In this paper, an intelligent electromagnetic environment reconstruction method is proposed based on a super-resolution generative adversarial network (SRGAN). The altitude matrices together with the low-resolution matrices obtained by measured power values are employed as inputs. Then, an electromagnetic environment reconstruction method capable of generating the high-resolution power coverage matrix in the selected area is designed. Data enhancement is employed to expand the dataset and a modified generator network with squeeze and excitation modules is used during the training process. To validate the proposed method, the simulation analyses are carried out in typical suburb environments based on a ray-tracing tool. Numerical results indicate that, compared with classical methods such as nearest-neighbor interpolation, bilinear interpolation, and bicubic interpolation, the proposed method can provide more accurate reconstruction results for power coverage. In addition, the peak signal to noise ratio (PSNR) of the proposed method is higher than those of the classical methods. The proposed intelligent electromagnetic environment reconstruction method can be useful for the planning, deployment, and optimization of wireless networks.

本文提出了一种基于超分辨率生成对抗网络（SRGAN）的智能电磁环境重构方法。高度矩阵与通过测量的功率值获得的低分辨率矩阵一起用作输入。然后，设计了一种能够在所选区域中生成高分辨率功率覆盖矩阵的电磁环境重建方法。数据增强被用来扩展数据集，并且在训练过程中使用带有挤压和激励模块的改进的发电机网络。为了验证所提出的方法，在典型的郊区环境中基于光线跟踪工具进行了仿真分析。数值结果表明，与经典方法（例如最近邻插值，双线性插值）相比，以及双三次插值，该方法可以为功率覆盖提供更准确的重建结果。另外，所提出的方法的峰值信噪比（PSNR）高于经典方法。所提出的智能电磁环境重建方法可用于无线网络的规划，部署和优化。