On the Use of Deep Convolutional Neural Networks in Microwave Imaging

In this work, intermediate-sized deep convolutional neural networks are investigated to solve inverse scattering problems for microwave imaging. The conventional approaches using inversion algorithms encounter challenges such as high contrast and high computational costs. Thus, various deep-learning techniques have been proposed recently to tackle these issues. Different network architectures, namely the DCEDnet, Unet-Lite, and Unet, have been implemented and tested with complex (real and imaginary) inputs and outputs. The inputs are the backpropagation of the measured scattered fields onto the imaging domain. The outputs are the real and imaginary relative complex permittivity values of objects of interest. The simulation results from different networks are compared against each other and against the conventional contrast source inversion algorithm. The findings indicate that the applied deep learning techniques can generate image reconstructions of superior quality while significantly reducing computational time.