Non-Uniform Synthetic Aperture Radiometer Image Reconstruction Based on Deep Convolutional Neural Network

Abstract

When observing the Earth from space, the synthetic aperture radiometer antenna array is sometimes set as a non-uniform array. In non-uniform synthetic aperture radiometer image reconstruction, the existing brightness temperature image reconstruction methods include the grid method and array factor forming (AFF) method. However, when using traditional methods for imaging, errors are usually introduced or some prior information is required. In this article, we propose a new IASR imaging method with deep convolution neural network (CNN). The frequency domain information is extracted through multiple convolutional layers, global pooling layers, and fully connected layers to achieve non-uniform synthetic aperture radiometer imaging. Through extensive numerical experiments, we demonstrate the performance of the proposed imaging method. Compared to traditional imaging methods such as the grid method and AFF method, the proposed method has advantages in image quality, computational efficiency, and noise suppression.