基于CNN-Transformer架构的电磁传播损耗预测算法

打开文本图片集

关键词：电磁传播；损耗预测；Transformer；CNN；斯皮尔曼系数法；地物类型；位置编码中图分类号：TN03-34；TP391.9 文献标识码：A 文章编号：1004-373X（2026）06-0043-06

Electromagnetic propagation lossprediction algorithm based on CNN-Transformerarchitecture

WANYong，LIJunjie，SUNWeifeng，DAIYongshou （CollgeofOceanographyand SpaceInformatics，China UniversityofPetroleum，Qingdao26658o，China）

Abstract：Inallusiontotheissueof insuffcient predictionaccuracyintraditional empirical propagationloss models，an electromagneticpropagationlosspredictionalgorithmbasedontheCNN-Transformerarchitectureisproposed，whichcarealize theaccuratepropagationloss predictionbyconstructingaregressionmodel.Theeffectivefeaturesareextractedbymeansof the Spearmacorrlationmethod，andCNNisused toextractshallowfeatures thatarehighlycorrelatedwithpropagationloss prediction.Thepositionalencoding isconductedonthesequence ofland featuresalong thepropagationpath obtained from satelite mages，toenhancetheunderstandingoftheimpactoftheorderofdiferentlandfeaturesinthepropagationpathon propagationlossTheshallow featuresextracted byCNNandtheposition-encoded land featuresequenceare input into the Transformermodel，wherethemulti-headself-atentionmechanismisusedtocapturetheglobalrelationshipsbetweenfeatures， efectivelycorrecting thepredictionresultsof thepropagationlossTheexperimentalresultsshowthattheproposedCNNTransformer methodcansignificantlyreducetherotmeansquareeror（RMSE）of propagation lossprediction to 3.3745dB， while maintaining a high coefficient of determination （ R2 ）of 0.895 6. The proposed electromagnetic propagation loss prediction algorithm can provide a valuable reference in this field and has a certain application value.

Keywords：electromagneticpropagation;lossprediction；Transformer;；CNN；Spearmancorelationcoefficientmethod; terrain type;position encoding

0 引言

电磁信号广泛用于通信、雷达、导航等多个领域，电磁传播损耗是指电磁信号在传输过程中受到各种因素影响，导致信号强度发生衰减的现象。（剩余8022字）