Improved Method for Planar Near-Field Measurement Based on Complex Convolutional Neural Networks

To achieve higher accuracy in planar near-field measurement, an improved approach based on complex convolutional neural networks is proposed, along with a dataset generation method for model training in this work. This improved method can calculate the spherical wave coefficients from the planar sampling data and then calculate the far-field pattern of the antenna under test. By employing a reasonable model structure and training design, the improved approach attains higher accuracy than conventional method on practical data, offering a new approach to enhance the planar near-field measurement.