We proposed a prior-induced interference suppression network (PISNet) to achieve RFI suppression and useful signal recovery in time-frequency domain. A dataset called SarTF is built to address the difficulty of data shortage to support the training of PISNet.
A dataset of the time-frequency spectrograms of SAR echoes with RFI, which can be used to train the end-to-end deep neural networks.
- Download
BaiduNetdisk: https://pan.baidu.com/s/12arHdlsZFg0lfrwxZYYiwA Password: i7zl
- Currently the network is suitable for dealing with interference intensity similar to that in SarTF Dataset. If the difference of the Signal-Interference-Ratio (SIR) is significant (eg. the SIR too small), the processing performance may be degraded. The SIR of the training set should be adjusted accordingly.
- For specific training details, please refer to the experiment and discussion chapters of our paper.
If you want to use this SarTF dataset or use PISNet as contrast model, please cite as follows
J. Shen, B. Han, Z. Pan, G. Li, Y. Hu and C. Ding, "Learning Time–Frequency Information With Prior for SAR Radio Frequency Interference Suppression," in IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-16, 2022, Art no. 5239716.