Polarimetric ISAR Super-Resolution Based on Group Residual Attention Network

Compared with optical imaging system, polarimetric inverse synthetic aperture radar (ISAR) can work all day and all-weather, which plays an important role in space surveillance. However, high-resolution ISAR images require large bandwidth and large coherent integration angle, which is limited by the physical conditions of the imaging system. In this vein, super-resolution of ISAR images is of vital importance. However, there is a lack of datasets for polarimetric ISAR super-resolution. Focused on this issue, a real physical degradation polarimetric ISAR super-resolution dataset are constructed utilizing electromagnetic simulation data of different bandwidths and different imaging apertures. On this basis, a group residual attention network (GRAN) for polarimetric ISAR image super-resolution is constructed. The main contribution of this work falls into three parts. Firstly, a polarimetric ISAR dataset is constructed. Secondly, since the polarimetric ISAR data is complex with multiple channels, the input data is divided into four groups for convolution processing to excavate the correlation between the real and imaginary parts of the complex data. Thirdly, a non-local attention module is added to the residual block to capture the long-range dependency and extract the channel relationship. Experiment results demonstrate that the proposed method has better visual quality and higher quantitative metric.