Deformation-Monitoring

The source code of Paper: Deep learning for localized rapid deformation detection and InSAR phase unwrapping

Authors: Zhipeng Wu, Student Member, IEEE, Teng Wang, Yingjie Wang, Robert Wang, Senior Member, IEEE, Daqing Ge

**The trained PUNet is mainly applied to unwrap very localized, rapidly deforming areas, particularly the mining subsidence as we stated in the manuscript. The proposed dataset generation strategy allows for richer samples, which can enable the retrained PUNet to cope with more complex situations. **

If you want to retrain the networks, see Deformation-Monitoring-Dev.

For code to generate the training dataset, see InterferogramSimulator.

Introduction

This is a TensorFlow (1.13.1) implementation of PUNet. It can be used to unwrap interferograms with dense fringes and severe noise. The resolution of the input interferogram should be about 20m×20m. Any size of the interferogram can be input, but 180×180 pixels is recommended. The input data should be binary files (*.wzp, float32) or Tiff files (*.tif, float32).

For reading and writing binary files, see imreadbin.m and imreadbin.m in MatlabToolbox.

Installation

The code was tested with Python 3.6.

1. Clone the repo:

   git clone https://github.com/Wu-Patrick/Deformation-Monitoring.git
   cd Deformation-Monitorin

2. Install dependencies:

   pip install -r requirements.txt

Unwrapping

1. Input arguments: (see full input arguments via python runPUNet.py --help):

usage: runPUNet.py [-h] [--input INPUT] [--output OUTPUT]
                   [--outputPng OUTPUTPNG] [--rows ROWS] [--cols COLS]
                   [--ckpt_dir CKPT_DIR]

2. Run the example：

python runPUNet.py
# or
python runPUNet.py --input data/dataset1/interf --output data/dataset1/unwrapped --outputPng 1 --rows 180 --cols 180
# or
python runPUNet.py --input data/dataset2/interf/300_20190120-20190201.wzp --output data/dataset2/unwrapped --outputPng 1 --rows 300 --cols 300

Citation

If you use this code, please cite the following:

@ARTICLE{9583229,
  author={Wu, Zhipeng and Wang, Teng and Wang, Yingjie and Wang, Robert and Ge, Daqing},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Deep Learning for the Detection and Phase Unwrapping of Mining-Induced Deformation in Large-Scale Interferograms}, 
  year={2022},
  volume={60},
  number={},
  pages={1-18},
  doi={10.1109/TGRS.2021.3121907}}

Acknowledgement

TensorFlow

Statement

The code can only be used for personal academic research testing.