26 March 2021 in IEEE Geoscience and Remote Sensing Letters
The high-quality training data sets are often insufficient in synthetic aperture radar (SAR) automatic target recognition (ATR) applications. The generative adversarial network (GAN) provides a way for SAR data augmentation. It is necessary to ensure the diversity, similarity, and correct category of the generated images so that these images can be served as the supplementary data set. In this letter, the multiconstraint GAN (MCGAN) is proposed to generate high-quality multicategory SAR images. First, an encoder is used to learn the features of the real images to enhance the similarity. Then, the encoded features are mixed with noise and category labels as the input of the generator to improve the diversity and category correctness. The generated images will be sent to a pretrained classifier to ensure the correct category. Finally, the improved Wasserstein loss with the gradient penalty is extended to the model to further improve the diversity and similarity of the generated images. The MSTAR data set is used to validate the proposed method on generation. The quality evaluation and classification tests are performed on the generated images, and the results show that the MCGAN can provide high-quality images, which could assist in achieving good classification accuracy.
10 August 2021 in IEEE Geoscience and Remote Sensing Letters
The synthetic aperture radar automatic target recognition (SAR ATR) application based on the deep convolutional network often faces data scarcity. SAR image simulation based on electromagnetic and geometric calculations can provide a large amount of data that contains interpretable physical features, such as shadow and contour. However, there is a big difference between the simulated SAR image and the real image, and it is difficult to directly use it for data augmentation. This letter proposes the adversarial encoding network to extract the physical-related features, which can be understood as the common features between the simulated and real data. By designing the adversarial learning between an encoder and a discriminator, the encoder can extract real features from the simulated images. The encoded features are sent to a classifier to ensure the correct category information. A decoder is used to reconstruct the encoded features into the input image so that the encoded feature retains the image information as much as possible. Ablation experiments and comparative experiments are used to verify the ability of each module and the performance of the proposed method. The results show that the proposed model can achieve 98.55% accuracy, especially when the real data are insufficient for classification, which verifies that the proposed method is effective for data augmentation.
15 October 2021 in IEEE Geoscience and Remote Sensing Letters
The realization of radiometric calibration relies on the known artificial point targets as the reference, such as corner reflectors. One of the criteria of the calibrator is that the 3 dB-beamwidth of the azimuth radar cross section (RCS) pattern needs to be larger than the synthetic aperture radar (SAR) antenna 3-dB beamwidth to ensure the SAR antenna, which has a narrow azimuth beamwidth, can align to the calibrator while also making the nominal RCS value of the calibrator approximately constant during the radiometric calibration. However, the RCS of the calibrator is actually inconstant within a synthetic aperture period, especially when the SAR antenna beamwidth is wide. The simplified treatment will introduce errors to calibration. In this letter, the azimuth RCS pattern is taken into account. The error caused by the constant RCS is modeled. Based on the error model, an improved criterion of the calibrator is discussed. The experiments use the known SAR system parameters for numerical simulation based on dihedral corner reflectors (DCRs) with different sizes. The results verify the error model and indicate that when the SAR beamwidth is larger than the beamwidth of the calibrator, and the introduced error cannot be ignored. The relationship between the size of the calibrator and the SAR beamwidth is further analyzed through the experiments.
27 October 2020 in IEEE Geoscience and Remote Sensing Letters
In the bistatic synthetic aperture radar (SAR), the comodulation effect induced by transmitting and receiving antennas to the image amplitude is the main factor affecting the radiometric accuracy. Therefore, the corresponding mechanism must be determined, and a correction method must be developed. However, there is no effective calibration target and method for measuring the bistatic SAR range round trip antenna pattern. In this letter, it is pointed out that the comodulation effect is not only related to the transmitting and receiving one-way antenna patterns but also to the range variation of the bistatic angle in the swath caused by the bistatic SAR observation geometry. To increase the radiometric accuracy in bistatic SAR images, a comodulation effect correction method based on accurate antenna models is proposed, and different error factors related to the range round trip antenna pattern are analyzed. Moreover, simulation experiments are conducted to verify the effectiveness of the proposed method.