Research on brightness temperature reconstruction of synthetic aperture based on Resnet

In order to further study the performance of deep learning network method in synthetic aperture bright temperature reconstruction, this paper proposes an inversion algorithm using residual network (Resnet) for synthetic aperture bright temperature reconstruction. Through simulation, the corresponding visibility function is calculated from the original bright temperature data and the systematic error and random error are added. Then the original bright temperature data and the visibility data with error are put into the network model for training, hoping to get the mapping relationship between them. The Resnet inversion method obtained a better reconstructed image, with the smallest RMSE(6.28K) and the largest PNSR (31.62dB) of the reconstructed image. Compared with traditional bright temperature inversion method, the bright temperature image can be restored better and the image resolution can be improved.