A Learning-Based Inversion Method for Phaseless Profiling of Circular Buried Objects

A deep learning algorithm for phaseless microwave imaging of circular buried targets, is proposed in this paper. To that end, a Convolutional Neural Network (CNN) is implemented, in which the amplitude of the measured scattered electric field is directly fed to the network as the input. By using phaseless data, the measurement setup cost and complexity can be reduced at the expense of bolding the ill-posedness challenge. To reduce the output unknown parameters, the commonly used assumption of circular cross-section is considered for defining the scatterers. For tackling the required large data set, the fast CG-FFT method is used for generating synthetic data. The obtained results reveal that the proposed simple CNN-based technique can accurately predict in real-time the location, radius, and the dielectric permittivity value of the subsurface objects of circular shapes.