PAFnet

Machine Learning for Synthetic Aperture Radar Autofocus


PAFnet

This is a PyTorch implementation of paper “AFnet and PAFnet: Fast and Accurate SAR Autofocus Based on Deep Learning”. pdf, doi

Autofocus plays a key role in synthetic aperture radar (SAR) imaging, especially for high-resolution imaging. In the literature, the minimum-entropy-based algorithms (MEA) have been proved to be robust and have been widely applied in SAR. However, this kind of method needs hundreds of iterations and is computationally expensive. In this paper, we proposed a non-iterative autofocus scheme based on deep learning and minimum-entropy criterion. It’s an unsupervised framework, which utilizes entropy as the loss function. In this scheme, deep neural networks are utilized for feature extraction and parameter estimation. Based on this scheme, two autofocus models (autofocus network and progressive autofocus network) are proposed. After training, the network learned the rules of autofocus from a large number of examples. Experimental results on real SAR data show that the proposed methods have focusing quality close to the state-of-the-art but with real-time focusing speed.

This code open-sourced fully now!

Dataset

Please see our paper and dataset page for details.

Prepare

Please update the value of SAR_AF_DATA_PATH in ‘data.yaml’ to your dataset path.

Please install our newest version of package torchbox and torchsar by pip install torchbox torchsar.

Training

python train.py --lr 1e-2 --num_epoch 1000, --device 'cuda:0'

Testing

python test.py --model_file [TRAINED MODEL PATH] --device 'cuda:0'

Visualization

The features of the above image extracted by AFnet (after 12-th layer) are shown as follows:

Features Extracted by Focuser0 Of Scene1

Features Extracted by Focuser0 Of Scene2

Features Extracted by Focuser0 Of Scene3

Citation

If you find the dataset or this code is useful, please kindly cite our paper and star our pakcage AutofocusSAR on GitHub::

@article{Liu2022PAFnet,
  title={AFnet and PAFnet: Fast and Accurate SAR Autofocus Based on Deep Learning},
  author={Liu, Zhi and Yang, Shuyuan and Gao, Quanwei and Feng, Zhixi and Wang, Min and Jiao, Licheng},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  year={2022},
  volume={60},
  number={},
  pages={1-13},
  doi={10.1109/TGRS.2022.3217063}
}