwrapping and Speckle Filtering by Using Convolutional Neural Networks

Performance of a U2-net Model for Phase Unwrapping

PUnet: A Robust Framework for Phase Unwrapping in Interferometric SAR

A Phase Filtering Method Based on Deep Learning Network.

Self-supervised Blind2Unblind Deep Learning Scheme for OCT Speckle Reductions

Phase Unwrapping in Optical Metrology Via Denoised and Convolutional Segmentation Networks

MCJ-UNet: A Dual/Multi-Channel-Joint Phase Unwrapping Network for Interferometric SAR

Deep Convolutional Neural Network Phase Unwrapping for Fringe Projection 3D Imaging.

Convolutional and fourier neural networks for speckle denoising of wrapped phase in digital holographic interferometry

Two-dimensional phase unwrapping by a high-resolution deep learning network