Long-Distance Terahertz Single-Photon 3-D Imaging Based on Scaling Training

Abstract

In this paper, to address the long-distance imaging problems, a novel forward-looking three-dimensional (3D) imaging method based on terahertz single-photon radar is presented. The imaging method is based on SISO architecture system and adopts a novel data-driven approach named scaling training. In the scaling training, by scaling the parameters of the longdistance scene to be imaged, a parameter scaling system is constructed, and then it is utilized to collect the training data to train the Artificial Neural Network (ANN) model. Once the training is completed, 3D image of the long-distance scene can be retrieved using the ANN model from solely the 1D photoncounting histogram echo recorded by terahertz single-photon radar. On the basis, to further improve the imaging quality, the depth correction method is implemented. Finally, numerical simulations are carried out to demonstrate the feasibility of the proposed method.