摘要
针对短波到达时差(time-difference-of-arrival,TDOA)定位问题,提出一套基于多源数据同化和卷积神经网络(convolutional neural network,CNN)重构的电离层虚高修正算法。该方法首先利用卡尔曼滤波对电离层斜向返回探测和全球导航卫星系统(global navigation satellite system,GNSS)层析结果进行数据同化;随后使用CNN代替普通的插值方式对区域电离层参数进行拟合以获取求解TDOA定位问题所需要的时空网格点下的电离层参数;最后利用解析型射线追踪对辐射源和虚高进行一体化估计。以斜测结果进行精度分析,该方法相对于GNSS层析、线性插值、临频峰高半厚的估计精度分别提升了62.93%,64.54%,20.48%。同时,本文利用时间平稳序列检验方法对电离层稳定性进行分析,通过对比短波TDOA定位性能,明确电离层稳定性与定位性能呈明显的正相关性。
Aiming at the short-wave time-difference-of-arrival(TDOA)localization problem,an ionospheric false height correction algorithm based on multi-source data assimilation and convolutional neural network(CNN)reconstruction is proposed.Firstly,this method uses Kalman filtering to assimilate the data from the ionospheric oblique return sounding and global navigation satellite system(GNSS)tomography results.Secondly,the CNN is used instead of ordinary interpolation to obtain the ionospheric parameters under the spatio-temporal grid points required for solving the TDOA localization problem.Finaly,the analytical ray-tracing is used for the integrated estimation of the position of radiation sources and false heights.Based on the oblique measurement results for accuracy analysis,this method has improved the estimation accuracy by 62.93%,64.54%,and 20.48%compared to GNSS tomography,linear interpolation,and peak height half thickness estimation,respectively.Meanwhile,this paper analyzes the stability of the ionosphere by using the time stationary series test method.By comparing the positioning performance of short-wave TDOA,it is clear that there is a significant positive correlation between the stability of the ionosphere and the positioning performance.
作者
孔令宸
刘桐辛
周晨
赵正予
KONG Lingchen;LIU Tongxin;ZHOU Chen;ZHAO Zhengyu(School of Earth and Space Science and Technology,Wuhan University,Wuhan 430072,China;School of Areospace Science,Harbin Institute of Technology(Shenzhen),Shenzhen 518055,China)
出处
《系统工程与电子技术》
北大核心
2025年第11期3521-3530,共10页
Systems Engineering and Electronics
关键词
多源数据同化
卷积神经网络
区域电离层模型
短波到达时差定位
multi-source data assimilation
convolution neural network(CNN)
regional ionospheric model
shortwave time-difference-of-arrival(TDOA)localization
