一种基于深度学习的电波传播场强预测方法  

A method for predicting the field strength of radio wave propagation based on deep learning

作　　者：孙绍哲 岑逸翔 谢玮松 SUN Shaozhe;CEN Yixiang(Army Engineering University of PLA,Nanjing XIE Weisong 210007,China)

机构地区：[1]陆军工程大学,江苏南京210007

出　　处：《通信与信息技术》2024年第4期7-9,67,共4页Communication & Information Technology

摘　　要：电波传播场强预测对于无线网络的规划具有重要意义,其预测精度与通信环境的大规模传播特性直接相关。为了提升电波传播场强预测的准确性,需尽可能地增加输入特征维度以最大化描述环境。卫星图像能直观地展示研究区域的环境特征,将其作为输入特征,采用卷积神经网络构建模型准确地预测电波传播情况。因此,提出了一种基于卫星图像数据的深度学习场强预测方法,输入数据为经纬度、高程和卫星图像,输出为预测场强。结果表明,与未使用卫星图像作为输入的模型预测值对比,添加卫星图像作为输入特征后,预测精度提高15.2%,对卫星图像数据进行增强,预测精度获得6.4%的提升。The prediction of radio wave propagation field strength is of great significance to the planning of wireless network,its prediction accuracy is directly related to the large-scale propagation characteristics of communication environment and.In order to im⁃prove the accuracy of radio wave propagation field intensity prediction,it is necessary to increase the input characteristic dimension as much as possible to maximize the description of the environment.The satellite image can intuitively display the environmental charac⁃teristics of the study area and take them as input characteristics.The convolutional neural network is used to build a model to accurately predict the radio wave propagation.Therefore,a deep learning field strength prediction method based on satellite image data is proposed in this paper.The input data are latitude and longitude,elevation and satellite image,and the output is predicted field strength.The results show that,compared with the model prediction value without satellite image as input,the prediction accuracy is improved by 15.2%after adding satellite image as input feature,and the prediction accuracy is improved by 6.4%when the satellite image data is enhanced.

关 键 词：电波传播场强预测 卫星图像 卷积神经网络 数据增强 

分 类 号：TN011[电子电信—物理电子学]