首页|期刊导航|计算机与现代化|基于深度学习的单辐射源三维定位方法

基于深度学习的单辐射源三维定位方法OA

Three-dimensional Single Radiation Source Localization Method Based on Deep Learning

中文摘要英文摘要

针对复杂山地与城市环境中非视距传播导致的单辐射源三维定位精度下降,以及传感器密集部署成本较高的问题,提出一种基于地理信息和深度学习的三维定位方法.首先,结合地理信息系统和主导路径模型构建三维电磁传播场景,并将稀疏传感器接收信号强度编码为三维体素矩阵;然后,采用融合ResNet残差模块的沙漏网络和加权损失函数,学习接收信号强度与辐射源三维位置之间的映射关系.实验结果表明,在传感器间隔为500 m×500 m×50 m的条件下,所提方法的平均三维定位误差为245.4 m,均方根误差为298.5 m,定位精度均优于MLP、FcCNN和SPCNN等基线模型.结果说明,该方法能够在大规模复杂环境和稀疏传感器条件下实现较稳定的单辐射源三维定位.

To address the reduced accuracy of three-dimensional localization of a single radiation source caused by non-line-of-sight propagation in complex mountainous and urban environments,as well as the high cost of dense sensor deployment,a three-dimensional localization method based on geographic information and deep learning is proposed.First,a three-dimensional elec-tromagnetic propagation scenario is constructed by integrating a geographic information system with the Dominant Path Model,and the received signal strength measurements collected by sparsely deployed sensors are encoded into a three-dimensional voxel matrix.Then,an hourglass network incorporating ResNet residual modules and a weighted loss function is employed to learn the mapping relationship between the received signal strength and the three-dimensional position of the radiation source.Experimen-tal results show that,with sensor spacings of 500 m×500 m×50 m,the proposed method achieves a mean three-dimensional lo-calization error of 245.4 m and a root mean square error of 298.5 m,outperforming the MLP,FcCNN,SPCNN and other baseline models.These results demonstrate that the proposed method can achieve stable three-dimensional localization of a single radia-tion source in large-scale complex environments with sparse sensor deployment.

路鑫友;姚昌华

安徽建筑大学电子与信息工程学院,安徽 合肥 230009南京信息工程大学电子与信息工程学院,江苏 南京 210044

信息技术与安全科学

单辐射源三维定位非视距传播主导路径模型深度学习沙漏网络

single radiation source 3D localizationnon-line-of-sight propagation(NLOS)dominant path modeldeep learn-inghourglass network

《计算机与现代化》 2026 (9)

86-92,7

国家自然科学基金资助项目(U22B2002,61971439)通信抗干扰全国重点实验室基础科研创新基金(稳定支持)项目资助(IFN20230207)

10.3969/j.issn.1006-2475.2026.09.010

评论