Equivalent Magnetic Dipole Matrix Reconstruction Method Based on Deep Learning for Magnitude-Only Near-Field Scanning

Least square method is traditionally used for establishing equivalent dipole matrix in electromagnetic interference problems. However, the accurate phase information is hard to be tested in the real measurement. A new deep learning method is proposed in this paper for dipole matrix reconstruction without considering the phase information on the near-field plane. Compared with least square method, a well-trained convolutional neural network model provides a higher accuracy and efficiency performance.